Datasets:

shrutisingh

/

dataset_recommendation_mcq_mc

like 0

[ "SNIPS", "University-1652", "RVL-CDIP", "DWIE", "NCBI Disease" ]

[ "AFLW", "CelebA" ]

[ { "dkey": "AFLW", "dval": "The Annotated Facial Landmarks in the Wild (AFLW) is a large-scale collection of annotated face images gathered from Flickr, exhibiting a large variety in appearance (e.g., pose, expression, ethnicity, age, gender) as well as general imaging and environmental conditions. In total about 25K faces are annotated with up to 21 landmarks per image." }, { "dkey": "CelebA", "dval": "CelebFaces Attributes dataset contains 202,599 face images of the size 178×218 from 10,177 celebrities, each annotated with 40 binary labels indicating facial attributes like hair color, gender and age." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "University-1652", "dval": "Contains data from three platforms, i.e., synthetic drones, satellites and ground cameras of 1,652 university buildings around the world. University-1652 is a drone-based geo-localization dataset and enables two new tasks, i.e., drone-view target localization and drone navigation." }, { "dkey": "RVL-CDIP", "dval": "The RVL-CDIP dataset consists of scanned document images belonging to 16 classes such as letter, form, email, resume, memo, etc. The dataset has 320,000 training, 40,000 validation and 40,000 test images. The images are characterized by low quality, noise, and low resolution, typically 100 dpi." }, { "dkey": "DWIE", "dval": "The 'Deutsche Welle corpus for Information Extraction' (DWIE) is a multi-task dataset that combines four main Information Extraction (IE) annotation sub-tasks: (i) Named Entity Recognition (NER), (ii) Coreference Resolution, (iii) Relation Extraction (RE), and (iv) Entity Linking. DWIE is conceived as an entity-centric dataset that describes interactions and properties of conceptual entities on the level of the complete document." }, { "dkey": "NCBI Disease", "dval": "The NCBI Disease corpus consists of 793 PubMed abstracts, which are separated into training (593), development (100) and test (100) subsets. The NCBI Disease corpus is annotated with disease mentions, using concept identifiers from either MeSH or OMIM." } ]

[ "Sentence Compression", "SALICON", "Set11", "Epinions", "Friendster", "BVI-DVC" ]

[ "COCO", "CIFAR-10" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "Sentence Compression", "dval": "Sentence Compression is a dataset where the syntactic trees of the compressions are subtrees of their uncompressed counterparts, and hence where supervised systems which require a structural alignment between the input and output can be successfully trained." }, { "dkey": "SALICON", "dval": "The SALIency in CONtext (SALICON) dataset contains 10,000 training images, 5,000 validation images and 5,000 test images for saliency prediction. This dataset has been created by annotating saliency in images from MS COCO.\nThe ground-truth saliency annotations include fixations generated from mouse trajectories. To improve the data quality, isolated fixations with low local density have been excluded.\nThe training and validation sets, provided with ground truth, contain the following data fields: image, resolution and gaze.\nThe testing data contains only the image and resolution fields." }, { "dkey": "Set11", "dval": "Set11 is a dataset of 11 grayscale images. It is a dataset used for image reconstruction and image compression." }, { "dkey": "Epinions", "dval": "The Epinions dataset is built form a who-trust-whom online social network of a general consumer review site Epinions.com. Members of the site can decide whether to ''trust'' each other. All the trust relationships interact and form the Web of Trust which is then combined with review ratings to determine which reviews are shown to the user.\nIt contains 75,879 nodes and 50,8837 edges." }, { "dkey": "Friendster", "dval": "Friendster is an on-line gaming network. Before re-launching as a game website, Friendster was a social networking site where users can form friendship edge each other. Friendster social network also allows users form a group which other members can then join. The Friendster dataset consist of ground-truth communities (based on user-defined groups) and the social network from induced subgraph of the nodes that either belong to at least one community or are connected to other nodes that belong to at least one community." }, { "dkey": "BVI-DVC", "dval": "Contains 800 sequences at various spatial resolutions from 270p to 2160p and has been evaluated on ten existing network architectures for four different coding tools." } ]

[ "ROSTD", "CUFSF", "StreetHazards", "IMDb Movie Reviews", "Covid-HeRA", "VAST", "APRICOT" ]

[ "CIFAR-10", "CelebA" ]

[ { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "CelebA", "dval": "CelebFaces Attributes dataset contains 202,599 face images of the size 178×218 from 10,177 celebrities, each annotated with 40 binary labels indicating facial attributes like hair color, gender and age." }, { "dkey": "ROSTD", "dval": "A dataset of 4K out-of-domain (OOD) examples for the publicly available dataset from (Schuster et al. 2019). In contrast to existing settings which synthesize OOD examples by holding out a subset of classes, the examples were authored by annotators with apriori instructions to be out-of-domain with respect to the sentences in an existing dataset." }, { "dkey": "CUFSF", "dval": "The CUHK Face Sketch FERET (CUFSF) is a dataset for research on face sketch synthesis and face sketch recognition. It contains two types of face images: photo and sketch. Total 1,194 images (one image per subject) were collected with lighting variations from the FERET dataset. For each subject, a sketch is drawn with shape exaggeration." }, { "dkey": "StreetHazards", "dval": "StreetHazards is a synthetic dataset for anomaly detection, created by inserting a diverse array of foreign objects into driving scenes and re-render the scenes with these novel objects." }, { "dkey": "IMDb Movie Reviews", "dval": "The IMDb Movie Reviews dataset is a binary sentiment analysis dataset consisting of 50,000 reviews from the Internet Movie Database (IMDb) labeled as positive or negative. The dataset contains an even number of positive and negative reviews. Only highly polarizing reviews are considered. A negative review has a score ≤ 4 out of 10, and a positive review has a score ≥ 7 out of 10. No more than 30 reviews are included per movie. The dataset contains additional unlabeled data." }, { "dkey": "Covid-HeRA", "dval": "Covid-HeRA is a dataset for health risk assessment and severity-informed decision making in the presence of COVID19 misinformation. It is a benchmark dataset for risk-aware health misinformation detection, related to the 2019 coronavirus pandemic. Social media posts (Twitter) are annotated based on the perceived likelihood of health behavioural changes and the perceived corresponding risks from following unreliable advice found online." }, { "dkey": "VAST", "dval": "VAST consists of a large range of topics covering broad themes, such as politics (e.g., ‘a Palestinian state’), education (e.g., ‘charter schools’), and public health (e.g., ‘childhood vaccination’). In addition, the data includes a wide range of similar expressions (e.g., ‘guns on campus’ versus ‘firearms on campus’). This variation captures how humans might realistically describe the same topic and contrasts with the lack of variation in existing datasets." }, { "dkey": "APRICOT", "dval": "APRICOT is a collection of over 1,000 annotated photographs of printed adversarial patches in public locations. The patches target several object categories for three COCO-trained detection models, and the photos represent natural variation in position, distance, lighting conditions, and viewing angle." } ]

[ "SNIPS", "LAMBADA", "Image and Video Advertisements", "FSDnoisy18k", "COVERAGE" ]

[ "ImageNet", "FFHQ" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "FFHQ", "dval": "Flickr-Faces-HQ (FFHQ) consists of 70,000 high-quality PNG images at 1024×1024 resolution and contains considerable variation in terms of age, ethnicity and image background. It also has good coverage of accessories such as eyeglasses, sunglasses, hats, etc. The images were crawled from Flickr, thus inheriting all the biases of that website, and automatically aligned and cropped using dlib. Only images under permissive licenses were collected. Various automatic filters were used to prune the set, and finally Amazon Mechanical Turk was used to remove the occasional statues, paintings, or photos of photos." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "LAMBADA", "dval": "The LAMBADA (LAnguage Modeling Broadened to Account for Discourse Aspects) benchmark is an open-ended cloze task which consists of about 10,000 passages from BooksCorpus where a missing target word is predicted in the last sentence of each passage. The missing word is constrained to always be the last word of the last sentence and there are no candidate words to choose from. Examples were filtered by humans to ensure they were possible to guess given the context, i.e., the sentences in the passage leading up to the last sentence. Examples were further filtered to ensure that missing words could not be guessed without the context, ensuring that models attempting the dataset would need to reason over the entire paragraph to answer questions." }, { "dkey": "Image and Video Advertisements", "dval": "The Image and Video Advertisements collection consists of an image dataset of 64,832 image ads, and a video dataset of 3,477 ads. The data contains rich annotations encompassing the topic and sentiment of the ads, questions and answers describing what actions the viewer is prompted to take and the reasoning that the ad presents to persuade the viewer (\"What should I do according to this ad, and why should I do it? \"), and symbolic references ads make (e.g. a dove symbolizes peace)." }, { "dkey": "FSDnoisy18k", "dval": "The FSDnoisy18k dataset is an open dataset containing 42.5 hours of audio across 20 sound event classes, including a small amount of manually-labeled data and a larger quantity of real-world noisy data. The audio content is taken from Freesound, and the dataset was curated using the Freesound Annotator. The noisy set of FSDnoisy18k consists of 15,813 audio clips (38.8h), and the test set consists of 947 audio clips (1.4h) with correct labels. The dataset features two main types of label noise: in-vocabulary (IV) and out-of-vocabulary (OOV). IV applies when, given an observed label that is incorrect or incomplete, the true or missing label is part of the target class set. Analogously, OOV means that the true or missing label is not covered by those 20 classes." }, { "dkey": "COVERAGE", "dval": "COVERAGE contains copymove forged (CMFD) images and their originals with similar but genuine objects (SGOs). COVERAGE is designed to highlight and address tamper detection ambiguity of popular methods, caused by self-similarity within natural images. In COVERAGE, forged–original pairs are annotated with (i) the duplicated and forged region masks, and (ii) the tampering factor/similarity metric. For benchmarking, forgery quality is evaluated using (i) computer vision-based methods, and (ii) human detection performance." } ]

[ "DublinCity", "THEODORE", "Waymo Open Dataset", "SemanticKITTI", "JRDB", "KITTI-trajectory-prediction" ]

[ "COCO", "KITTI" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "KITTI", "dval": "KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) is one of the most popular datasets for use in mobile robotics and autonomous driving. It consists of hours of traffic scenarios recorded with a variety of sensor modalities, including high-resolution RGB, grayscale stereo cameras, and a 3D laser scanner. Despite its popularity, the dataset itself does not contain ground truth for semantic segmentation. However, various researchers have manually annotated parts of the dataset to fit their necessities. Álvarez et al. generated ground truth for 323 images from the road detection challenge with three classes: road, vertical, and sky. Zhang et al. annotated 252 (140 for training and 112 for testing) acquisitions – RGB and Velodyne scans – from the tracking challenge for ten object categories: building, sky, road, vegetation, sidewalk, car, pedestrian, cyclist, sign/pole, and fence. Ros et al. labeled 170 training images and 46 testing images (from the visual odometry challenge) with 11 classes: building, tree, sky, car, sign, road, pedestrian, fence, pole, sidewalk, and bicyclist." }, { "dkey": "DublinCity", "dval": "A novel benchmark dataset that includes a manually annotated point cloud for over 260 million laser scanning points into 100'000 (approx.) assets from Dublin LiDAR point cloud [12] in 2015. Objects are labelled into 13 classes using hierarchical levels of detail from large (i.e., building, vegetation and ground) to refined (i.e., window, door and tree) elements." }, { "dkey": "THEODORE", "dval": "Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset." }, { "dkey": "Waymo Open Dataset", "dval": "The Waymo Open Dataset is comprised of high resolution sensor data collected by autonomous vehicles operated by the Waymo Driver in a wide variety of conditions. \n\nThe Waymo Open Dataset currently contains 1,950 segments. The authors plan to grow this dataset in the future. Currently the datasets includes:\n\n\n1,950 segments of 20s each, collected at 10Hz (390,000 frames) in diverse geographies and conditions\nSensor data\n1 mid-range lidar\n4 short-range lidars\n5 cameras (front and sides)\nSynchronized lidar and camera data\nLidar to camera projections\nSensor calibrations and vehicle poses\n\n\nLabeled data\nLabels for 4 object classes - Vehicles, Pedestrians, Cyclists, Signs\nHigh-quality labels for lidar data in 1,200 segments\n12.6M 3D bounding box labels with tracking IDs on lidar data\nHigh-quality labels for camera data in 1,000 segments\n11.8M 2D bounding box labels with tracking IDs on camera data" }, { "dkey": "SemanticKITTI", "dval": "SemanticKITTI is a large-scale outdoor-scene dataset for point cloud semantic segmentation. It is derived from the KITTI Vision Odometry Benchmark which it extends with dense point-wise annotations for the complete 360 field-of-view of the employed automotive LiDAR. The dataset consists of 22 sequences. Overall, the dataset provides 23201 point clouds for training and 20351 for testing." }, { "dkey": "JRDB", "dval": "A novel egocentric dataset collected from social mobile manipulator JackRabbot. The dataset includes 64 minutes of annotated multimodal sensor data including stereo cylindrical 360 degrees RGB video at 15 fps, 3D point clouds from two Velodyne 16 Lidars, line 3D point clouds from two Sick Lidars, audio signal, RGB-D video at 30 fps, 360 degrees spherical image from a fisheye camera and encoder values from the robot's wheels." }, { "dkey": "KITTI-trajectory-prediction", "dval": "KITTI is a well established dataset in the computer vision community. It has often been used for trajectory prediction despite not having a well defined split, generating non comparable baselines in different works. This dataset aims at bridging this gap and proposes a well defined split of the KITTI data.\nSamples are collected as 6 seconds chunks (2seconds for past and 4 for future) in a sliding window fashion from all trajectories in the dataset, including the egovehicle. There are a total of 8613 top-view trajectories for training and 2907 for testing.\nSince top-view maps are not provided by KITTI, semantic labels of static categories obtained with DeepLab-v3+ from all frames are projected in a common top-view map using the Velodyne 3D point cloud and IMU. The resulting maps have a spatial resolution of 0.5 meters and are provided along with the trajectories." } ]

[ "ROCStories", "E2E", "DDD20", "DIPS", "iVQA" ]

[ "CARLA", "KITTI" ]

[ { "dkey": "CARLA", "dval": "CARLA (CAR Learning to Act) is an open simulator for urban driving, developed as an open-source layer over Unreal Engine 4. Technically, it operates similarly to, as an open source layer over Unreal Engine 4 that provides sensors in the form of RGB cameras (with customizable positions), ground truth depth maps, ground truth semantic segmentation maps with 12 semantic classes designed for driving (road, lane marking, traffic sign, sidewalk and so on), bounding boxes for dynamic objects in the environment, and measurements of the agent itself (vehicle location and orientation)." }, { "dkey": "KITTI", "dval": "KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) is one of the most popular datasets for use in mobile robotics and autonomous driving. It consists of hours of traffic scenarios recorded with a variety of sensor modalities, including high-resolution RGB, grayscale stereo cameras, and a 3D laser scanner. Despite its popularity, the dataset itself does not contain ground truth for semantic segmentation. However, various researchers have manually annotated parts of the dataset to fit their necessities. Álvarez et al. generated ground truth for 323 images from the road detection challenge with three classes: road, vertical, and sky. Zhang et al. annotated 252 (140 for training and 112 for testing) acquisitions – RGB and Velodyne scans – from the tracking challenge for ten object categories: building, sky, road, vegetation, sidewalk, car, pedestrian, cyclist, sign/pole, and fence. Ros et al. labeled 170 training images and 46 testing images (from the visual odometry challenge) with 11 classes: building, tree, sky, car, sign, road, pedestrian, fence, pole, sidewalk, and bicyclist." }, { "dkey": "ROCStories", "dval": "ROCStories is a collection of commonsense short stories. The corpus consists of 100,000 five-sentence stories. Each story logically follows everyday topics created by Amazon Mechanical Turk workers. These stories contain a variety of commonsense causal and temporal relations between everyday events. Writers also develop an additional 3,742 Story Cloze Test stories which contain a four-sentence-long body and two candidate endings. The endings were collected by asking Mechanical Turk workers to write both a right ending and a wrong ending after eliminating original endings of given short stories. Both endings were required to make logical sense and include at least one character from the main story line. The published ROCStories dataset is constructed with ROCStories as a training set that includes 98,162 stories that exclude candidate wrong endings, an evaluation set, and a test set, which have the same structure (1 body + 2 candidate endings) and a size of 1,871." }, { "dkey": "E2E", "dval": "End-to-End NLG Challenge (E2E) aims to assess whether recent end-to-end NLG systems can generate more complex output by learning from datasets containing higher lexical richness, syntactic complexity and diverse discourse phenomena." }, { "dkey": "DDD20", "dval": "The dataset was captured with a DAVIS camera that concurrently streams both dynamic vision sensor (DVS) brightness change events and active pixel sensor (APS) intensity frames. DDD20 is the longest event camera end-to-end driving dataset to date with 51h of DAVIS event+frame camera and vehicle human control data collected from 4000km of highway and urban driving under a variety of lighting conditions." }, { "dkey": "DIPS", "dval": "Contains biases but is two orders of magnitude larger than those used previously." }, { "dkey": "iVQA", "dval": "An open-ended VideoQA benchmark that aims to: i) provide a well-defined evaluation by including five correct answer annotations per question and ii) avoid questions which can be answered without the video. \n\niVQA contains 10,000 video clips with one question and five corresponding answers per clip. Moreover, we manually reduce the language bias by excluding questions that could be answered without watching the video." } ]

[ "IIRC", "PHM2017", "CommonsenseQA", "RuBQ" ]

[ "DuReader", "SQuAD" ]

[ { "dkey": "DuReader", "dval": "DuReader is a large-scale open-domain Chinese machine reading comprehension dataset. The dataset consists of 200K questions, 420K answers and 1M documents. The questions and documents are based on Baidu Search and Baidu Zhidao. The answers are manually generated. The dataset additionally provides question type annotations – each question was manually annotated as either Entity, Description or YesNo and one of Fact or Opinion." }, { "dkey": "SQuAD", "dval": "The Stanford Question Answering Dataset (SQuAD) is a collection of question-answer pairs derived from Wikipedia articles. In SQuAD, the correct answers of questions can be any sequence of tokens in the given text. Because the questions and answers are produced by humans through crowdsourcing, it is more diverse than some other question-answering datasets. SQuAD 1.1 contains 107,785 question-answer pairs on 536 articles. SQuAD2.0 (open-domain SQuAD, SQuAD-Open), the latest version, combines the 100,000 questions in SQuAD1.1 with over 50,000 un-answerable questions written adversarially by crowdworkers in forms that are similar to the answerable ones." }, { "dkey": "IIRC", "dval": "Contains more than 13K questions over paragraphs from English Wikipedia that provide only partial information to answer them, with the missing information occurring in one or more linked documents. The questions were written by crowd workers who did not have access to any of the linked documents, leading to questions that have little lexical overlap with the contexts where the answers appear." }, { "dkey": "PHM2017", "dval": "PHM2017 is a new dataset consisting of 7,192 English tweets across six diseases and conditions: Alzheimer’s Disease, heart attack (any severity), Parkinson’s disease, cancer (any type), Depression (any severity), and Stroke. The Twitter search API was used to retrieve the data using the colloquial disease names as search keywords, with the expectation of retrieving a high-recall, low precision dataset. After removing the re-tweets and replies, the tweets were manually annotated. The labels are:\n\n\nself-mention. The tweet contains a health mention with a health self-report of the Twitter account owner, e.g., \"However, I worked hard and ran for Tokyo Mayer Election Campaign in January through February, 2014, without publicizing the cancer.\"\nother-mention. The tweet contains a health mention of a health report about someone other than the account owner, e.g., \"Designer with Parkinson’s couldn’t work then engineer invents bracelet + changes her world\"\nawareness. The tweet contains the disease name, but does not mention a specific person, e.g., \"A Month Before a Heart Attack, Your Body Will Warn You With These 8 Signals\"\nnon-health. The tweet contains the disease name, but the tweet topic is not about health. \"Now I can have cancer on my wall for all to see <3\"" }, { "dkey": "CommonsenseQA", "dval": "The CommonsenseQA is a dataset for commonsense question answering task. The dataset consists of 12,247 questions with 5 choices each.\nThe dataset was generated by Amazon Mechanical Turk workers in the following process (an example is provided in parentheses):\n\n\na crowd worker observes a source concept from ConceptNet (“River”) and three target concepts (“Waterfall”, “Bridge”, “Valley”) that are all related by the same ConceptNet relation (“AtLocation”),\nthe worker authors three questions, one per target concept, such that only that particular target concept is the answer, while the other two distractor concepts are not, (“Where on a river can you hold a cup upright to catch water on a sunny day?”, “Where can I stand on a river to see water falling without getting wet?”, “I’m crossing the river, my feet are wet but my body is dry, where am I?”)\nfor each question, another worker chooses one additional distractor from Concept Net (“pebble”, “stream”, “bank”), and the author another distractor (“mountain”, “bottom”, “island”) manually." }, { "dkey": "RuBQ", "dval": "The first Russian knowledge base question answering (KBQA) dataset. The high-quality dataset consists of 1,500 Russian questions of varying complexity, their English machine translations, SPARQL queries to Wikidata, reference answers, as well as a Wikidata sample of triples containing entities with Russian labels. The dataset creation started with a large collection of question-answer pairs from online quizzes. The data underwent automatic filtering, crowd-assisted entity linking, automatic generation of SPARQL queries, and their subsequent in-house verification." } ]

[ "DublinCity", "OCID", "THEODORE", "ScanObjectNN" ]

[ "ScanNet", "ModelNet" ]

[ { "dkey": "ScanNet", "dval": "ScanNet is an instance-level indoor RGB-D dataset that includes both 2D and 3D data. It is a collection of labeled voxels rather than points or objects. Up to now, ScanNet v2, the newest version of ScanNet, has collected 1513 annotated scans with an approximate 90% surface coverage. In the semantic segmentation task, this dataset is marked in 20 classes of annotated 3D voxelized objects." }, { "dkey": "ModelNet", "dval": "The ModelNet40 dataset contains synthetic object point clouds. As the most widely used benchmark for point cloud analysis, ModelNet40 is popular because of its various categories, clean shapes, well-constructed dataset, etc. The original ModelNet40 consists of 12,311 CAD-generated meshes in 40 categories (such as airplane, car, plant, lamp), of which 9,843 are used for training while the rest 2,468 are reserved for testing. The corresponding point cloud data points are uniformly sampled from the mesh surfaces, and then further preprocessed by moving to the origin and scaling into a unit sphere." }, { "dkey": "DublinCity", "dval": "A novel benchmark dataset that includes a manually annotated point cloud for over 260 million laser scanning points into 100'000 (approx.) assets from Dublin LiDAR point cloud [12] in 2015. Objects are labelled into 13 classes using hierarchical levels of detail from large (i.e., building, vegetation and ground) to refined (i.e., window, door and tree) elements." }, { "dkey": "OCID", "dval": "Developing robot perception systems for handling objects in the real-world requires computer vision algorithms to be carefully scrutinized with respect to the expected operating domain. This demands large quantities of ground truth data to rigorously evaluate the performance of algorithms.\n\nThe Object Cluttered Indoor Dataset is an RGBD-dataset containing point-wise labeled point-clouds for each object. The data was captured using two ASUS-PRO Xtion cameras that are positioned at different heights. It captures diverse settings of objects, background, context, sensor to scene distance, viewpoint angle and lighting conditions. The main purpose of OCID is to allow systematic comparison of existing object segmentation methods in scenes with increasing amount of clutter. In addition OCID does also provide ground-truth data for other vision tasks like object-classification and recognition." }, { "dkey": "THEODORE", "dval": "Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset." }, { "dkey": "ScanObjectNN", "dval": "ScanObjectNN is a newly published real-world dataset comprising of 2902 3D objects in 15 categories. It is a challenging point cloud classification datasets due to the background, missing parts and deformations." } ]

[ "Localized Narratives", "DeeperForensics-1.0", "7-Scenes", "WiderPerson", "KAIST Multispectral Pedestrian Detection Benchmark" ]

[ "AFLW", "Florence" ]

[ { "dkey": "AFLW", "dval": "The Annotated Facial Landmarks in the Wild (AFLW) is a large-scale collection of annotated face images gathered from Flickr, exhibiting a large variety in appearance (e.g., pose, expression, ethnicity, age, gender) as well as general imaging and environmental conditions. In total about 25K faces are annotated with up to 21 landmarks per image." }, { "dkey": "Florence", "dval": "The Florence 3D faces dataset consists of:\n\n\nHigh-resolution 3D scans of human faces from many subjects.\nSeveral video sequences of varying resolution, conditions and zoom level for each subject.\nEach subject is recorded in the following situations:\nIn a controlled setting in HD video.\nIn a less-constrained (but still indoor) setting using a standard, PTZ surveillance camera.\nIn an unconstrained, outdoor environment under challenging recording conditions." }, { "dkey": "Localized Narratives", "dval": "We propose Localized Narratives, a new form of multimodal image annotations connecting vision and language. We ask annotators to describe an image with their voice while simultaneously hovering their mouse over the region they are describing. Since the voice and the mouse pointer are synchronized, we can localize every single word in the description. This dense visual grounding takes the form of a mouse trace segment per word and is unique to our data. We annotated 849k images with Localized Narratives: the whole COCO, Flickr30k, and ADE20K datasets, and 671k images of Open Images, all of which we make publicly available. We provide an extensive analysis of these annotations showing they are diverse, accurate, and efficient to produce. We also demonstrate their utility on the application of controlled image captioning." }, { "dkey": "DeeperForensics-1.0", "dval": "DeeperForensics-1.0 represents the largest face forgery detection dataset by far, with 60,000 videos constituted by a total of 17.6 million frames, 10 times larger than existing datasets of the same kind. The full dataset includes 48,475 source videos and 11,000 manipulated videos. The source videos are collected on 100 paid and consented actors from 26 countries, and the manipulated videos are generated by a newly proposed many-to-many end-to-end face swapping method, DF-VAE. 7 types of real-world perturbations at 5 intensity levels are employed to ensure a larger scale and higher diversity." }, { "dkey": "7-Scenes", "dval": "The 7-Scenes dataset is a collection of tracked RGB-D camera frames. The dataset may be used for evaluation of methods for different applications such as dense tracking and mapping and relocalization techniques.\nAll scenes were recorded from a handheld Kinect RGB-D camera at 640×480 resolution. The dataset creators use an implementation of the KinectFusion system to obtain the ‘ground truth’ camera tracks, and a dense 3D model. Several sequences were recorded per scene by different users, and split into distinct training and testing sequence sets." }, { "dkey": "WiderPerson", "dval": "WiderPerson contains a total of 13,382 images with 399,786 annotations, i.e., 29.87 annotations per image, which means this dataset contains dense pedestrians with various kinds of occlusions. Hence, pedestrians in the proposed dataset are extremely challenging due to large variations in the scenario and occlusion, which is suitable to evaluate pedestrian detectors in the wild." }, { "dkey": "KAIST Multispectral Pedestrian Detection Benchmark", "dval": "KAIST Multispectral Pedestrian Dataset\n\nThe KAIST Multispectral Pedestrian Dataset is imaging hardware consisting of a color camera, a thermal camera and a beam splitter to capture the aligned multispectral (RGB color + Thermal) images. With this hardware, we captured various regular traffic scenes at day and night time to consider changes in light conditions. and, consists of 95k color-thermal pairs (640x480, 20Hz) taken from a vehicle. All the pairs are manually annotated (person, people, cyclist) for the total of 103,128 dense annotations and 1,182 unique pedestrians. The annotation includes temporal correspondence between bounding boxes like Caltech Pedestrian Dataset.\n\nFor more information, read Multispectral Pedestrian Detection: Benchmark Dataset and Baseline (CVPR 2015) or visit this website" } ]

[ "VisDial", "EPIC-KITCHENS-100", "TDIUC", "Localized Narratives" ]

[ "COCO", "Flickr30k" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "Flickr30k", "dval": "The Flickr30k dataset contains 31,000 images collected from Flickr, together with 5 reference sentences provided by human annotators." }, { "dkey": "VisDial", "dval": "Visual Dialog (VisDial) dataset contains human annotated questions based on images of MS COCO dataset. This dataset was developed by pairing two subjects on Amazon Mechanical Turk to chat about an image. One person was assigned the job of a ‘questioner’ and the other person acted as an ‘answerer’. The questioner sees only the text description of an image (i.e., an image caption from MS COCO dataset) and the original image remains hidden to the questioner. Their task is to ask questions about this hidden image to “imagine the scene better”. The answerer sees the image, caption and answers the questions asked by the questioner. The two of them can continue the conversation by asking and answering questions for 10 rounds at max.\n\nVisDial v1.0 contains 123K dialogues on MS COCO (2017 training set) for training split, 2K dialogues with validation images for validation split and 8K dialogues on test set for test-standard set. The previously released v0.5 and v0.9 versions of VisDial dataset (corresponding to older splits of MS COCO) are considered deprecated." }, { "dkey": "EPIC-KITCHENS-100", "dval": "This paper introduces the pipeline to scale the largest dataset in egocentric vision EPIC-KITCHENS. The effort culminates in EPIC-KITCHENS-100, a collection of 100 hours, 20M frames, 90K actions in 700 variable-length videos, capturing long-term unscripted activities in 45 environments, using head-mounted cameras. Compared to its previous version (EPIC-KITCHENS-55), EPIC-KITCHENS-100 has been annotated using a novel pipeline that allows denser (54% more actions per minute) and more complete annotations of fine-grained actions (+128% more action segments). This collection also enables evaluating the \"test of time\" - i.e. whether models trained on data collected in 2018 can generalise to new footage collected under the same hypotheses albeit \"two years on\".\nThe dataset is aligned with 6 challenges: action recognition (full and weak supervision), action detection, action anticipation, cross-modal retrieval (from captions), as well as unsupervised domain adaptation for action recognition. For each challenge, we define the task, provide baselines and evaluation metrics." }, { "dkey": "TDIUC", "dval": "Task Directed Image Understanding Challenge (TDIUC) dataset is a Visual Question Answering dataset which consists of 1.6M questions and 170K images sourced from MS COCO and the Visual Genome Dataset. The image-question pairs are split into 12 categories and 4 additional evaluation matrices which help evaluate models’ robustness against answer imbalance and its ability to answer questions that require higher reasoning capability. The TDIUC dataset divides the VQA paradigm into 12 different task directed question types. These include questions that require a simpler task (e.g., object presence, color attribute) and more complex tasks (e.g., counting, positional reasoning). The dataset includes also an “Absurd” question category in which questions are irrelevant to the image contents to help balance the dataset." }, { "dkey": "Localized Narratives", "dval": "We propose Localized Narratives, a new form of multimodal image annotations connecting vision and language. We ask annotators to describe an image with their voice while simultaneously hovering their mouse over the region they are describing. Since the voice and the mouse pointer are synchronized, we can localize every single word in the description. This dense visual grounding takes the form of a mouse trace segment per word and is unique to our data. We annotated 849k images with Localized Narratives: the whole COCO, Flickr30k, and ADE20K datasets, and 671k images of Open Images, all of which we make publicly available. We provide an extensive analysis of these annotations showing they are diverse, accurate, and efficient to produce. We also demonstrate their utility on the application of controlled image captioning." } ]

[ "CoQA", "TQA", "TechQA", "CommonsenseQA", "iVQA", "VizWiz" ]

[ "COCO", "CLEVR" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "CLEVR", "dval": "CLEVR (Compositional Language and Elementary Visual Reasoning) is a synthetic Visual Question Answering dataset. It contains images of 3D-rendered objects; each image comes with a number of highly compositional questions that fall into different categories. Those categories fall into 5 classes of tasks: Exist, Count, Compare Integer, Query Attribute and Compare Attribute. The CLEVR dataset consists of: a training set of 70k images and 700k questions, a validation set of 15k images and 150k questions, A test set of 15k images and 150k questions about objects, answers, scene graphs and functional programs for all train and validation images and questions. Each object present in the scene, aside of position, is characterized by a set of four attributes: 2 sizes: large, small, 3 shapes: square, cylinder, sphere, 2 material types: rubber, metal, 8 color types: gray, blue, brown, yellow, red, green, purple, cyan, resulting in 96 unique combinations." }, { "dkey": "CoQA", "dval": "CoQA is a large-scale dataset for building Conversational Question Answering systems. The goal of the CoQA challenge is to measure the ability of machines to understand a text passage and answer a series of interconnected questions that appear in a conversation.\n\nCoQA contains 127,000+ questions with answers collected from 8000+ conversations. Each conversation is collected by pairing two crowdworkers to chat about a passage in the form of questions and answers. The unique features of CoQA include 1) the questions are conversational; 2) the answers can be free-form text; 3) each answer also comes with an evidence subsequence highlighted in the passage; and 4) the passages are collected from seven diverse domains. CoQA has a lot of challenging phenomena not present in existing reading comprehension datasets, e.g., coreference and pragmatic reasoning." }, { "dkey": "TQA", "dval": "The TextbookQuestionAnswering (TQA) dataset is drawn from middle school science curricula. It consists of 1,076 lessons from Life Science, Earth Science and Physical Science textbooks. This includes 26,260 questions, including 12,567 that have an accompanying diagram.\n\nThe TQA dataset encourages work on the task of Multi-Modal Machine Comprehension (M3C) task. The M3C task builds on the popular Visual Question Answering (VQA) and Machine Comprehension (MC) paradigms by framing question answering as a machine comprehension task, where the context needed to answer questions is provided and composed of both text and images. The dataset constructed to showcase this task has been built from a middle school science curriculum that pairs a given question to a limited span of knowledge needed to answer it." }, { "dkey": "TechQA", "dval": "TECHQA is a domain-adaptation question answering dataset for the technical support domain. The TECHQA corpus highlights two real-world issues from the automated customer support domain. First, it contains actual questions posed by users on a technical forum, rather than questions generated specifically for a competition or a task. Second, it has a real-world size – 600 training, 310 dev, and 490 evaluation question/answer pairs – thus reflecting the cost of creating large labeled datasets with actual data. Consequently, TECHQA is meant to stimulate research in domain adaptation rather than being a resource to build QA systems from scratch. The dataset was obtained by crawling the IBM Developer and IBM DeveloperWorks forums for questions with accepted answers that appear in a published IBM Technote—a technical document that addresses a specific technical issue." }, { "dkey": "CommonsenseQA", "dval": "The CommonsenseQA is a dataset for commonsense question answering task. The dataset consists of 12,247 questions with 5 choices each.\nThe dataset was generated by Amazon Mechanical Turk workers in the following process (an example is provided in parentheses):\n\n\na crowd worker observes a source concept from ConceptNet (“River”) and three target concepts (“Waterfall”, “Bridge”, “Valley”) that are all related by the same ConceptNet relation (“AtLocation”),\nthe worker authors three questions, one per target concept, such that only that particular target concept is the answer, while the other two distractor concepts are not, (“Where on a river can you hold a cup upright to catch water on a sunny day?”, “Where can I stand on a river to see water falling without getting wet?”, “I’m crossing the river, my feet are wet but my body is dry, where am I?”)\nfor each question, another worker chooses one additional distractor from Concept Net (“pebble”, “stream”, “bank”), and the author another distractor (“mountain”, “bottom”, “island”) manually." }, { "dkey": "iVQA", "dval": "An open-ended VideoQA benchmark that aims to: i) provide a well-defined evaluation by including five correct answer annotations per question and ii) avoid questions which can be answered without the video. \n\niVQA contains 10,000 video clips with one question and five corresponding answers per clip. Moreover, we manually reduce the language bias by excluding questions that could be answered without watching the video." }, { "dkey": "VizWiz", "dval": "The VizWiz-VQA dataset originates from a natural visual question answering setting where blind people each took an image and recorded a spoken question about it, together with 10 crowdsourced answers per visual question. The proposed challenge addresses the following two tasks for this dataset: predict the answer to a visual question and (2) predict whether a visual question cannot be answered." } ]

[ "YouTube-8M", "CLUECorpus2020", "ConvAI2", "COVERAGE", "SNIPS", "Twitter100k", "Word Sense Disambiguation: a Unified Evaluation Framework and Empirical Comparison" ]

[ "ImageNet", "COCO", "VRD" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "VRD", "dval": "The Visual Relationship Dataset (VRD) contains 4000 images for training and 1000 for testing annotated with visual relationships. Bounding boxes are annotated with a label containing 100 unary predicates. These labels refer to animals, vehicles, clothes and generic objects. Pairs of bounding boxes are annotated with a label containing 70 binary predicates. These labels refer to actions, prepositions, spatial relations, comparatives or preposition phrases. The dataset has 37993 instances of visual relationships and 6672 types of relationships. 1877 instances of relationships occur only in the test set and they are used to evaluate the zero-shot learning scenario." }, { "dkey": "YouTube-8M", "dval": "The YouTube-8M dataset is a large scale video dataset, which includes more than 7 million videos with 4716 classes labeled by the annotation system. The dataset consists of three parts: training set, validate set, and test set. In the training set, each class contains at least 100 training videos. Features of these videos are extracted by the state-of-the-art popular pre-trained models and released for public use. Each video contains audio and visual modality. Based on the visual information, videos are divided into 24 topics, such as sports, game, arts & entertainment, etc" }, { "dkey": "CLUECorpus2020", "dval": "CLUECorpus2020 is a large-scale corpus that can be used directly for self-supervised learning such as pre-training of a language model, or language generation. It has 100G raw corpus with 35 billion Chinese characters, which is retrieved from Common Crawl." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "COVERAGE", "dval": "COVERAGE contains copymove forged (CMFD) images and their originals with similar but genuine objects (SGOs). COVERAGE is designed to highlight and address tamper detection ambiguity of popular methods, caused by self-similarity within natural images. In COVERAGE, forged–original pairs are annotated with (i) the duplicated and forged region masks, and (ii) the tampering factor/similarity metric. For benchmarking, forgery quality is evaluated using (i) computer vision-based methods, and (ii) human detection performance." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "Twitter100k", "dval": "Twitter100k is a large-scale dataset for weakly supervised cross-media retrieval." }, { "dkey": "Word Sense Disambiguation: a Unified Evaluation Framework and Empirical Comparison", "dval": "The Evaluation framework of Raganato et al. 2017 includes two training sets (SemCor-Miller et al., 1993- and OMSTI-Taghipour and Ng, 2015-) and five test sets from the Senseval/SemEval series (Edmonds and Cotton, 2001; Snyder and Palmer, 2004; Pradhan et al., 2007; Navigli et al., 2013; Moro and Navigli, 2015), standardized to the same format and sense inventory (i.e. WordNet 3.0).\n\nTypically, there are two kinds of approach for WSD: supervised (which make use of sense-annotated training data) and knowledge-based (which make use of the properties of lexical resources).\n\nSupervised: The most widely used training corpus used is SemCor, with 226,036 sense annotations from 352 documents manually annotated. All supervised systems in the evaluation table are trained on SemCor. Some supervised methods, particularly neural architectures, usually employ the SemEval 2007 dataset as development set (marked by *). The most usual baseline is the Most Frequent Sense (MFS) heuristic, which selects for each target word the most frequent sense in the training data.\n\nKnowledge-based: Knowledge-based systems usually exploit WordNet or BabelNet as semantic network. The first sense given by the underlying sense inventory (i.e. WordNet 3.0) is included as a baseline.\n\nDescription from NLP Progress" } ]

[ "SuperGLUE", "CSQA", "SimpleQuestions", "VizWiz" ]

[ "WikiQA", "SQuAD" ]

[ { "dkey": "WikiQA", "dval": "The WikiQA corpus is a publicly available set of question and sentence pairs, collected and annotated for research on open-domain question answering. In order to reflect the true information need of general users, Bing query logs were used as the question source. Each question is linked to a Wikipedia page that potentially has the answer. Because the summary section of a Wikipedia page provides the basic and usually most important information about the topic, sentences in this section were used as the candidate answers. The corpus includes 3,047 questions and 29,258 sentences, where 1,473 sentences were labeled as answer sentences to their corresponding questions." }, { "dkey": "SQuAD", "dval": "The Stanford Question Answering Dataset (SQuAD) is a collection of question-answer pairs derived from Wikipedia articles. In SQuAD, the correct answers of questions can be any sequence of tokens in the given text. Because the questions and answers are produced by humans through crowdsourcing, it is more diverse than some other question-answering datasets. SQuAD 1.1 contains 107,785 question-answer pairs on 536 articles. SQuAD2.0 (open-domain SQuAD, SQuAD-Open), the latest version, combines the 100,000 questions in SQuAD1.1 with over 50,000 un-answerable questions written adversarially by crowdworkers in forms that are similar to the answerable ones." }, { "dkey": "SuperGLUE", "dval": "SuperGLUE is a benchmark dataset designed to pose a more rigorous test of language understanding than GLUE. SuperGLUE has the same high-level motivation as GLUE: to provide a simple, hard-to-game measure of progress toward general-purpose language understanding technologies for English. SuperGLUE follows the basic design of GLUE: It consists of a public leaderboard built around eight language understanding tasks, drawing on existing data, accompanied by a single-number\nperformance metric, and an analysis toolkit. However, it improves upon GLUE in several ways:\n\n\nMore challenging tasks: SuperGLUE retains the two hardest tasks in GLUE. The remaining tasks were identified from those submitted to an open call for task proposals and were selected based on difficulty for current NLP approaches.\nMore diverse task formats: The task formats in GLUE are limited to sentence- and sentence-pair classification. The authors expand the set of task formats in SuperGLUE to include\ncoreference resolution and question answering (QA).\nComprehensive human baselines: the authors include human performance estimates for all benchmark tasks, which verify that substantial headroom exists between a strong BERT-based baseline and human performance.\nImproved code support: SuperGLUE is distributed with a new, modular toolkit for work on pretraining, multi-task learning, and transfer learning in NLP, built around standard tools including PyTorch (Paszke et al., 2017) and AllenNLP (Gardner et al., 2017).\nRefined usage rules: The conditions for inclusion on the SuperGLUE leaderboard were revamped to ensure fair competition, an informative leaderboard, and full credit\nassignment to data and task creators." }, { "dkey": "CSQA", "dval": "Contains around 200K dialogs with a total of 1.6M turns. Further, unlike existing large scale QA datasets which contain simple questions that can be answered from a single tuple, the questions in the dialogs require a larger subgraph of the KG." }, { "dkey": "SimpleQuestions", "dval": "SimpleQuestions is a large-scale factoid question answering dataset. It consists of 108,442 natural language questions, each paired with a corresponding fact from Freebase knowledge base. Each fact is a triple (subject, relation, object) and the answer to the question is always the object. The dataset is divided into training, validation, and test sets with 75,910, 10,845 and 21,687 questions respectively." }, { "dkey": "VizWiz", "dval": "The VizWiz-VQA dataset originates from a natural visual question answering setting where blind people each took an image and recorded a spoken question about it, together with 10 crowdsourced answers per visual question. The proposed challenge addresses the following two tasks for this dataset: predict the answer to a visual question and (2) predict whether a visual question cannot be answered." } ]

[ "HotpotQA", "MultiReQA", "OpenBookQA", "XQA" ]

[ "SQuAD", "WebQuestions", "TriviaQA" ]

[ { "dkey": "SQuAD", "dval": "The Stanford Question Answering Dataset (SQuAD) is a collection of question-answer pairs derived from Wikipedia articles. In SQuAD, the correct answers of questions can be any sequence of tokens in the given text. Because the questions and answers are produced by humans through crowdsourcing, it is more diverse than some other question-answering datasets. SQuAD 1.1 contains 107,785 question-answer pairs on 536 articles. SQuAD2.0 (open-domain SQuAD, SQuAD-Open), the latest version, combines the 100,000 questions in SQuAD1.1 with over 50,000 un-answerable questions written adversarially by crowdworkers in forms that are similar to the answerable ones." }, { "dkey": "WebQuestions", "dval": "The WebQuestions dataset is a question answering dataset using Freebase as the knowledge base and contains 6,642 question-answer pairs. It was created by crawling questions through the Google Suggest API, and then obtaining answers using Amazon Mechanical Turk. The original split uses 3,778 examples for training and 2,032 for testing. All answers are defined as Freebase entities.\n\nExample questions (answers) in the dataset include “Where did Edgar Allan Poe died?” (baltimore) or “What degrees did Barack Obama get?” (bachelor_of_arts, juris_doctor)." }, { "dkey": "TriviaQA", "dval": "TriviaQA is a realistic text-based question answering dataset which includes 950K question-answer pairs from 662K documents collected from Wikipedia and the web. This dataset is more challenging than standard QA benchmark datasets such as Stanford Question Answering Dataset (SQuAD), as the answers for a question may not be directly obtained by span prediction and the context is very long. TriviaQA dataset consists of both human-verified and machine-generated QA subsets." }, { "dkey": "HotpotQA", "dval": "HotpotQA is a question answering dataset collected on the English Wikipedia, containing about 113K crowd-sourced questions that are constructed to require the introduction paragraphs of two Wikipedia articles to answer. Each question in the dataset comes with the two gold paragraphs, as well as a list of sentences in these paragraphs that crowdworkers identify as supporting facts necessary to answer the question. \n\nA diverse range of reasoning strategies are featured in HotpotQA, including questions involving missing entities in the question, intersection questions (What satisfies property A and property B?), and comparison questions, where two entities are compared by a common attribute, among others. In the few-document distractor setting, the QA models are given ten paragraphs in which the gold paragraphs are guaranteed to be found; in the open-domain fullwiki setting, the models are only given the question and the entire Wikipedia. Models are evaluated on their answer accuracy and explainability, where the former is measured as overlap between the predicted and gold answers with exact match (EM) and unigram F1, and the latter concerns how well the predicted supporting fact sentences match human annotation (Supporting Fact EM/F1). A joint metric is also reported on this dataset, which encourages systems to perform well on both tasks simultaneously." }, { "dkey": "MultiReQA", "dval": "MultiReQA is a cross-domain evaluation for retrieval question answering models. Retrieval question answering (ReQA) is the task of retrieving a sentence-level answer to a question from an open corpus. MultiReQA is a new multi-domain ReQA evaluation suite composed of eight retrieval QA tasks drawn from publicly available QA datasets from the MRQA shared task.\nMultiReQA contains the sentence boundary annotation from eight publicly available QA datasets including SearchQA, TriviaQA, HotpotQA, NaturalQuestions, SQuAD, BioASQ, RelationExtraction, and TextbookQA. Five of these datasets, including SearchQA, TriviaQA, HotpotQA, NaturalQuestions, SQuAD, contain both training and test data, and three, in cluding BioASQ, RelationExtraction, TextbookQA, contain only the test data." }, { "dkey": "OpenBookQA", "dval": "OpenBookQA is a new kind of question-answering dataset modeled after open book exams for assessing human understanding of a subject. It consists of 5,957 multiple-choice elementary-level science questions (4,957 train, 500 dev, 500 test), which probe the understanding of a small “book” of 1,326 core science facts and the application of these facts to novel situations. For training, the dataset includes a mapping from each question to the core science fact it was designed to probe. Answering OpenBookQA questions requires additional broad common knowledge, not contained in the book. The questions, by design, are answered incorrectly by both a retrieval-based algorithm and a word co-occurrence algorithm.\nAdditionally, the dataset includes a collection of 5,167 crowd-sourced common knowledge facts, and an expanded version of the train/dev/test questions where each question is associated with its originating core fact, a human accuracy score, a clarity score, and an anonymized crowd-worker ID." }, { "dkey": "XQA", "dval": "XQA is a data which consists of a total amount of 90k question-answer pairs in nine languages for cross-lingual open-domain question answering." } ]

[ "DDD20", "EyeCar", "DIPS", "WikiReading", "iSUN", "Jamendo Lyrics" ]

[ "KITTI", "Cityscapes" ]

[ { "dkey": "KITTI", "dval": "KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) is one of the most popular datasets for use in mobile robotics and autonomous driving. It consists of hours of traffic scenarios recorded with a variety of sensor modalities, including high-resolution RGB, grayscale stereo cameras, and a 3D laser scanner. Despite its popularity, the dataset itself does not contain ground truth for semantic segmentation. However, various researchers have manually annotated parts of the dataset to fit their necessities. Álvarez et al. generated ground truth for 323 images from the road detection challenge with three classes: road, vertical, and sky. Zhang et al. annotated 252 (140 for training and 112 for testing) acquisitions – RGB and Velodyne scans – from the tracking challenge for ten object categories: building, sky, road, vegetation, sidewalk, car, pedestrian, cyclist, sign/pole, and fence. Ros et al. labeled 170 training images and 46 testing images (from the visual odometry challenge) with 11 classes: building, tree, sky, car, sign, road, pedestrian, fence, pole, sidewalk, and bicyclist." }, { "dkey": "Cityscapes", "dval": "Cityscapes is a large-scale database which focuses on semantic understanding of urban street scenes. It provides semantic, instance-wise, and dense pixel annotations for 30 classes grouped into 8 categories (flat surfaces, humans, vehicles, constructions, objects, nature, sky, and void). The dataset consists of around 5000 fine annotated images and 20000 coarse annotated ones. Data was captured in 50 cities during several months, daytimes, and good weather conditions. It was originally recorded as video so the frames were manually selected to have the following features: large number of dynamic objects, varying scene layout, and varying background." }, { "dkey": "DDD20", "dval": "The dataset was captured with a DAVIS camera that concurrently streams both dynamic vision sensor (DVS) brightness change events and active pixel sensor (APS) intensity frames. DDD20 is the longest event camera end-to-end driving dataset to date with 51h of DAVIS event+frame camera and vehicle human control data collected from 4000km of highway and urban driving under a variety of lighting conditions." }, { "dkey": "EyeCar", "dval": "EyeCar is a dataset of driving videos of vehicles involved in rear-end collisions paired with eye fixation data captured from human subjects. It contains 21 front-view videos that were captured in various traffic, weather, and day light conditions. Each video is 30sec in length and contains typical driving tasks (e.g., lanekeeping, merging-in, and braking) ending to rear-end collisions." }, { "dkey": "DIPS", "dval": "Contains biases but is two orders of magnitude larger than those used previously." }, { "dkey": "WikiReading", "dval": "WikiReading is a large-scale natural language understanding task and publicly-available dataset with 18 million instances. The task is to predict textual values from the structured knowledge base Wikidata by reading the text of the corresponding Wikipedia articles. The task contains a rich variety of challenging classification and extraction sub-tasks, making it well-suited for end-to-end models such as deep neural networks (DNNs)." }, { "dkey": "iSUN", "dval": "iSUN is a ground truth of gaze traces on images from the SUN dataset. The collection is partitioned into 6,000 images for training, 926 for validation and 2,000 for test." }, { "dkey": "Jamendo Lyrics", "dval": "Dataset for lyrics alignment and transcription evaluation. It contains 20 music pieces under CC license from the Jamendo website along with their lyrics, with:\n\n\nManual annotations indicating the start time of each word in the audio file\nPredictions of start and end times for each word from both of the models presented in the paper" } ]

[ "BLURB", "BL30K", "GGPONC", "MeDAL", "SuperGLUE", "C4" ]

[ "MRPC", "GLUE" ]

[ { "dkey": "MRPC", "dval": "Microsoft Research Paraphrase Corpus (MRPC) is a corpus consists of 5,801 sentence pairs collected from newswire articles. Each pair is labelled if it is a paraphrase or not by human annotators. The whole set is divided into a training subset (4,076 sentence pairs of which 2,753 are paraphrases) and a test subset (1,725 pairs of which 1,147 are paraphrases)." }, { "dkey": "GLUE", "dval": "General Language Understanding Evaluation (GLUE) benchmark is a collection of nine natural language understanding tasks, including single-sentence tasks CoLA and SST-2, similarity and paraphrasing tasks MRPC, STS-B and QQP, and natural language inference tasks MNLI, QNLI, RTE and WNLI." }, { "dkey": "BLURB", "dval": "BLURB is a collection of resources for biomedical natural language processing. In general domains such as newswire and the Web, comprehensive benchmarks and leaderboards such as GLUE have greatly accelerated progress in open-domain NLP. In biomedicine, however, such resources are ostensibly scarce. In the past, there have been a plethora of shared tasks in biomedical NLP, such as BioCreative, BioNLP Shared Tasks, SemEval, and BioASQ, to name just a few. These efforts have played a significant role in fueling interest and progress by the research community, but they typically focus on individual tasks. The advent of neural language models such as BERTs provides a unifying foundation to leverage transfer learning from unlabeled text to support a wide range of NLP applications. To accelerate progress in biomedical pretraining strategies and task-specific methods, it is thus imperative to create a broad-coverage benchmark encompassing diverse biomedical tasks.\n\nInspired by prior efforts toward this direction (e.g., BLUE), BLURB (short for Biomedical Language Understanding and Reasoning Benchmark) was created. BLURB comprises of a comprehensive benchmark for PubMed-based biomedical NLP applications, as well as a leaderboard for tracking progress by the community. BLURB includes thirteen publicly available datasets in six diverse tasks. To avoid placing undue emphasis on tasks with many available datasets, such as named entity recognition (NER), BLURB reports the macro average across all tasks as the main score. The BLURB leaderboard is model-agnostic. Any system capable of producing the test predictions using the same training and development data can participate. The main goal of BLURB is to lower the entry barrier in biomedical NLP and help accelerate progress in this vitally important field for positive societal and human impact." }, { "dkey": "BL30K", "dval": "BL30K is a synthetic dataset rendered using Blender with ShapeNet's data. We break the dataset into six segments, each with approximately 5K videos. The videos are organized in a similar format as DAVIS and YouTubeVOS, so dataloaders for those datasets can be used directly. Each video is 160 frames long, and each frame has a resolution of 768*512. There are 3-5 objects per video, and each object has a random smooth trajectory -- we tried to optimize the trajectories in a greedy fashion to minimize object intersection (not guaranteed), with occlusions still possible (happen a lot in reality). See MiVOS for details." }, { "dkey": "GGPONC", "dval": "German Guideline Program in Oncology NLP Corpus (GGPONC) is a German language corpus based on clinical practice guidelines for oncology. This corpus is one of the largest ever built from German medical documents. Unlike clinical documents, clinical guidelines do not contain any patient-related information and can therefore be used without data protection restrictions." }, { "dkey": "MeDAL", "dval": "The Medical Dataset for Abbreviation Disambiguation for Natural Language Understanding (MeDAL) is a large medical text dataset curated for abbreviation disambiguation, designed for natural language understanding pre-training in the medical domain. It was published at the ClinicalNLP workshop at EMNLP." }, { "dkey": "SuperGLUE", "dval": "SuperGLUE is a benchmark dataset designed to pose a more rigorous test of language understanding than GLUE. SuperGLUE has the same high-level motivation as GLUE: to provide a simple, hard-to-game measure of progress toward general-purpose language understanding technologies for English. SuperGLUE follows the basic design of GLUE: It consists of a public leaderboard built around eight language understanding tasks, drawing on existing data, accompanied by a single-number\nperformance metric, and an analysis toolkit. However, it improves upon GLUE in several ways:\n\n\nMore challenging tasks: SuperGLUE retains the two hardest tasks in GLUE. The remaining tasks were identified from those submitted to an open call for task proposals and were selected based on difficulty for current NLP approaches.\nMore diverse task formats: The task formats in GLUE are limited to sentence- and sentence-pair classification. The authors expand the set of task formats in SuperGLUE to include\ncoreference resolution and question answering (QA).\nComprehensive human baselines: the authors include human performance estimates for all benchmark tasks, which verify that substantial headroom exists between a strong BERT-based baseline and human performance.\nImproved code support: SuperGLUE is distributed with a new, modular toolkit for work on pretraining, multi-task learning, and transfer learning in NLP, built around standard tools including PyTorch (Paszke et al., 2017) and AllenNLP (Gardner et al., 2017).\nRefined usage rules: The conditions for inclusion on the SuperGLUE leaderboard were revamped to ensure fair competition, an informative leaderboard, and full credit\nassignment to data and task creators." }, { "dkey": "C4", "dval": "C4 is a colossal, cleaned version of Common Crawl's web crawl corpus. It was based on Common Crawl dataset: https://commoncrawl.org. It was used to train the T5 text-to-text Transformer models.\n\nThe dataset can be downloaded in a pre-processed form from allennlp." } ]

[ "SNIPS", "COCO-Tasks", "Open Images V4", "ConvAI2", "FaceForensics" ]

[ "ImageNet", "Food-101", "WebVision" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "Food-101", "dval": "The Food-101 dataset consists of 101 food categories with 750 training and 250 test images per category, making a total of 101k images. The labels for the test images have been manually cleaned, while the training set contains some noise." }, { "dkey": "WebVision", "dval": "The WebVision dataset is designed to facilitate the research on learning visual representation from noisy web data. It is a large scale web images dataset that contains more than 2.4 million of images crawled from the Flickr website and Google Images search. \n\nThe same 1,000 concepts as the ILSVRC 2012 dataset are used for querying images, such that a bunch of existing approaches can be directly investigated and compared to the models trained from the ILSVRC 2012 dataset, and also makes it possible to study the dataset bias issue in the large scale scenario. The textual information accompanied with those images (e.g., caption, user tags, or description) are also provided as additional meta information. A validation set contains 50,000 images (50 images per category) is provided to facilitate the algorithmic development." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "COCO-Tasks", "dval": "Comprises about 40,000 images where the most suitable objects for 14 tasks have been annotated." }, { "dkey": "Open Images V4", "dval": "Open Images V4 offers large scale across several dimensions: 30.1M image-level labels for 19.8k concepts, 15.4M bounding boxes for 600 object classes, and 375k visual relationship annotations involving 57 classes. For object detection in particular, 15x more bounding boxes than the next largest datasets (15.4M boxes on 1.9M images) are provided. The images often show complex scenes with several objects (8 annotated objects per image on average). Visual relationships between them are annotated, which support visual relationship detection, an emerging task that requires structured reasoning." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "FaceForensics", "dval": "FaceForensics is a video dataset consisting of more than 500,000 frames containing faces from 1004 videos that can be used to study image or video forgeries. All videos are downloaded from Youtube and are cut down to short continuous clips that contain mostly frontal faces. This dataset has two versions:\n\n\n\nSource-to-Target: where the authors reenact over 1000 videos with new facial expressions extracted from other videos, which e.g. can be used to train a classifier to detect fake images or videos.\n\n\n\nSelfreenactment: where the authors use Face2Face to reenact the facial expressions of videos with their own facial expressions as input to get pairs of videos, which e.g. can be used to train supervised generative refinement models." } ]

[ "DailyDialog++", "SNIPS", "Syn2Real", "Cumulo", "ACDC" ]

[ "ImageNet", "CIFAR-10" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "DailyDialog++", "dval": "Consists of (i) five relevant responses for each context and (ii) five adversarially crafted irrelevant responses for each context." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "Syn2Real", "dval": "Syn2Real, a synthetic-to-real visual domain adaptation benchmark meant to encourage further development of robust domain transfer methods. The goal is to train a model on a synthetic \"source\" domain and then update it so that its performance improves on a real \"target\" domain, without using any target annotations. It includes three tasks, illustrated in figures above: the more traditional closed-set classification task with a known set of categories; the less studied open-set classification task with unknown object categories in the target domain; and the object detection task, which involves localizing instances of objects by predicting their bounding boxes and corresponding class labels." }, { "dkey": "Cumulo", "dval": "A benchmark dataset for training and evaluating global cloud classification models. It consists of one year of 1km resolution MODIS hyperspectral imagery merged with pixel-width 'tracks' of CloudSat cloud labels." }, { "dkey": "ACDC", "dval": "The goal of the Automated Cardiac Diagnosis Challenge (ACDC) challenge is to:\n\n\ncompare the performance of automatic methods on the segmentation of the left ventricular endocardium and epicardium as the right ventricular endocardium for both end diastolic and end systolic phase instances;\ncompare the performance of automatic methods for the classification of the examinations in five classes (normal case, heart failure with infarction, dilated cardiomyopathy, hypertrophic cardiomyopathy, abnormal right ventricle).\n\nThe overall ACDC dataset was created from real clinical exams acquired at the University Hospital of Dijon. Acquired data were fully anonymized and handled within the regulations set by the local ethical committee of the Hospital of Dijon (France). Our dataset covers several well-defined pathologies with enough cases to (1) properly train machine learning methods and (2) clearly assess the variations of the main physiological parameters obtained from cine-MRI (in particular diastolic volume and ejection fraction). The dataset is composed of 150 exams (all from different patients) divided into 5 evenly distributed subgroups (4 pathological plus 1 healthy subject groups) as described below. Furthermore, each patient comes with the following additional information : weight, height, as well as the diastolic and systolic phase instants.\n\nThe database is made available to participants through two datasets from the dedicated online evaluation website after a personal registration: i) a training dataset of 100 patients along with the corresponding manual references based on the analysis of one clinical expert; ii) a testing dataset composed of 50 new patients, without manual annotations but with the patient information given above. The raw input images are provided through the Nifti format." } ]

[ "MARS", "Cluttered Omniglot", "NYU-VP", "Dialogue State Tracking Challenge", "PASCAL-5i" ]

[ "DAVIS", "GOT-10k", "TrackingNet", "VOT2018" ]

[ { "dkey": "DAVIS", "dval": "The Densely Annotation Video Segmentation dataset (DAVIS) is a high quality and high resolution densely annotated video segmentation dataset under two resolutions, 480p and 1080p. There are 50 video sequences with 3455 densely annotated frames in pixel level. 30 videos with 2079 frames are for training and 20 videos with 1376 frames are for validation." }, { "dkey": "GOT-10k", "dval": "The GOT-10k dataset contains more than 10,000 video segments of real-world moving objects and over 1.5 million manually labelled bounding boxes. The dataset contains more than 560 classes of real-world moving objects and 80+ classes of motion patterns." }, { "dkey": "TrackingNet", "dval": "TrackingNet is a large-scale tracking dataset consisting of videos in the wild. It has a total of 30,643 videos split into 30,132 training videos and 511 testing videos, with an average of 470,9 frames." }, { "dkey": "VOT2018", "dval": "VOT2018 is a dataset for visual object tracking. It consists of 60 challenging videos collected from real-life datasets." }, { "dkey": "MARS", "dval": "MARS (Motion Analysis and Re-identification Set) is a large scale video based person reidentification dataset, an extension of the Market-1501 dataset. It has been collected from six near-synchronized cameras. It consists of 1,261 different pedestrians, who are captured by at least 2 cameras. The variations in poses, colors and illuminations of pedestrians, as well as the poor image quality, make it very difficult to yield high matching accuracy. Moreover, the dataset contains 3,248 distractors in order to make it more realistic. Deformable Part Model and GMMCP tracker were used to automatically generate the tracklets (mostly 25-50 frames long)." }, { "dkey": "Cluttered Omniglot", "dval": "Dataset for one-shot segmentation." }, { "dkey": "NYU-VP", "dval": "NYU-VP is a new dataset for multi-model fitting, vanishing point (VP) estimation in this case. Each image is annotated with up to eight vanishing points, and pre-extracted line segments are provided which act as data points for a robust estimator. Due to its size, the dataset is the first to allow for supervised learning of a multi-model fitting task." }, { "dkey": "Dialogue State Tracking Challenge", "dval": "The Dialog State Tracking Challenges 2 & 3 (DSTC2&3) were research challenge focused on improving the state of the art in tracking the state of spoken dialog systems. State tracking, sometimes called belief tracking, refers to accurately estimating the user's goal as a dialog progresses. Accurate state tracking is desirable because it provides robustness to errors in speech recognition, and helps reduce ambiguity inherent in language within a temporal process like dialog.\nIn these challenges, participants were given labelled corpora of dialogs to develop state tracking algorithms. The trackers were then evaluated on a common set of held-out dialogs, which were released, un-labelled, during a one week period.\n\nThe corpus was collected using Amazon Mechanical Turk, and consists of dialogs in two domains: restaurant information, and tourist information. Tourist information subsumes restaurant information, and includes bars, cafés etc. as well as multiple new slots. There were two rounds of evaluation using this data:\n\nDSTC 2 released a large number of training dialogs related to restaurant search. Compared to DSTC (which was in the bus timetables domain), DSTC 2 introduces changing user goals, tracking 'requested slots' as well as the new restaurants domain. Results from DSTC 2 were presented at SIGDIAL 2014.\nDSTC 3 addressed the problem of adaption to a new domain - tourist information. DSTC 3 releases a small amount of labelled data in the tourist information domain; participants will use this data plus the restaurant data from DSTC 2 for training.\nDialogs used for training are fully labelled; user transcriptions, user dialog-act semantics and dialog state are all annotated. (This corpus therefore is also suitable for studies in Spoken Language Understanding.)" }, { "dkey": "PASCAL-5i", "dval": "PASCAL-5i is a dataset used to evaluate few-shot segmentation. The dataset is sub-divided into 4 folds each containing 5 classes. A fold contains labelled samples from 5 classes that are used for evaluating the few-shot learning method. The rest 15 classes are used for training." } ]

[ "IJB-A", "WMCA", "MSU-MFSD", "FEVER", "FaceForensics", "Large Age-Gap" ]

[ "CASIA-WebFace", "FFHQ" ]

[ { "dkey": "CASIA-WebFace", "dval": "The CASIA-WebFace dataset is used for face verification and face identification tasks. The dataset contains 494,414 face images of 10,575 real identities collected from the web." }, { "dkey": "FFHQ", "dval": "Flickr-Faces-HQ (FFHQ) consists of 70,000 high-quality PNG images at 1024×1024 resolution and contains considerable variation in terms of age, ethnicity and image background. It also has good coverage of accessories such as eyeglasses, sunglasses, hats, etc. The images were crawled from Flickr, thus inheriting all the biases of that website, and automatically aligned and cropped using dlib. Only images under permissive licenses were collected. Various automatic filters were used to prune the set, and finally Amazon Mechanical Turk was used to remove the occasional statues, paintings, or photos of photos." }, { "dkey": "IJB-A", "dval": "The IARPA Janus Benchmark A (IJB-A) database is developed with the aim to augment more challenges to the face recognition task by collecting facial images with a wide variations in pose, illumination, expression, resolution and occlusion. IJB-A is constructed by collecting 5,712 images and 2,085 videos from 500 identities, with an average of 11.4 images and 4.2 videos per identity." }, { "dkey": "WMCA", "dval": "The Wide Multi Channel Presentation Attack (WMCA) database consists of 1941 short video recordings of both bonafide and presentation attacks from 72 different identities. The data is recorded from several channels including color, depth, infra-red, and thermal.\n\nAdditionally, the pulse reading data for bonafide recordings is also provided.\n\nPreprocessed images for some of the channels are also provided for part of the data used in the reference publication.\n\nThe WMCA database is produced at Idiap within the framework of “IARPA BATL” and “H2020 TESLA” projects and it is intended for investigation of presentation attack detection (PAD) methods for face recognition systems." }, { "dkey": "MSU-MFSD", "dval": "The MSU-MFSD dataset contains 280 video recordings of genuine and attack faces. 35 individuals have participated in the development of this database with a total of 280 videos. Two kinds of cameras with different resolutions (720×480 and 640×480) were used to record the videos from the 35 individuals. For the real accesses, each individual has two video recordings captured with the Laptop cameras and Android, respectively. For the video attacks, two types of cameras, the iPhone and Canon cameras were used to capture high definition videos on each of the subject. The videos taken with Canon camera were then replayed on iPad Air screen to generate the HD replay attacks while the videos recorded by the iPhone mobile were replayed itself to generate the mobile replay attacks. Photo attacks were produced by printing the 35 subjects’ photos on A3 papers using HP colour printer. The recording videos with respect to the 35 individuals were divided into training (15 subjects with 120 videos) and testing (40 subjects with 160 videos) datasets, respectively." }, { "dkey": "FEVER", "dval": "FEVER is a publicly available dataset for fact extraction and verification against textual sources.\n\nIt consists of 185,445 claims manually verified against the introductory sections of Wikipedia pages and classified as SUPPORTED, REFUTED or NOTENOUGHINFO. For the first two classes, systems and annotators need to also return the combination of sentences forming the necessary evidence supporting or refuting the claim.\n\nThe claims were generated by human annotators extracting claims from Wikipedia and mutating them in a variety of ways, some of which were meaning-altering. The verification of each claim was conducted in a separate annotation process by annotators who were aware of the page but not the sentence from which original claim was\nextracted and thus in 31.75% of the claims more than one sentence was considered appropriate evidence. Claims require composition of evidence from multiple sentences in 16.82% of cases. Furthermore, in 12.15% of the claims, this evidence was taken from multiple pages." }, { "dkey": "FaceForensics", "dval": "FaceForensics is a video dataset consisting of more than 500,000 frames containing faces from 1004 videos that can be used to study image or video forgeries. All videos are downloaded from Youtube and are cut down to short continuous clips that contain mostly frontal faces. This dataset has two versions:\n\n\n\nSource-to-Target: where the authors reenact over 1000 videos with new facial expressions extracted from other videos, which e.g. can be used to train a classifier to detect fake images or videos.\n\n\n\nSelfreenactment: where the authors use Face2Face to reenact the facial expressions of videos with their own facial expressions as input to get pairs of videos, which e.g. can be used to train supervised generative refinement models." }, { "dkey": "Large Age-Gap", "dval": "Large Age-Gap (LAG) is a dataset for face verification, The dataset contains 3,828 images of 1,010 celebrities. For each identity at least one child/young image and one adult/old image are present." } ]

[ "P-DESTRE", "DukeMTMC-reID", "Airport", "CUHK02" ]

[ "Market-1501", "MARS" ]

[ { "dkey": "Market-1501", "dval": "Market-1501 is a large-scale public benchmark dataset for person re-identification. It contains 1501 identities which are captured by six different cameras, and 32,668 pedestrian image bounding-boxes obtained using the Deformable Part Models pedestrian detector. Each person has 3.6 images on average at each viewpoint. The dataset is split into two parts: 750 identities are utilized for training and the remaining 751 identities are used for testing. In the official testing protocol 3,368 query images are selected as probe set to find the correct match across 19,732 reference gallery images." }, { "dkey": "MARS", "dval": "MARS (Motion Analysis and Re-identification Set) is a large scale video based person reidentification dataset, an extension of the Market-1501 dataset. It has been collected from six near-synchronized cameras. It consists of 1,261 different pedestrians, who are captured by at least 2 cameras. The variations in poses, colors and illuminations of pedestrians, as well as the poor image quality, make it very difficult to yield high matching accuracy. Moreover, the dataset contains 3,248 distractors in order to make it more realistic. Deformable Part Model and GMMCP tracker were used to automatically generate the tracklets (mostly 25-50 frames long)." }, { "dkey": "P-DESTRE", "dval": "Provides consistent ID annotations across multiple days, making it suitable for the extremely challenging problem of person search, i.e., where no clothing information can be reliably used. Apart this feature, the P-DESTRE annotations enable the research on UAV-based pedestrian detection, tracking, re-identification and soft biometric solutions." }, { "dkey": "DukeMTMC-reID", "dval": "The DukeMTMC-reID (Duke Multi-Tracking Multi-Camera ReIDentification) dataset is a subset of the DukeMTMC for image-based person re-ID. The dataset is created from high-resolution videos from 8 different cameras. It is one of the largest pedestrian image datasets wherein images are cropped by hand-drawn bounding boxes. The dataset consists 16,522 training images of 702 identities, 2,228 query images of the other 702 identities and 17,661 gallery images.\n\nNOTE: This dataset has been retracted." }, { "dkey": "Airport", "dval": "The Airport dataset is a dataset for person re-identification which consists of 39,902 images and 9,651 identities across six cameras." }, { "dkey": "CUHK02", "dval": "CUHK02 is a dataset for person re-identification. It contains 1,816 identities from two disjoint camera views. Each identity has two samples per camera view making a total of 7,264 images. It is used for Person Re-identification." } ]

[ "Drive&Act", "PRID2011", "NTU RGB+D", "Cam2BEV" ]

[ "UCF101", "HMDB51" ]

[ { "dkey": "UCF101", "dval": "UCF101 dataset is an extension of UCF50 and consists of 13,320 video clips, which are classified into 101 categories. These 101 categories can be classified into 5 types (Body motion, Human-human interactions, Human-object interactions, Playing musical instruments and Sports). The total length of these video clips is over 27 hours. All the videos are collected from YouTube and have a fixed frame rate of 25 FPS with the resolution of 320 × 240." }, { "dkey": "HMDB51", "dval": "The HMDB51 dataset is a large collection of realistic videos from various sources, including movies and web videos. The dataset is composed of 6,766 video clips from 51 action categories (such as “jump”, “kiss” and “laugh”), with each category containing at least 101 clips. The original evaluation scheme uses three different training/testing splits. In each split, each action class has 70 clips for training and 30 clips for testing. The average accuracy over these three splits is used to measure the final performance." }, { "dkey": "Drive&Act", "dval": "The Drive&Act dataset is a state of the art multi modal benchmark for driver behavior recognition. The dataset includes 3D skeletons in addition to frame-wise hierarchical labels of 9.6 Million frames captured by 6 different views and 3 modalities (RGB, IR and depth).\n\nIt offers following key features:\n\n\n12h of video data in 29 long sequences\nCalibrated multi view camera system with 5 views\nMulti modal videos: NIR, Depth and Color data\nMarkerless motion capture: 3D Body Pose and Head Pose\nModel of the static interior of the car\n83 manually annotated hierarchical activity labels:\nLevel 1: Long running tasks (12)\nLevel 2: Semantic actions (34)\nLevel 3: Object Interaction tripplets [action|object|location] (6|17|14)" }, { "dkey": "PRID2011", "dval": "PRID 2011 is a person reidentification dataset that provides multiple person trajectories recorded from two different static surveillance cameras, monitoring crosswalks and sidewalks. The dataset shows a clean background, and the people in the dataset are rarely occluded. In the dataset, 200 people appear in both views. Among the 200 people, 178 people have more than 20 appearances" }, { "dkey": "NTU RGB+D", "dval": "NTU RGB+D is a large-scale dataset for RGB-D human action recognition. It involves 56,880 samples of 60 action classes collected from 40 subjects. The actions can be generally divided into three categories: 40 daily actions (e.g., drinking, eating, reading), nine health-related actions (e.g., sneezing, staggering, falling down), and 11 mutual actions (e.g., punching, kicking, hugging). These actions take place under 17 different scene conditions corresponding to 17 video sequences (i.e., S001–S017). The actions were captured using three cameras with different horizontal imaging viewpoints, namely, −45∘,0∘, and +45∘. Multi-modality information is provided for action characterization, including depth maps, 3D skeleton joint position, RGB frames, and infrared sequences. The performance evaluation is performed by a cross-subject test that split the 40 subjects into training and test groups, and by a cross-view test that employed one camera (+45∘) for testing, and the other two cameras for training." }, { "dkey": "Cam2BEV", "dval": "The dataset contains two subsets of synthetic, semantically segmented road-scene images, which have been created for developing and applying the methodology described in the paper \"A Sim2Real Deep Learning Approach for the Transformation of Images from Multiple Vehicle-Mounted Cameras to a Semantically Segmented Image in Bird’s Eye View\" (IEEE Xplore, arXiv, YouTube)\n\nThe dataset can be used through the official code implementation of the Cam2BEV methodology described on Github.\n\n| Dataset | # Training Samples | # Validation Samples | # Vehicle Cameras | # Semantic Classes | Contained Images (examples) |\n| --- | --- | --- | --- | --- | --- |\n| Dataset 1: 360° Surround | 33199 | 3731 | 4 (front, rear, left, right) | 30 (CityScapes) | front camera, rear camera, left camera, right camera, bird's eye view, bird's eye view incl. occlusion, homography view |\n| Dataset 2: Front Camera only | 32246 | 3172 | 1 (front) | 30 (CityScapes) | front camera, bird's eye view, bird's eye view incl. occlusion, homography view |" } ]

[ "MOT17", "ORVS", "VOT2017", "TrackingNet", "Word Sense Disambiguation: a Unified Evaluation Framework and Empirical Comparison", "VOT2016" ]

[ "GOT-10k", "COCO" ]

[ { "dkey": "GOT-10k", "dval": "The GOT-10k dataset contains more than 10,000 video segments of real-world moving objects and over 1.5 million manually labelled bounding boxes. The dataset contains more than 560 classes of real-world moving objects and 80+ classes of motion patterns." }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "MOT17", "dval": "The Multiple Object Tracking 17 (MOT17) dataset is a dataset for multiple object tracking. Similar to its previous version MOT16, this challenge contains seven different indoor and outdoor scenes of public places with pedestrians as the objects of interest. A video for each scene is divided into two clips, one for training and the other for testing. The dataset provides detections of objects in the video frames with three detectors, namely SDP, Faster-RCNN and DPM. The challenge accepts both on-line and off-line tracking approaches, where the latter are allowed to use the future video frames to predict tracks." }, { "dkey": "ORVS", "dval": "The ORVS dataset has been newly established as a collaboration between the computer science and visual-science departments at the University of Calgary.\n\nThis dataset contains 49 images (42 training and seven testing images) collected from a clinic in Calgary-Canada. All images were acquired with a Zeiss Visucam 200 with 30 degrees field of view (FOV). The image size is 1444×1444 with 24 bits per pixel. Images and are stored in JPEG format with low compression, which is common in ophthalmology practice. All images were manually traced by an expert who a has been working in the field of retinal-image analysis and went through training. The expert was asked to label all pixels belonging to retinal vessels. The Windows Paint 3D tool was used to manually label the images." }, { "dkey": "VOT2017", "dval": "VOT2017 is a Visual Object Tracking dataset for different tasks that contains 60 short sequences annotated with 6 different attributes." }, { "dkey": "TrackingNet", "dval": "TrackingNet is a large-scale tracking dataset consisting of videos in the wild. It has a total of 30,643 videos split into 30,132 training videos and 511 testing videos, with an average of 470,9 frames." }, { "dkey": "Word Sense Disambiguation: a Unified Evaluation Framework and Empirical Comparison", "dval": "The Evaluation framework of Raganato et al. 2017 includes two training sets (SemCor-Miller et al., 1993- and OMSTI-Taghipour and Ng, 2015-) and five test sets from the Senseval/SemEval series (Edmonds and Cotton, 2001; Snyder and Palmer, 2004; Pradhan et al., 2007; Navigli et al., 2013; Moro and Navigli, 2015), standardized to the same format and sense inventory (i.e. WordNet 3.0).\n\nTypically, there are two kinds of approach for WSD: supervised (which make use of sense-annotated training data) and knowledge-based (which make use of the properties of lexical resources).\n\nSupervised: The most widely used training corpus used is SemCor, with 226,036 sense annotations from 352 documents manually annotated. All supervised systems in the evaluation table are trained on SemCor. Some supervised methods, particularly neural architectures, usually employ the SemEval 2007 dataset as development set (marked by *). The most usual baseline is the Most Frequent Sense (MFS) heuristic, which selects for each target word the most frequent sense in the training data.\n\nKnowledge-based: Knowledge-based systems usually exploit WordNet or BabelNet as semantic network. The first sense given by the underlying sense inventory (i.e. WordNet 3.0) is included as a baseline.\n\nDescription from NLP Progress" }, { "dkey": "VOT2016", "dval": "VOT2016 is a video dataset for visual object tracking. It contains 60 video clips and 21,646 corresponding ground truth maps with pixel-wise annotation of salient objects." } ]

[ "Visual Genome", "DuoRC", "BiPaR", "QuAC", "HotpotQA", "ComplexWebQuestions", "ARCD" ]

[ "SQuAD", "TriviaQA" ]

[ { "dkey": "SQuAD", "dval": "The Stanford Question Answering Dataset (SQuAD) is a collection of question-answer pairs derived from Wikipedia articles. In SQuAD, the correct answers of questions can be any sequence of tokens in the given text. Because the questions and answers are produced by humans through crowdsourcing, it is more diverse than some other question-answering datasets. SQuAD 1.1 contains 107,785 question-answer pairs on 536 articles. SQuAD2.0 (open-domain SQuAD, SQuAD-Open), the latest version, combines the 100,000 questions in SQuAD1.1 with over 50,000 un-answerable questions written adversarially by crowdworkers in forms that are similar to the answerable ones." }, { "dkey": "TriviaQA", "dval": "TriviaQA is a realistic text-based question answering dataset which includes 950K question-answer pairs from 662K documents collected from Wikipedia and the web. This dataset is more challenging than standard QA benchmark datasets such as Stanford Question Answering Dataset (SQuAD), as the answers for a question may not be directly obtained by span prediction and the context is very long. TriviaQA dataset consists of both human-verified and machine-generated QA subsets." }, { "dkey": "Visual Genome", "dval": "Visual Genome contains Visual Question Answering data in a multi-choice setting. It consists of 101,174 images from MSCOCO with 1.7 million QA pairs, 17 questions per image on average. Compared to the Visual Question Answering dataset, Visual Genome represents a more balanced distribution over 6 question types: What, Where, When, Who, Why and How. The Visual Genome dataset also presents 108K images with densely annotated objects, attributes and relationships." }, { "dkey": "DuoRC", "dval": "DuoRC contains 186,089 unique question-answer pairs created from a collection of 7680 pairs of movie plots where each pair in the collection reflects two versions of the same movie.\n\nWhy another RC dataset?\n\nDuoRC pushes the NLP community to address challenges on incorporating knowledge and reasoning in neural architectures for reading comprehension. It poses several interesting challenges such as:\n\n\nDuoRC using parallel plots is especially designed to contain a large number of questions with low lexical overlap between questions and their corresponding passages\nIt requires models to go beyond the content of the given passage itself and incorporate world-knowledge, background knowledge, and common-sense knowledge to arrive at the answer\nIt revolves around narrative passages from movie plots describing complex events and therefore naturally require complex reasoning (e.g. temporal reasoning, entailment, long-distance anaphoras, etc.) across multiple sentences to infer the answer to questions\nSeveral of the questions in DuoRC, while seeming relevant, cannot actually be answered from the given passage. This requires the model to detect the unanswerability of questions. This aspect is important for machines to achieve in industrial settings in particular" }, { "dkey": "BiPaR", "dval": "BiPaR is a manually annotated bilingual parallel novel-style machine reading comprehension (MRC) dataset, developed to support monolingual, multilingual and cross-lingual reading comprehension on novels. The biggest difference between BiPaR and existing reading comprehension datasets is that each triple (Passage, Question, Answer) in BiPaR is written in parallel in two languages. BiPaR is diverse in prefixes of questions, answer types and relationships between questions and passages. Answering the questions requires reading comprehension skills of coreference resolution, multi-sentence reasoning, and understanding of implicit causality." }, { "dkey": "QuAC", "dval": "Question Answering in Context is a large-scale dataset that consists of around 14K crowdsourced Question Answering dialogs with 98K question-answer pairs in total. Data instances consist of an interactive dialog between two crowd workers: (1) a student who poses a sequence of freeform questions to learn as much as possible about a hidden Wikipedia text, and (2) a teacher who answers the questions by providing short excerpts (spans) from the text." }, { "dkey": "HotpotQA", "dval": "HotpotQA is a question answering dataset collected on the English Wikipedia, containing about 113K crowd-sourced questions that are constructed to require the introduction paragraphs of two Wikipedia articles to answer. Each question in the dataset comes with the two gold paragraphs, as well as a list of sentences in these paragraphs that crowdworkers identify as supporting facts necessary to answer the question. \n\nA diverse range of reasoning strategies are featured in HotpotQA, including questions involving missing entities in the question, intersection questions (What satisfies property A and property B?), and comparison questions, where two entities are compared by a common attribute, among others. In the few-document distractor setting, the QA models are given ten paragraphs in which the gold paragraphs are guaranteed to be found; in the open-domain fullwiki setting, the models are only given the question and the entire Wikipedia. Models are evaluated on their answer accuracy and explainability, where the former is measured as overlap between the predicted and gold answers with exact match (EM) and unigram F1, and the latter concerns how well the predicted supporting fact sentences match human annotation (Supporting Fact EM/F1). A joint metric is also reported on this dataset, which encourages systems to perform well on both tasks simultaneously." }, { "dkey": "ComplexWebQuestions", "dval": "ComplexWebQuestions is a dataset for answering complex questions that require reasoning over multiple web snippets. It contains a large set of complex questions in natural language, and can be used in multiple ways:\n\n\nBy interacting with a search engine;\nAs a reading comprehension task: the authors release 12,725,989 web snippets that are relevant for the questions, and were collected during the development of their model;\nAs a semantic parsing task: each question is paired with a SPARQL query that can be executed against Freebase to retrieve the answer." }, { "dkey": "ARCD", "dval": "Composed of 1,395 questions posed by crowdworkers on Wikipedia articles, and a machine translation of the Stanford Question Answering Dataset (Arabic-SQuAD)." } ]

[ "OCID", "JRDB", "Completion3D", "DublinCity", "RobustPointSet" ]

[ "ScanNet", "nuScenes" ]

[ { "dkey": "ScanNet", "dval": "ScanNet is an instance-level indoor RGB-D dataset that includes both 2D and 3D data. It is a collection of labeled voxels rather than points or objects. Up to now, ScanNet v2, the newest version of ScanNet, has collected 1513 annotated scans with an approximate 90% surface coverage. In the semantic segmentation task, this dataset is marked in 20 classes of annotated 3D voxelized objects." }, { "dkey": "nuScenes", "dval": "The nuScenes dataset is a large-scale autonomous driving dataset. The dataset has 3D bounding boxes for 1000 scenes collected in Boston and Singapore. Each scene is 20 seconds long and annotated at 2Hz. This results in a total of 28130 samples for training, 6019 samples for validation and 6008 samples for testing. The dataset has the full autonomous vehicle data suite: 32-beam LiDAR, 6 cameras and radars with complete 360° coverage. The 3D object detection challenge evaluates the performance on 10 classes: cars, trucks, buses, trailers, construction vehicles, pedestrians, motorcycles, bicycles, traffic cones and barriers." }, { "dkey": "OCID", "dval": "Developing robot perception systems for handling objects in the real-world requires computer vision algorithms to be carefully scrutinized with respect to the expected operating domain. This demands large quantities of ground truth data to rigorously evaluate the performance of algorithms.\n\nThe Object Cluttered Indoor Dataset is an RGBD-dataset containing point-wise labeled point-clouds for each object. The data was captured using two ASUS-PRO Xtion cameras that are positioned at different heights. It captures diverse settings of objects, background, context, sensor to scene distance, viewpoint angle and lighting conditions. The main purpose of OCID is to allow systematic comparison of existing object segmentation methods in scenes with increasing amount of clutter. In addition OCID does also provide ground-truth data for other vision tasks like object-classification and recognition." }, { "dkey": "JRDB", "dval": "A novel egocentric dataset collected from social mobile manipulator JackRabbot. The dataset includes 64 minutes of annotated multimodal sensor data including stereo cylindrical 360 degrees RGB video at 15 fps, 3D point clouds from two Velodyne 16 Lidars, line 3D point clouds from two Sick Lidars, audio signal, RGB-D video at 30 fps, 360 degrees spherical image from a fisheye camera and encoder values from the robot's wheels." }, { "dkey": "Completion3D", "dval": "The Completion3D benchmark is a dataset for evaluating state-of-the-art 3D Object Point Cloud Completion methods. Ggiven a partial 3D object point cloud the goal is to infer a complete 3D point cloud for the object." }, { "dkey": "DublinCity", "dval": "A novel benchmark dataset that includes a manually annotated point cloud for over 260 million laser scanning points into 100'000 (approx.) assets from Dublin LiDAR point cloud [12] in 2015. Objects are labelled into 13 classes using hierarchical levels of detail from large (i.e., building, vegetation and ground) to refined (i.e., window, door and tree) elements." }, { "dkey": "RobustPointSet", "dval": "A dataset for robustness analysis of point cloud classification models (independent of data augmentation) to input transformations." } ]

[ "SNIPS", "PHM2017", "Image and Video Advertisements", "CLIRMatrix", "DiDeMo" ]

[ "VIPeR", "DukeMTMC-reID" ]

[ { "dkey": "VIPeR", "dval": "The Viewpoint Invariant Pedestrian Recognition (VIPeR) dataset includes 632 people and two outdoor cameras under different viewpoints and light conditions. Each person has one image per camera and each image has been scaled to be 128×48 pixels. It provides the pose angle of each person as 0° (front), 45°, 90° (right), 135°, and 180° (back)." }, { "dkey": "DukeMTMC-reID", "dval": "The DukeMTMC-reID (Duke Multi-Tracking Multi-Camera ReIDentification) dataset is a subset of the DukeMTMC for image-based person re-ID. The dataset is created from high-resolution videos from 8 different cameras. It is one of the largest pedestrian image datasets wherein images are cropped by hand-drawn bounding boxes. The dataset consists 16,522 training images of 702 identities, 2,228 query images of the other 702 identities and 17,661 gallery images.\n\nNOTE: This dataset has been retracted." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "PHM2017", "dval": "PHM2017 is a new dataset consisting of 7,192 English tweets across six diseases and conditions: Alzheimer’s Disease, heart attack (any severity), Parkinson’s disease, cancer (any type), Depression (any severity), and Stroke. The Twitter search API was used to retrieve the data using the colloquial disease names as search keywords, with the expectation of retrieving a high-recall, low precision dataset. After removing the re-tweets and replies, the tweets were manually annotated. The labels are:\n\n\nself-mention. The tweet contains a health mention with a health self-report of the Twitter account owner, e.g., \"However, I worked hard and ran for Tokyo Mayer Election Campaign in January through February, 2014, without publicizing the cancer.\"\nother-mention. The tweet contains a health mention of a health report about someone other than the account owner, e.g., \"Designer with Parkinson’s couldn’t work then engineer invents bracelet + changes her world\"\nawareness. The tweet contains the disease name, but does not mention a specific person, e.g., \"A Month Before a Heart Attack, Your Body Will Warn You With These 8 Signals\"\nnon-health. The tweet contains the disease name, but the tweet topic is not about health. \"Now I can have cancer on my wall for all to see <3\"" }, { "dkey": "Image and Video Advertisements", "dval": "The Image and Video Advertisements collection consists of an image dataset of 64,832 image ads, and a video dataset of 3,477 ads. The data contains rich annotations encompassing the topic and sentiment of the ads, questions and answers describing what actions the viewer is prompted to take and the reasoning that the ad presents to persuade the viewer (\"What should I do according to this ad, and why should I do it? \"), and symbolic references ads make (e.g. a dove symbolizes peace)." }, { "dkey": "CLIRMatrix", "dval": "CLIRMatrix is a large collection of bilingual and multilingual datasets for Cross-Lingual Information Retrieval. It includes:\n\n\nBI-139: A bilingual dataset of queries in one language matched with relevant documents in another language for 139x138=19,182 language pairs,\nMULTI-8, a multilingual dataset of queries and documents jointly aligned in 8 different languages.\n\nIn total, 49 million unique queries and 34 billion (query, document, label) triplets were mined, making CLIRMatrix the largest and most comprehensive CLIR dataset to date." }, { "dkey": "DiDeMo", "dval": "The Distinct Describable Moments (DiDeMo) dataset is one of the largest and most diverse datasets for the temporal localization of events in videos given natural language descriptions. The videos are collected from Flickr and each video is trimmed to a maximum of 30 seconds. The videos in the dataset are divided into 5-second segments to reduce the complexity of annotation. The dataset is split into training, validation and test sets containing 8,395, 1,065 and 1,004 videos respectively. The dataset contains a total of 26,892 moments and one moment could be associated with descriptions from multiple annotators. The descriptions in DiDeMo dataset are detailed and contain camera movement, temporal transition indicators, and activities. Moreover, the descriptions in DiDeMo are verified so that each description refers to a single moment." } ]

[ "DIV2K", "BSD", "TextZoom", "Make3D", "OST300" ]

[ "CASIA-WebFace", "CelebA" ]

[ { "dkey": "CASIA-WebFace", "dval": "The CASIA-WebFace dataset is used for face verification and face identification tasks. The dataset contains 494,414 face images of 10,575 real identities collected from the web." }, { "dkey": "CelebA", "dval": "CelebFaces Attributes dataset contains 202,599 face images of the size 178×218 from 10,177 celebrities, each annotated with 40 binary labels indicating facial attributes like hair color, gender and age." }, { "dkey": "DIV2K", "dval": "DIV2K is a popular single-image super-resolution dataset which contains 1,000 images with different scenes and is splitted to 800 for training, 100 for validation and 100 for testing. It was collected for NTIRE2017 and NTIRE2018 Super-Resolution Challenges in order to encourage research on image super-resolution with more realistic degradation. This dataset contains low resolution images with different types of degradations. Apart from the standard bicubic downsampling, several types of degradations are considered in synthesizing low resolution images for different tracks of the challenges. Track 2 of NTIRE 2017 contains low resolution images with unknown x4 downscaling. Track 2 and track 4 of NTIRE 2018 correspond to realistic mild ×4 and realistic wild ×4 adverse conditions, respectively. Low-resolution images under realistic mild x4 setting suffer from motion blur, Poisson noise and pixel shifting. Degradations under realistic wild x4 setting are further extended to be of different levels from image to image." }, { "dkey": "BSD", "dval": "BSD is a dataset used frequently for image denoising and super-resolution. Of the subdatasets, BSD100 is aclassical image dataset having 100 test images proposed by Martin et al.. The dataset is composed of a large variety of images ranging from natural images to object-specific such as plants, people, food etc. BSD100 is the testing set of the Berkeley segmentation dataset BSD300." }, { "dkey": "TextZoom", "dval": "TextZoom is a super-resolution dataset that consists of paired Low Resolution – High Resolution scene text images. The images are captured by cameras with different focal length in the wild." }, { "dkey": "Make3D", "dval": "The Make3D dataset is a monocular Depth Estimation dataset that contains 400 single training RGB and depth map pairs, and 134 test samples. The RGB images have high resolution, while the depth maps are provided at low resolution." }, { "dkey": "OST300", "dval": "OST300 is an outdoor scene dataset with 300 test images of outdoor scenes, and a training set of 7 categories of images with rich textures." } ]

[ "MovieQA", "COCO-QA", "LEAF-QA", "TweetQA", "Visual Genome", "VQA-E" ]

[ "FigureQA", "CLEVR" ]

[ { "dkey": "FigureQA", "dval": "FigureQA is a visual reasoning corpus of over one million question-answer pairs grounded in over 100,000 images. The images are synthetic, scientific-style figures from five classes: line plots, dot-line plots, vertical and horizontal bar graphs, and pie charts." }, { "dkey": "CLEVR", "dval": "CLEVR (Compositional Language and Elementary Visual Reasoning) is a synthetic Visual Question Answering dataset. It contains images of 3D-rendered objects; each image comes with a number of highly compositional questions that fall into different categories. Those categories fall into 5 classes of tasks: Exist, Count, Compare Integer, Query Attribute and Compare Attribute. The CLEVR dataset consists of: a training set of 70k images and 700k questions, a validation set of 15k images and 150k questions, A test set of 15k images and 150k questions about objects, answers, scene graphs and functional programs for all train and validation images and questions. Each object present in the scene, aside of position, is characterized by a set of four attributes: 2 sizes: large, small, 3 shapes: square, cylinder, sphere, 2 material types: rubber, metal, 8 color types: gray, blue, brown, yellow, red, green, purple, cyan, resulting in 96 unique combinations." }, { "dkey": "MovieQA", "dval": "The MovieQA dataset is a dataset for movie question answering. to evaluate automatic story comprehension from both video and text. The data set consists of almost 15,000 multiple choice question answers obtained from over 400 movies and features high semantic diversity. Each question comes with a set of five highly plausible answers; only one of which is correct. The questions can be answered using multiple sources of information: movie clips, plots, subtitles, and for a subset scripts and DVS." }, { "dkey": "COCO-QA", "dval": "COCO-QA is a dataset for visual question answering. It consists of:\n\n\n123287 images\n78736 train questions\n38948 test questions\n4 types of questions: object, number, color, location\nAnswers are all one-word." }, { "dkey": "LEAF-QA", "dval": "LEAF-QA, a comprehensive dataset of 250,000 densely annotated figures/charts, constructed from real-world open data sources, along with ~2 million question-answer (QA) pairs querying the structure and semantics of these charts. LEAF-QA highlights the problem of multimodal QA, which is notably different from conventional visual QA (VQA), and has recently gained interest in the community. Furthermore, LEAF-QA is significantly more complex than previous attempts at chart QA, viz. FigureQA and DVQA, which present only limited variations in chart data. LEAF-QA being constructed from real-world sources, requires a novel architecture to enable question answering." }, { "dkey": "TweetQA", "dval": "With social media becoming increasingly popular on which lots of news and real-time events are reported, developing automated question answering systems is critical to the effectiveness of many applications that rely on real-time knowledge. While previous question answering (QA) datasets have concentrated on formal text like news and Wikipedia, the first large-scale dataset for QA over social media data is presented. To make sure the tweets are meaningful and contain interesting information, tweets used by journalists to write news articles are gathered. Then human annotators are asked to write questions and answers upon these tweets. Unlike other QA datasets like SQuAD in which the answers are extractive, the answer are allowed to be abstractive. The task requires model to read a short tweet and a question and outputs a text phrase (does not need to be in the tweet) as the answer." }, { "dkey": "Visual Genome", "dval": "Visual Genome contains Visual Question Answering data in a multi-choice setting. It consists of 101,174 images from MSCOCO with 1.7 million QA pairs, 17 questions per image on average. Compared to the Visual Question Answering dataset, Visual Genome represents a more balanced distribution over 6 question types: What, Where, When, Who, Why and How. The Visual Genome dataset also presents 108K images with densely annotated objects, attributes and relationships." }, { "dkey": "VQA-E", "dval": "VQA-E is a dataset for Visual Question Answering with Explanation, where the models are required to generate and explanation with the predicted answer. The VQA-E dataset is automatically derived from the VQA v2 dataset by synthesizing a textual explanation for each image-question-answer triple." } ]

[ "Stream-51", "StreetHazards", "MVTecAD", "Freiburg Forest" ]

[ "COCO", "Cityscapes" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "Cityscapes", "dval": "Cityscapes is a large-scale database which focuses on semantic understanding of urban street scenes. It provides semantic, instance-wise, and dense pixel annotations for 30 classes grouped into 8 categories (flat surfaces, humans, vehicles, constructions, objects, nature, sky, and void). The dataset consists of around 5000 fine annotated images and 20000 coarse annotated ones. Data was captured in 50 cities during several months, daytimes, and good weather conditions. It was originally recorded as video so the frames were manually selected to have the following features: large number of dynamic objects, varying scene layout, and varying background." }, { "dkey": "Stream-51", "dval": "A new dataset for streaming classification consisting of temporally correlated images from 51 distinct object categories and additional evaluation classes outside of the training distribution to test novelty recognition." }, { "dkey": "StreetHazards", "dval": "StreetHazards is a synthetic dataset for anomaly detection, created by inserting a diverse array of foreign objects into driving scenes and re-render the scenes with these novel objects." }, { "dkey": "MVTecAD", "dval": "MVTec AD is a dataset for benchmarking anomaly detection methods with a focus on industrial inspection. It contains over 5000 high-resolution images divided into fifteen different object and texture categories. Each category comprises a set of defect-free training images and a test set of images with various kinds of defects as well as images without defects.\n\nThere are two common metrics: Detection AUROC and Segmentation (or pixelwise) AUROC\n\nDetection (or, classification) methods output single float (anomaly score) per input test image. \n\nSegmentation methods output anomaly probability for each pixel. \n\"To assess segmentation performance, we evaluate the relative per-region overlap of the segmentation with the ground truth. To get an additional performance measure that is independent of the determined threshold, we compute the area under the receiver operating characteristic curve (ROC AUC). We define the true positive rate as the percentage of pixels that were correctly classified as anomalous\" [1]\nLater segmentation metric was improved to balance regions with small and large area, see PRO-AUC and other in [2]\n\n[1] Paul Bergmann et al, \"MVTec AD — A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection\"\n[2] Bergmann, P., Batzner, K., Fauser, M. et al. The MVTec Anomaly Detection Dataset: A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection. Int J Comput Vis (2021). https://doi.org/10.1007/s11263-020-01400-4" }, { "dkey": "Freiburg Forest", "dval": "The Freiburg Forest dataset was collected using a Viona autonomous mobile robot platform equipped with cameras for capturing multi-spectral and multi-modal images. The dataset may be used for evaluation of different perception algorithms for segmentation, detection, classification, etc. All scenes were recorded at 20 Hz with a camera resolution of 1024x768 pixels. The data was collected on three different days to have enough variability in lighting conditions as shadows and sun angles play a crucial role in the quality of acquired images. The robot traversed about 4.7 km each day. The dataset creators provide manually annotated pixel-wise ground truth segmentation masks for 6 classes: Obstacle, Trail, Sky, Grass, Vegetation, and Void." } ]

[ "SNIPS", "FGADR", "BAR", "OpenEDS", "DialogueFairness", "I-HAZE" ]

[ "TVQA", "MovieQA" ]

[ { "dkey": "TVQA", "dval": "The TVQA dataset is a large-scale vido dataset for video question answering. It is based on 6 popular TV shows (Friends, The Big Bang Theory, How I Met Your Mother, House M.D., Grey's Anatomy, Castle). It includes 152,545 QA pairs from 21,793 TV show clips. The QA pairs are split into the ratio of 8:1:1 for training, validation, and test sets. The TVQA dataset provides the sequence of video frames extracted at 3 FPS, the corresponding subtitles with the video clips, and the query consisting of a question and four answer candidates. Among the four answer candidates, there is only one correct answer." }, { "dkey": "MovieQA", "dval": "The MovieQA dataset is a dataset for movie question answering. to evaluate automatic story comprehension from both video and text. The data set consists of almost 15,000 multiple choice question answers obtained from over 400 movies and features high semantic diversity. Each question comes with a set of five highly plausible answers; only one of which is correct. The questions can be answered using multiple sources of information: movie clips, plots, subtitles, and for a subset scripts and DVS." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "FGADR", "dval": "This dataset has 1,842 images with pixel-level DR-related lesion annotations, and 1,000 images with image-level labels graded by six board-certified ophthalmologists with intra-rater consistency. The proposed dataset will enable extensive studies on DR diagnosis." }, { "dkey": "BAR", "dval": "Biased Action Recognition (BAR) dataset is a real-world image dataset categorized as six action classes which are biased to distinct places. The authors settle these six action classes by inspecting imSitu, which provides still action images from Google Image Search with action and place labels. In detail, the authors choose action classes where images for each of these candidate actions share common place characteristics. At the same time, the place characteristics of action class candidates should be distinct in order to classify the action only from place attributes. The select pairs are six typical action-place pairs: (Climbing, RockWall), (Diving, Underwater), (Fishing, WaterSurface), (Racing, APavedTrack), (Throwing, PlayingField),and (Vaulting, Sky)." }, { "dkey": "OpenEDS", "dval": "OpenEDS (Open Eye Dataset) is a large scale data set of eye-images captured using a virtual-reality (VR) head mounted display mounted with two synchronized eyefacing cameras at a frame rate of 200 Hz under controlled illumination. This dataset is compiled from video capture of the eye-region collected from 152 individual participants and is divided into four subsets: (i) 12,759 images with pixel-level annotations for key eye-regions: iris, pupil and sclera (ii) 252,690 unlabelled eye-images, (iii) 91,200 frames from randomly selected video sequence of 1.5 seconds in duration and (iv) 143 pairs of left and right point cloud data compiled from corneal topography of eye regions collected from a subset, 143 out of 152, participants in the study." }, { "dkey": "DialogueFairness", "dval": "The Dialogue Fairness dataset is used to evaluate and understand fairness in dialogue models, focusing on gender and racial biases." }, { "dkey": "I-HAZE", "dval": "The I-Haze dataset contains 25 indoor hazy images (size 2833×4657 pixels) training. It has 5 hazy images for validation along with their corresponding ground truth images." } ]

[ "BraTS 2017", "RITE", "ISIC 2018 Task 1", "IntrA", "CoNSeP", "CHiME-Home" ]

[ "HRF", "DRIVE" ]

[ { "dkey": "HRF", "dval": "The HRF dataset is a dataset for retinal vessel segmentation which comprises 45 images and is organized as 15 subsets. Each subset contains one healthy fundus image, one image of patient with diabetic retinopathy and one glaucoma image. The image sizes are 3,304 x 2,336, with a training/testing image split of 22/23." }, { "dkey": "DRIVE", "dval": "The Digital Retinal Images for Vessel Extraction (DRIVE) dataset is a dataset for retinal vessel segmentation. It consists of a total of JPEG 40 color fundus images; including 7 abnormal pathology cases. The images were obtained from a diabetic retinopathy screening program in the Netherlands. The images were acquired using Canon CR5 non-mydriatic 3CCD camera with FOV equals to 45 degrees. Each image resolution is 584*565 pixels with eight bits per color channel (3 channels). \n\nThe set of 40 images was equally divided into 20 images for the training set and 20 images for the testing set. Inside both sets, for each image, there is circular field of view (FOV) mask of diameter that is approximately 540 pixels. Inside training set, for each image, one manual segmentation by an ophthalmological expert has been applied. Inside testing set, for each image, two manual segmentations have been applied by two different observers, where the first observer segmentation is accepted as the ground-truth for performance evaluation." }, { "dkey": "BraTS 2017", "dval": "The BRATS2017 dataset. It contains 285 brain tumor MRI scans, with four MRI modalities as T1, T1ce, T2, and Flair for each scan. The dataset also provides full masks for brain tumors, with labels for ED, ET, NET/NCR. The segmentation evaluation is based on three tasks: WT, TC and ET segmentation." }, { "dkey": "RITE", "dval": "The RITE (Retinal Images vessel Tree Extraction) is a database that enables comparative studies on segmentation or classification of arteries and veins on retinal fundus images, which is established based on the public available DRIVE database (Digital Retinal Images for Vessel Extraction).\n\nRITE contains 40 sets of images, equally separated into a training subset and a test subset, the same as DRIVE. The two subsets are built from the corresponding two subsets in DRIVE. For each set, there is a fundus photograph, a vessel reference standard, and a Arteries/Veins (A/V) reference standard. \n\n\nThe fundus photograph is inherited from DRIVE. \nFor the training set, the vessel reference standard is a modified version of 1st_manual from DRIVE. \nFor the test set, the vessel reference standard is 2nd_manual from DRIVE. \nFor the A/V reference standard, four types of vessels are labelled using four colors based on the vessel reference standard. \nArteries are labelled in red; veins are labelled in blue; the overlapping of arteries and veins are labelled in green; the vessels which are uncertain are labelled in white. \nThe fundus photograph is in tif format. And the vessel reference standard and the A/V reference standard are in png format. \n\nThe dataset is described in more detail in our paper, which you will cite if you use the dataset in any way: \n\nHu Q, Abràmoff MD, Garvin MK. Automated separation of binary overlapping trees in low-contrast color retinal images. Med Image Comput Comput Assist Interv. 2013;16(Pt 2):436-43. PubMed PMID: 24579170 https://doi.org/10.1007/978-3-642-40763-5_54" }, { "dkey": "ISIC 2018 Task 1", "dval": "The ISIC 2018 dataset was published by the International Skin Imaging Collaboration (ISIC) as a large-scale dataset of dermoscopy images. This Task 1 dataset is the challenge on lesion segmentation. It includes 2594 images." }, { "dkey": "IntrA", "dval": "IntrA is an open-access 3D intracranial aneurysm dataset that makes the application of points-based and mesh-based classification and segmentation models available. This dataset can be used to diagnose intracranial aneurysms and to extract the neck for a clipping operation in medicine and other areas of deep learning, such as normal estimation and surface reconstruction.\n\n103 3D models of entire brain vessels are collected by reconstructing scanned 2D MRA images of patients (the raw 2D MRA images are not published due to medical ethics).\n1909 blood vessel segments are generated automatically from the complete models, including 1694 healthy vessel segments and 215 aneurysm segments for diagnosis.\n116 aneurysm segments are divided and annotated manually by medical experts; the scale of each aneurysm segment is based on the need for a preoperative examination.\nGeodesic distance matrices are computed and included for each annotated 3D segment, because the expression of the geodesic distance is more accurate than Euclidean distance according to the shape of vessels." }, { "dkey": "CoNSeP", "dval": "The colorectal nuclear segmentation and phenotypes (CoNSeP) dataset consists of 41 H&E stained image tiles, each of size 1,000×1,000 pixels at 40× objective magnification. The images were extracted from 16 colorectal adenocarcinoma (CRA) WSIs, each belonging to an individual patient, and scanned with an Omnyx VL120 scanner within the department of pathology at University Hospitals Coventry and Warwickshire, UK." }, { "dkey": "CHiME-Home", "dval": "CHiME-Home is a dataset for sound source recognition in a domestic environment. It uses around 6.8 hours of domestic environment audio recordings. The recordings were obtained from the CHiME projects – computational hearing in multisource environments – where recording equipment was positioned inside an English Victorian semi-detached house. The recordings were selected from 22 sessions totalling 19.5 hours, with each session made between 7:30 in the morning and 20:00 in the evening. In the considered recordings, the equipment was placed in the lounge (sitting room) near the door opening onto a hallway, with the hallway opening onto a kitchen with no door. With the lounge door typically open, prominent sounds thus may originate from sources both in the lounge and kitchen.\n\nThe choice of permitted labels was motivated by the sources present in the considered acoustic environment: Human speakers (c,m,f); human activity (p); television (v); household appliances (b). Further labels o,S,U respectively relate to any other identifiable sounds, silence, unidentifiable sounds. Labels S,U may respectively only be assigned in isolation. Annotators were acquired to assign at least one label to a chunk, thus annotators may either assign one or more labels from the set {c,m,f,v,p,b,o}, or may alternatively ‘flag’ the chunk using a single label from the set {S,U}." } ]

[ "SNIPS", "LSUN", "CINIC-10", "Places205", "I-HAZE", "COCO-Tasks" ]

[ "BSDS500", "COCO" ]

[ { "dkey": "BSDS500", "dval": "Berkeley Segmentation Data Set 500 (BSDS500) is a standard benchmark for contour detection. This dataset is designed for evaluating natural edge detection that includes not only object contours but also object interior boundaries and background boundaries. It includes 500 natural images with carefully annotated boundaries collected from multiple users. The dataset is divided into three parts: 200 for training, 100 for validation and the rest 200 for test." }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "LSUN", "dval": "The Large-scale Scene Understanding (LSUN) challenge aims to provide a different benchmark for large-scale scene classification and understanding. The LSUN classification dataset contains 10 scene categories, such as dining room, bedroom, chicken, outdoor church, and so on. For training data, each category contains a huge number of images, ranging from around 120,000 to 3,000,000. The validation data includes 300 images, and the test data has 1000 images for each category." }, { "dkey": "CINIC-10", "dval": "CINIC-10 is a dataset for image classification. It has a total of 270,000 images, 4.5 times that of CIFAR-10. It is constructed from two different sources: ImageNet and CIFAR-10. Specifically, it was compiled as a bridge between CIFAR-10 and ImageNet. It is split into three equal subsets - train, validation, and test - each of which contain 90,000 images." }, { "dkey": "Places205", "dval": "The Places205 dataset is a large-scale scene-centric dataset with 205 common scene categories. The training dataset contains around 2,500,000 images from these categories. In the training set, each scene category has the minimum 5,000 and maximum 15,000 images. The validation set contains 100 images per category (a total of 20,500 images), and the testing set includes 200 images per category (a total of 41,000 images)." }, { "dkey": "I-HAZE", "dval": "The I-Haze dataset contains 25 indoor hazy images (size 2833×4657 pixels) training. It has 5 hazy images for validation along with their corresponding ground truth images." }, { "dkey": "COCO-Tasks", "dval": "Comprises about 40,000 images where the most suitable objects for 14 tasks have been annotated." } ]

[ "ConvAI2", "CommonsenseQA", "AV Digits Database", "PHM2017" ]

[ "UCF101", "HMDB51" ]

[ { "dkey": "UCF101", "dval": "UCF101 dataset is an extension of UCF50 and consists of 13,320 video clips, which are classified into 101 categories. These 101 categories can be classified into 5 types (Body motion, Human-human interactions, Human-object interactions, Playing musical instruments and Sports). The total length of these video clips is over 27 hours. All the videos are collected from YouTube and have a fixed frame rate of 25 FPS with the resolution of 320 × 240." }, { "dkey": "HMDB51", "dval": "The HMDB51 dataset is a large collection of realistic videos from various sources, including movies and web videos. The dataset is composed of 6,766 video clips from 51 action categories (such as “jump”, “kiss” and “laugh”), with each category containing at least 101 clips. The original evaluation scheme uses three different training/testing splits. In each split, each action class has 70 clips for training and 30 clips for testing. The average accuracy over these three splits is used to measure the final performance." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "CommonsenseQA", "dval": "The CommonsenseQA is a dataset for commonsense question answering task. The dataset consists of 12,247 questions with 5 choices each.\nThe dataset was generated by Amazon Mechanical Turk workers in the following process (an example is provided in parentheses):\n\n\na crowd worker observes a source concept from ConceptNet (“River”) and three target concepts (“Waterfall”, “Bridge”, “Valley”) that are all related by the same ConceptNet relation (“AtLocation”),\nthe worker authors three questions, one per target concept, such that only that particular target concept is the answer, while the other two distractor concepts are not, (“Where on a river can you hold a cup upright to catch water on a sunny day?”, “Where can I stand on a river to see water falling without getting wet?”, “I’m crossing the river, my feet are wet but my body is dry, where am I?”)\nfor each question, another worker chooses one additional distractor from Concept Net (“pebble”, “stream”, “bank”), and the author another distractor (“mountain”, “bottom”, “island”) manually." }, { "dkey": "AV Digits Database", "dval": "AV Digits Database is an audiovisual database which contains normal, whispered and silent speech. 53 participants were recorded from 3 different views (frontal, 45 and profile) pronouncing digits and phrases in three speech modes.\n\nThe database consists of two parts: digits and short phrases. In the first part, participants were asked to read 10 digits, from 0 to 9, in English in random order five times. In case of non-native English speakers this part was also repeated in the participant’s native language. In total, 53 participants (41 males and 12 females) from 16 nationalities, were recorded with a mean age and standard deviation of 26.7 and 4.3 years, respectively.\n\nIn the second part, participants were asked to read 10 short phrases. The phrases are the same as the ones used in the OuluVS2 database: “Excuse me”, “Goodbye”, “Hello”, “How are you”, “Nice to meet you”, “See you”, “I am sorry”, “Thank you”, “Have a good time”, “You are welcome”. Again, each phrase was repeated five times in 3 different modes, neutral, whisper and silent speech. Thirty nine participants (32 males and 7 females) were recorded for this part with a mean age and standard deviation of 26.3 and 3.8 years, respectively." }, { "dkey": "PHM2017", "dval": "PHM2017 is a new dataset consisting of 7,192 English tweets across six diseases and conditions: Alzheimer’s Disease, heart attack (any severity), Parkinson’s disease, cancer (any type), Depression (any severity), and Stroke. The Twitter search API was used to retrieve the data using the colloquial disease names as search keywords, with the expectation of retrieving a high-recall, low precision dataset. After removing the re-tweets and replies, the tweets were manually annotated. The labels are:\n\n\nself-mention. The tweet contains a health mention with a health self-report of the Twitter account owner, e.g., \"However, I worked hard and ran for Tokyo Mayer Election Campaign in January through February, 2014, without publicizing the cancer.\"\nother-mention. The tweet contains a health mention of a health report about someone other than the account owner, e.g., \"Designer with Parkinson’s couldn’t work then engineer invents bracelet + changes her world\"\nawareness. The tweet contains the disease name, but does not mention a specific person, e.g., \"A Month Before a Heart Attack, Your Body Will Warn You With These 8 Signals\"\nnon-health. The tweet contains the disease name, but the tweet topic is not about health. \"Now I can have cancer on my wall for all to see <3\"" } ]

[ "NATS-Bench", "Cholec80", "NAS-Bench-201", "NAS-Bench-101", "30MQA", "COG", "LEAF-QA" ]

[ "COCO", "Flickr30k" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "Flickr30k", "dval": "The Flickr30k dataset contains 31,000 images collected from Flickr, together with 5 reference sentences provided by human annotators." }, { "dkey": "NATS-Bench", "dval": "A unified benchmark on searching for both topology and size, for (almost) any up-to-date NAS algorithm. NATS-Bench includes the search space of 15,625 neural cell candidates for architecture topology and 32,768 for architecture size on three datasets." }, { "dkey": "Cholec80", "dval": "Cholec80 is an endoscopic video dataset containing 80 videos of cholecystectomy surgeries performed by 13 surgeons. The videos are captured at 25 fps and downsampled to 1 fps for processing. The whole dataset is labeled with the phase and tool presence annotations. The phases have been defined by a senior surgeon in Strasbourg hospital, France. Since the tools are sometimes hardly visible in the images and thus difficult to be recognized visually, a tool is defined as present in an image if at least half of the tool tip is visible.\n\n[https://arxiv.org/pdf/1602.03012.pdf]" }, { "dkey": "NAS-Bench-201", "dval": "NAS-Bench-201 is a benchmark (and search space) for neural architecture search. Each architecture consists of a predefined skeleton with a stack of the searched cell. In this way, architecture search is transformed into the problem of searching a good cell." }, { "dkey": "NAS-Bench-101", "dval": "NAS-Bench-101 is the first public architecture dataset for NAS research. To build NASBench-101, the authors carefully constructed a compact, yet expressive, search space, exploiting graph isomorphisms to identify 423k unique convolutional\narchitectures. The authors trained and evaluated all of these architectures multiple times on CIFAR-10 and compiled the results into a large dataset of over 5 million trained models. This allows researchers to evaluate the quality of a diverse range of models in milliseconds by querying the precomputed dataset." }, { "dkey": "30MQA", "dval": "An enormous question answer pair corpus produced by applying a novel neural network architecture on the knowledge base Freebase to transduce facts into natural language questions." }, { "dkey": "COG", "dval": "A configurable visual question and answer dataset (COG) to parallel experiments in humans and animals. COG is much simpler than the general problem of video analysis, yet it addresses many of the problems relating to visual and logical reasoning and memory -- problems that remain challenging for modern deep learning architectures." }, { "dkey": "LEAF-QA", "dval": "LEAF-QA, a comprehensive dataset of 250,000 densely annotated figures/charts, constructed from real-world open data sources, along with ~2 million question-answer (QA) pairs querying the structure and semantics of these charts. LEAF-QA highlights the problem of multimodal QA, which is notably different from conventional visual QA (VQA), and has recently gained interest in the community. Furthermore, LEAF-QA is significantly more complex than previous attempts at chart QA, viz. FigureQA and DVQA, which present only limited variations in chart data. LEAF-QA being constructed from real-world sources, requires a novel architecture to enable question answering." } ]

[ "Hollywood 3D dataset", "Kinetics-600", "Kinetics", "AVA", "Moments in Time" ]

[ "HMDB51", "KTH" ]

[ { "dkey": "HMDB51", "dval": "The HMDB51 dataset is a large collection of realistic videos from various sources, including movies and web videos. The dataset is composed of 6,766 video clips from 51 action categories (such as “jump”, “kiss” and “laugh”), with each category containing at least 101 clips. The original evaluation scheme uses three different training/testing splits. In each split, each action class has 70 clips for training and 30 clips for testing. The average accuracy over these three splits is used to measure the final performance." }, { "dkey": "KTH", "dval": "The efforts to create a non-trivial and publicly available dataset for action recognition was initiated at the KTH Royal Institute of Technology in 2004. The KTH dataset is one of the most standard datasets, which contains six actions: walk, jog, run, box, hand-wave, and hand clap. To account for performance nuance, each action is performed by 25 different individuals, and the setting is systematically altered for each action per actor. Setting variations include: outdoor (s1), outdoor with scale variation (s2), outdoor with different clothes (s3), and indoor (s4). These variations test the ability of each algorithm to identify actions independent of the background, appearance of the actors, and the scale of the actors." }, { "dkey": "Hollywood 3D dataset", "dval": "A dataset for benchmarking action recognition algorithms in natural environments, while making use of 3D information. The dataset contains around 650 video clips, across 14 classes. In addition, two state of the art action recognition algorithms are extended to make use of the 3D data, and five new interest point detection strategies are also proposed, that extend to the 3D data." }, { "dkey": "Kinetics-600", "dval": "The Kinetics-600 is a large-scale action recognition dataset which consists of around 480K videos from 600 action categories. The 480K videos are divided into 390K, 30K, 60K for training, validation and test sets, respectively. Each video in the dataset is a 10-second clip of action moment annotated from raw YouTube video. It is an extensions of the Kinetics-400 dataset." }, { "dkey": "Kinetics", "dval": "The Kinetics dataset is a large-scale, high-quality dataset for human action recognition in videos. The dataset consists of around 500,000 video clips covering 600 human action classes with at least 600 video clips for each action class. Each video clip lasts around 10 seconds and is labeled with a single action class. The videos are collected from YouTube." }, { "dkey": "AVA", "dval": "AVA is a project that provides audiovisual annotations of video for improving our understanding of human activity. Each of the video clips has been exhaustively annotated by human annotators, and together they represent a rich variety of scenes, recording conditions, and expressions of human activity. There are annotations for:\n\n\nKinetics (AVA-Kinetics) - a crossover between AVA and Kinetics. In order to provide localized action labels on a wider variety of visual scenes, authors provide AVA action labels on videos from Kinetics-700, nearly doubling the number of total annotations, and increasing the number of unique videos by over 500x. \nActions (AvA Actions) - the AVA dataset densely annotates 80 atomic visual actions in 430 15-minute movie clips, where actions are localized in space and time, resulting in 1.62M action labels with multiple labels per human occurring frequently. \nSpoken Activity (AVA ActiveSpeaker, AVA Speech). AVA ActiveSpeaker: associates speaking activity with a visible face, on the AVA v1.0 videos, resulting in 3.65 million frames labeled across ~39K face tracks. AVA Speech densely annotates audio-based speech activity in AVA v1.0 videos, and explicitly labels 3 background noise conditions, resulting in ~46K labeled segments spanning 45 hours of data." }, { "dkey": "Moments in Time", "dval": "Moments in Time is a large-scale dataset for recognizing and understanding action in videos. The dataset includes a collection of one million labeled 3 second videos, involving people, animals, objects or natural phenomena, that capture the gist of a dynamic scene." } ]

[ "AQUA", "YouTube-8M", "AtariARI", "SUN397", "CDTB" ]

[ "COCO", "MovieQA", "DAQUAR" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "MovieQA", "dval": "The MovieQA dataset is a dataset for movie question answering. to evaluate automatic story comprehension from both video and text. The data set consists of almost 15,000 multiple choice question answers obtained from over 400 movies and features high semantic diversity. Each question comes with a set of five highly plausible answers; only one of which is correct. The questions can be answered using multiple sources of information: movie clips, plots, subtitles, and for a subset scripts and DVS." }, { "dkey": "DAQUAR", "dval": "DAQUAR (DAtaset for QUestion Answering on Real-world images) is a dataset of human question answer pairs about images." }, { "dkey": "AQUA", "dval": "The question-answer (QA) pairs are automatically generated using state-of-the-art question generation methods based on paintings and comments provided in an existing art understanding dataset. The QA pairs are cleansed by crowdsourcing workers with respect to their grammatical correctness, answerability, and answers' correctness. The dataset inherently consists of visual (painting-based) and knowledge (comment-based) questions." }, { "dkey": "YouTube-8M", "dval": "The YouTube-8M dataset is a large scale video dataset, which includes more than 7 million videos with 4716 classes labeled by the annotation system. The dataset consists of three parts: training set, validate set, and test set. In the training set, each class contains at least 100 training videos. Features of these videos are extracted by the state-of-the-art popular pre-trained models and released for public use. Each video contains audio and visual modality. Based on the visual information, videos are divided into 24 topics, such as sports, game, arts & entertainment, etc" }, { "dkey": "AtariARI", "dval": "The AtariARI (Atari Annotated RAM Interface) is an environment for representation learning. The Atari Arcade Learning Environment (ALE) does not explicitly expose any ground truth state information. However, ALE does expose the RAM state (128 bytes per timestep) which are used by the game programmer to store important state information such as the location of sprites, the state of the clock, or the current room the agent is in. To extract these variables, the dataset creators consulted commented disassemblies (or source code) of Atari 2600 games which were made available by Engelhardt and Jentzsch and CPUWIZ. The dataset creators were able to find and verify important state variables for a total of 22 games. Once this information was acquired, combining it with the ALE interface produced a wrapper that can automatically output a state label for every example frame generated from the game. The dataset creators make this available with an easy-to-use gym wrapper, which returns this information with no change to existing code using gym interfaces." }, { "dkey": "SUN397", "dval": "The Scene UNderstanding (SUN) database contains 899 categories and 130,519 images. There are 397 well-sampled categories to evaluate numerous state-of-the-art algorithms for scene recognition." }, { "dkey": "CDTB", "dval": "dataset is recorded by several passive and active RGB-D setups and contains indoor as well as outdoor sequences acquired in direct sunlight. The sequences were recorded to contain significant object pose change, clutter, occlusion, and periods of long-term target absence to enable tracker evaluation under realistic conditions. Sequences are per-frame annotated with 13 visual attributes for detailed analysis. It contains around 100,000 samples.)" } ]

[ "GVGAI", "Image Paragraph Captioning", "Localized Narratives", "SWAG" ]

[ "Flickr30k", "COCO" ]

[ { "dkey": "Flickr30k", "dval": "The Flickr30k dataset contains 31,000 images collected from Flickr, together with 5 reference sentences provided by human annotators." }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "GVGAI", "dval": "The General Video Game AI (GVGAI) framework is widely used in research which features a corpus of over 100 single-player games and 60 two-player games. These are fairly small games, each focusing on specific mechanics or skills the players should be able to demonstrate, including clones of classic arcade games such as Space Invaders, puzzle games like Sokoban, adventure games like Zelda or game-theory problems such as the Iterative Prisoners Dilemma. All games are real-time and require players to make decisions in only 40ms at every game tick, although not all games explicitly reward or require fast reactions; in fact, some of the best game-playing approaches add up the time in the beginning of the game to run Breadth-First Search in puzzle games in order to find an accurate solution. However, given the large variety of games (many of which are stochastic and difficult to predict accurately), scoring systems and termination conditions, all unknown to the players, highly-adaptive general methods are needed to tackle the diverse challenges proposed." }, { "dkey": "Image Paragraph Captioning", "dval": "The Image Paragraph Captioning dataset allows researchers to benchmark their progress in generating paragraphs that tell a story about an image. The dataset contains 19,561 images from the Visual Genome dataset. Each image contains one paragraph. The training/val/test sets contains 14,575/2,487/2,489 images.\n\nSince all the images are also part of the Visual Genome dataset, each image also contains 50 region descriptions (short phrases describing parts of an image), 35 objects, 26 attributes and 21 relationships and 17 question-answer pairs." }, { "dkey": "Localized Narratives", "dval": "We propose Localized Narratives, a new form of multimodal image annotations connecting vision and language. We ask annotators to describe an image with their voice while simultaneously hovering their mouse over the region they are describing. Since the voice and the mouse pointer are synchronized, we can localize every single word in the description. This dense visual grounding takes the form of a mouse trace segment per word and is unique to our data. We annotated 849k images with Localized Narratives: the whole COCO, Flickr30k, and ADE20K datasets, and 671k images of Open Images, all of which we make publicly available. We provide an extensive analysis of these annotations showing they are diverse, accurate, and efficient to produce. We also demonstrate their utility on the application of controlled image captioning." }, { "dkey": "SWAG", "dval": "Given a partial description like \"she opened the hood of the car,\" humans can reason about the situation and anticipate what might come next (\"then, she examined the engine\"). SWAG (Situations With Adversarial Generations) is a large-scale dataset for this task of grounded commonsense inference, unifying natural language inference and physically grounded reasoning.\n\nThe dataset consists of 113k multiple choice questions about grounded situations. Each question is a video caption from LSMDC or ActivityNet Captions, with four answer choices about what might happen next in the scene. The correct answer is the (real) video caption for the next event in the video; the three incorrect answers are adversarially generated and human verified, so as to fool machines but not humans. The authors aim for SWAG to be a benchmark for evaluating grounded commonsense NLI and for learning representations." } ]

[ "ImageCLEF-DA", "Libri-Adapt", "VisDA-2017", "EPIC-KITCHENS-100", "MNIST-M", "EMNIST" ]

[ "GTA5", "CompCars" ]

[ { "dkey": "GTA5", "dval": "The GTA5 dataset contains 24966 synthetic images with pixel level semantic annotation. The images have been rendered using the open-world video game Grand Theft Auto 5 and are all from the car perspective in the streets of American-style virtual cities. There are 19 semantic classes which are compatible with the ones of Cityscapes dataset." }, { "dkey": "CompCars", "dval": "The Comprehensive Cars (CompCars) dataset contains data from two scenarios, including images from web-nature and surveillance-nature. The web-nature data contains 163 car makes with 1,716 car models. There are a total of 136,726 images capturing the entire cars and 27,618 images capturing the car parts. The full car images are labeled with bounding boxes and viewpoints. Each car model is labeled with five attributes, including maximum speed, displacement, number of doors, number of seats, and type of car. The surveillance-nature data contains 50,000 car images captured in the front view. \n\nThe dataset can be used for the tasks of:\n\n\nFine-grained classification\nAttribute prediction\nCar model verification\n\nThe dataset can be also used for other tasks such as image ranking, multi-task learning, and 3D reconstruction." }, { "dkey": "ImageCLEF-DA", "dval": "The ImageCLEF-DA dataset is a benchmark dataset for ImageCLEF 2014 domain adaptation challenge, which contains three domains: Caltech-256 (C), ImageNet ILSVRC 2012 (I) and Pascal VOC 2012 (P). For each domain, there are 12 categories and 50 images in each category." }, { "dkey": "Libri-Adapt", "dval": "Libri-Adapt aims to support unsupervised domain adaptation research on speech recognition models." }, { "dkey": "VisDA-2017", "dval": "VisDA-2017 is a simulation-to-real dataset for domain adaptation with over 280,000 images across 12 categories in the training, validation and testing domains. The training images are generated from the same object under different circumstances, while the validation images are collected from MSCOCO.." }, { "dkey": "EPIC-KITCHENS-100", "dval": "This paper introduces the pipeline to scale the largest dataset in egocentric vision EPIC-KITCHENS. The effort culminates in EPIC-KITCHENS-100, a collection of 100 hours, 20M frames, 90K actions in 700 variable-length videos, capturing long-term unscripted activities in 45 environments, using head-mounted cameras. Compared to its previous version (EPIC-KITCHENS-55), EPIC-KITCHENS-100 has been annotated using a novel pipeline that allows denser (54% more actions per minute) and more complete annotations of fine-grained actions (+128% more action segments). This collection also enables evaluating the \"test of time\" - i.e. whether models trained on data collected in 2018 can generalise to new footage collected under the same hypotheses albeit \"two years on\".\nThe dataset is aligned with 6 challenges: action recognition (full and weak supervision), action detection, action anticipation, cross-modal retrieval (from captions), as well as unsupervised domain adaptation for action recognition. For each challenge, we define the task, provide baselines and evaluation metrics." }, { "dkey": "MNIST-M", "dval": "MNIST-M is created by combining MNIST digits with the patches randomly extracted from color photos of BSDS500 as their background. It contains 59,001 training and 90,001 test images." }, { "dkey": "EMNIST", "dval": "EMNIST (extended MNIST) has 4 times more data than MNIST. It is a set of handwritten digits with a 28 x 28 format." } ]

[ "ArtEmis", "JAAD", "DAiSEE", "Aff-Wild2", "MOR-UAV" ]

[ "ImageNet", "AudioSet" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "AudioSet", "dval": "Audioset is an audio event dataset, which consists of over 2M human-annotated 10-second video clips. These clips are collected from YouTube, therefore many of which are in poor-quality and contain multiple sound-sources. A hierarchical ontology of 632 event classes is employed to annotate these data, which means that the same sound could be annotated as different labels. For example, the sound of barking is annotated as Animal, Pets, and Dog. All the videos are split into Evaluation/Balanced-Train/Unbalanced-Train set." }, { "dkey": "ArtEmis", "dval": "ArtEmis is a large-scale dataset aimed at providing a detailed understanding of the interplay between visual content, its emotional effect, and explanations for the latter in language. In contrast to most existing annotation datasets in computer vision, this dataset focuses on the affective experience triggered by visual artworks an the annotators were asked to indicate the dominant emotion they feel for a given image and, crucially, to also provide a grounded verbal explanation for their emotion choice. This leads to a rich set of signals for both the objective content and the affective impact of an image, creating associations with abstract concepts (e.g., “freedom” or “love”), or references that go beyond what is directly visible, including visual similes and metaphors, or subjective references to personal experiences. \n\nThis dataset focuses on visual art (e.g., paintings, artistic photographs) as it is a prime example of imagery created to elicit emotional responses from its viewers. ArtEmis contains 439K emotion attributions and explanations from humans, on 81K artworks from WikiArt.\n\nPaper: ArtEmis: Affective Language for Visual Art" }, { "dkey": "JAAD", "dval": "JAAD is a dataset for studying joint attention in the context of autonomous driving. The focus is on pedestrian and driver behaviors at the point of crossing and factors that influence them. To this end, JAAD dataset provides a richly annotated collection of 346 short video clips (5-10 sec long) extracted from over 240 hours of driving footage. These videos filmed in several locations in North America and Eastern Europe represent scenes typical for everyday urban driving in various weather conditions.\n\nBounding boxes with occlusion tags are provided for all pedestrians making this dataset suitable for pedestrian detection.\n\nBehavior annotations specify behaviors for pedestrians that interact with or require attention of the driver. For each video there are several tags (weather, locations, etc.) and timestamped behavior labels from a fixed list (e.g. stopped, walking, looking, etc.). In addition, a list of demographic attributes is provided for each pedestrian (e.g. age, gender, direction of motion, etc.) as well as a list of visible traffic scene elements (e.g. stop sign, traffic signal, etc.) for each frame.\n\nPaper: Are They Going to Cross? A Benchmark Dataset and Baseline for Pedestrian Crosswalk Behavior" }, { "dkey": "DAiSEE", "dval": "DAiSEE is a multi-label video classification dataset comprising of 9,068 video snippets captured from 112 users for recognizing the user affective states of boredom, confusion, engagement, and frustration \"in the wild\". The dataset has four levels of labels namely - very low, low, high, and very high for each of the affective states, which are crowd annotated and correlated with a gold standard annotation created using a team of expert psychologists." }, { "dkey": "Aff-Wild2", "dval": "Aff-Wild2 is an extension of the Aff-Wild dataset for affect recognition. It approximately doubles the number of included video frames and the number of subjects; thus, improving the variability of the included behaviors and of the involved persons." }, { "dkey": "MOR-UAV", "dval": "A large-scale video dataset for MOR in aerial videos." } ]

[ "UASOL", "2D-3D Match Dataset", "Places", "Localized Narratives", "ICL-NUIM", "3DMatch" ]

[ "Charades", "HMDB51" ]

[ { "dkey": "Charades", "dval": "The Charades dataset is composed of 9,848 videos of daily indoors activities with an average length of 30 seconds, involving interactions with 46 objects classes in 15 types of indoor scenes and containing a vocabulary of 30 verbs leading to 157 action classes. Each video in this dataset is annotated by multiple free-text descriptions, action labels, action intervals and classes of interacting objects. 267 different users were presented with a sentence, which includes objects and actions from a fixed vocabulary, and they recorded a video acting out the sentence. In total, the dataset contains 66,500 temporal annotations for 157 action classes, 41,104 labels for 46 object classes, and 27,847 textual descriptions of the videos. In the standard split there are7,986 training video and 1,863 validation video." }, { "dkey": "HMDB51", "dval": "The HMDB51 dataset is a large collection of realistic videos from various sources, including movies and web videos. The dataset is composed of 6,766 video clips from 51 action categories (such as “jump”, “kiss” and “laugh”), with each category containing at least 101 clips. The original evaluation scheme uses three different training/testing splits. In each split, each action class has 70 clips for training and 30 clips for testing. The average accuracy over these three splits is used to measure the final performance." }, { "dkey": "UASOL", "dval": "The UASOL an RGB-D stereo dataset, that contains 160902 frames, filmed at 33 different scenes, each with between 2 k and 10 k frames. The frames show different paths from the perspective of a pedestrian, including sidewalks, trails, roads, etc. The images were extracted from video files with 15 fps at HD2K resolution with a size of 2280 × 1282 pixels. The dataset also provides a GPS geolocalization tag for each second of the sequences and reflects different climatological conditions. It also involved up to 4 different persons filming the dataset at different moments of the day.\n\nWe propose a train, validation and test split to train the network. \nAdditionally, we introduce a subset of 676 pairs of RGB Stereo images and their respective depth, which we extracted randomly from the entire dataset. This given test set is introduced to make comparability possible between the different methods trained with the dataset." }, { "dkey": "2D-3D Match Dataset", "dval": "2D-3D Match Dataset is a new dataset of 2D-3D correspondences by leveraging the availability of several 3D datasets from RGB-D scans. Specifically, the data from SceneNN and 3DMatch are used. The training dataset consists of 110 RGB-D scans, of which 56 scenes are from SceneNN and 54 scenes are from 3DMatch. The 2D-3D correspondence data is generated as follows. Given a 3D point which is randomly sampled from a 3D point cloud, a set of 3D patches from different scanning views are extracted. To find a 2D-3D correspondence, for each 3D patch, its 3D position is re-projected into all RGB-D frames for which the point lies in the camera frustum, taking occlusion into account. The corresponding local 2D patches around the re-projected point are extracted. In total, around 1.4 millions 2D-3D correspondences are collected." }, { "dkey": "Places", "dval": "The Places dataset is proposed for scene recognition and contains more than 2.5 million images covering more than 205 scene categories with more than 5,000 images per category." }, { "dkey": "Localized Narratives", "dval": "We propose Localized Narratives, a new form of multimodal image annotations connecting vision and language. We ask annotators to describe an image with their voice while simultaneously hovering their mouse over the region they are describing. Since the voice and the mouse pointer are synchronized, we can localize every single word in the description. This dense visual grounding takes the form of a mouse trace segment per word and is unique to our data. We annotated 849k images with Localized Narratives: the whole COCO, Flickr30k, and ADE20K datasets, and 671k images of Open Images, all of which we make publicly available. We provide an extensive analysis of these annotations showing they are diverse, accurate, and efficient to produce. We also demonstrate their utility on the application of controlled image captioning." }, { "dkey": "ICL-NUIM", "dval": "The ICL-NUIM dataset aims at benchmarking RGB-D, Visual Odometry and SLAM algorithms. Two different scenes (the living room and the office room scene) are provided with ground truth. Living room has 3D surface ground truth together with the depth-maps as well as camera poses and as a result perfectly suits not just for benchmarking camera trajectory but also reconstruction. Office room scene comes with only trajectory data and does not have any explicit 3D model with it.\n\nAll data is compatible with the evaluation tools available for the TUM RGB-D dataset, and if your system can take TUM RGB-D format PNGs as input, the authors’ TUM RGB-D Compatible data will also work (given the correct camera parameters)." }, { "dkey": "3DMatch", "dval": "The 3DMATCH benchmark evaluates how well descriptors (both 2D and 3D) can establish correspondences between RGB-D frames of different views. The dataset contains 2D RGB-D patches and 3D patches (local TDF voxel grid volumes) of wide-baselined correspondences. \n\nThe pixel size of each 2D patch is determined by the projection of the 0.3m3 local 3D patch around the interest point onto the image plane." } ]

[ "LSP", "THEODORE", "MPI-INF-3DHP", "UMDFaces", "PASCAL3D+", "MLFP" ]

[ "COCO", "MPII" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "MPII", "dval": "The MPII Human Pose Dataset for single person pose estimation is composed of about 25K images of which 15K are training samples, 3K are validation samples and 7K are testing samples (which labels are withheld by the authors). The images are taken from YouTube videos covering 410 different human activities and the poses are manually annotated with up to 16 body joints." }, { "dkey": "LSP", "dval": "The Leeds Sports Pose (LSP) dataset is widely used as the benchmark for human pose estimation. The original LSP dataset contains 2,000 images of sportspersons gathered from Flickr, 1000 for training and 1000 for testing. Each image is annotated with 14 joint locations, where left and right joints are consistently labelled from a person-centric viewpoint. The extended LSP dataset contains additional 10,000 images labeled for training.\n\nImage: Sumer et al" }, { "dkey": "THEODORE", "dval": "Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset." }, { "dkey": "MPI-INF-3DHP", "dval": "MPI-INF-3DHP is a 3D human body pose estimation dataset consisting of both constrained indoor and complex outdoor scenes. It records 8 actors performing 8 activities from 14 camera views. It consists on >1.3M frames captured from the 14 cameras." }, { "dkey": "UMDFaces", "dval": "UMDFaces is a face dataset divided into two parts:\n\n\nStill Images - 367,888 face annotations for 8,277 subjects.\nVideo Frames - Over 3.7 million annotated video frames from over 22,000 videos of 3100 subjects.\n\nPart 1 - Still Images\n\nThe dataset contains 367,888 face annotations for 8,277 subjects divided into 3 batches. The annotations contain human curated bounding boxes for faces and estimated pose (yaw, pitch, and roll), locations of twenty-one keypoints, and gender information generated by a pre-trained neural network.\n\nPart 2 - Video Frames\n\nThe second part contains 3,735,476 annotated video frames extracted from a total of 22,075 for 3,107 subjects. The annotations contain the estimated pose (yaw, pitch, and roll), locations of twenty-one keypoints, and gender information generated by a pre-trained neural network." }, { "dkey": "PASCAL3D+", "dval": "The Pascal3D+ multi-view dataset consists of images in the wild, i.e., images of object categories exhibiting high variability, captured under uncontrolled settings, in cluttered scenes and under many different poses. Pascal3D+ contains 12 categories of rigid objects selected from the PASCAL VOC 2012 dataset. These objects are annotated with pose information (azimuth, elevation and distance to camera). Pascal3D+ also adds pose annotated images of these 12 categories from the ImageNet dataset." }, { "dkey": "MLFP", "dval": "The MLFP dataset consists of face presentation attacks captured with seven 3D latex masks and three 2D print attacks. The dataset contains videos captured from color, thermal and infrared channels." } ]

[ "CrowdFlow", "MVSEC", "SlowFlow", "VIsual PERception (VIPER)", "Creative Flow+ Dataset", "JHMDB" ]

[ "KITTI", "SYNTHIA" ]

[ { "dkey": "KITTI", "dval": "KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) is one of the most popular datasets for use in mobile robotics and autonomous driving. It consists of hours of traffic scenarios recorded with a variety of sensor modalities, including high-resolution RGB, grayscale stereo cameras, and a 3D laser scanner. Despite its popularity, the dataset itself does not contain ground truth for semantic segmentation. However, various researchers have manually annotated parts of the dataset to fit their necessities. Álvarez et al. generated ground truth for 323 images from the road detection challenge with three classes: road, vertical, and sky. Zhang et al. annotated 252 (140 for training and 112 for testing) acquisitions – RGB and Velodyne scans – from the tracking challenge for ten object categories: building, sky, road, vegetation, sidewalk, car, pedestrian, cyclist, sign/pole, and fence. Ros et al. labeled 170 training images and 46 testing images (from the visual odometry challenge) with 11 classes: building, tree, sky, car, sign, road, pedestrian, fence, pole, sidewalk, and bicyclist." }, { "dkey": "SYNTHIA", "dval": "The SYNTHIA dataset is a synthetic dataset that consists of 9400 multi-viewpoint photo-realistic frames rendered from a virtual city and comes with pixel-level semantic annotations for 13 classes. Each frame has resolution of 1280 × 960." }, { "dkey": "CrowdFlow", "dval": "The TUB CrowdFlow is a synthetic dataset that contains 10 sequences showing 5 scenes. Each scene is rendered twice: with a static point of view and a dynamic camera to simulate drone/UAV based surveillance. The scenes are render using Unreal Engine at HD resolution (1280x720) at 25 fps, which is typical for current commercial CCTV surveillance systems. The total number of frames is 3200.\n\nEach sequence has the following ground-truth data:\n\n\nOptical flow fields\nPerson trajectories (up to 1451)\nDense pixel trajectories" }, { "dkey": "MVSEC", "dval": "The Multi Vehicle Stereo Event Camera (MVSEC) dataset is a collection of data designed for the development of novel 3D perception algorithms for event based cameras. Stereo event data is collected from car, motorbike, hexacopter and handheld data, and fused with lidar, IMU, motion capture and GPS to provide ground truth pose and depth images." }, { "dkey": "SlowFlow", "dval": "SlowFlow is an optical flow dataset collected by applying Slow Flow technique on data from a high-speed camera and analyzing the performance of the state-of-the-art in optical flow under various levels of motion blur." }, { "dkey": "VIsual PERception (VIPER)", "dval": "VIPER is a benchmark suite for visual perception. The benchmark is based on more than 250K high-resolution video frames, all annotated with ground-truth data for both low-level and high-level vision tasks, including optical flow, semantic instance segmentation, object detection and tracking, object-level 3D scene layout, and visual odometry. Ground-truth data for all tasks is available for every frame. The data was collected while driving, riding, and walking a total of 184 kilometers in diverse ambient conditions in a realistic virtual world." }, { "dkey": "Creative Flow+ Dataset", "dval": "Includes 3000 animated sequences rendered using styles randomly selected from 40 textured line styles and 38 shading styles, spanning the range between flat cartoon fill and wildly sketchy shading. The dataset includes 124K+ train set frames and 10K test set frames rendered at 1500x1500 resolution, far surpassing the largest available optical flow datasets in size." }, { "dkey": "JHMDB", "dval": "JHMDB is an action recognition dataset that consists of 960 video sequences belonging to 21 actions. It is a subset of the larger HMDB51 dataset collected from digitized movies and YouTube videos. The dataset contains video and annotation for puppet flow per frame (approximated optimal flow on the person), puppet mask per frame, joint positions per frame, action label per clip and meta label per clip (camera motion, visible body parts, camera viewpoint, number of people, video quality)." } ]

[ "Talk2Car", "Syn2Real", "2D-3D-S", "DailyDialog++", "SNIPS" ]

[ "ImageNet", "ShapeNet" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "ShapeNet", "dval": "ShapeNet is a large scale repository for 3D CAD models developed by researchers from Stanford University, Princeton University and the Toyota Technological Institute at Chicago, USA. The repository contains over 300M models with 220,000 classified into 3,135 classes arranged using WordNet hypernym-hyponym relationships. ShapeNet Parts subset contains 31,693 meshes categorised into 16 common object classes (i.e. table, chair, plane etc.). Each shapes ground truth contains 2-5 parts (with a total of 50 part classes)." }, { "dkey": "Talk2Car", "dval": "The Talk2Car dataset finds itself at the intersection of various research domains, promoting the development of cross-disciplinary solutions for improving the state-of-the-art in grounding natural language into visual space. The annotations were gathered with the following aspects in mind:\nFree-form high quality natural language commands, that stimulate the development of solutions that can operate in the wild.\nA realistic task setting. Specifically, the authors consider an autonomous driving setting, where a passenger can control the actions of an Autonomous Vehicle by giving commands in natural language.\nThe Talk2Car dataset was build on top of the nuScenes dataset to include an extensive suite of sensor modalities, i.e. semantic maps, GPS, LIDAR, RADAR and 360-degree RGB images annotated with 3D bounding boxes. Such variety of input modalities sets the object referral task on the Talk2Car dataset apart from related challenges, where additional sensor modalities are generally missing." }, { "dkey": "Syn2Real", "dval": "Syn2Real, a synthetic-to-real visual domain adaptation benchmark meant to encourage further development of robust domain transfer methods. The goal is to train a model on a synthetic \"source\" domain and then update it so that its performance improves on a real \"target\" domain, without using any target annotations. It includes three tasks, illustrated in figures above: the more traditional closed-set classification task with a known set of categories; the less studied open-set classification task with unknown object categories in the target domain; and the object detection task, which involves localizing instances of objects by predicting their bounding boxes and corresponding class labels." }, { "dkey": "2D-3D-S", "dval": "The 2D-3D-S dataset provides a variety of mutually registered modalities from 2D, 2.5D and 3D domains, with instance-level semantic and geometric annotations. It covers over 6,000 m2 collected in 6 large-scale indoor areas that originate from 3 different buildings. It contains over 70,000 RGB images, along with the corresponding depths, surface normals, semantic annotations, global XYZ images (all in forms of both regular and 360° equirectangular images) as well as camera information. It also includes registered raw and semantically annotated 3D meshes and point clouds. The dataset enables development of joint and cross-modal learning models and potentially unsupervised approaches utilizing the regularities present in large-scale indoor spaces." }, { "dkey": "DailyDialog++", "dval": "Consists of (i) five relevant responses for each context and (ii) five adversarially crafted irrelevant responses for each context." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." } ]

[ "BREAK", "VoxPopuli", "AtariARI", "Icentia11K", "EmoBank" ]

[ "SNLI", "SentEval", "SST", "WikiText-2" ]

[ { "dkey": "SNLI", "dval": "The SNLI dataset (Stanford Natural Language Inference) consists of 570k sentence-pairs manually labeled as entailment, contradiction, and neutral. Premises are image captions from Flickr30k, while hypotheses were generated by crowd-sourced annotators who were shown a premise and asked to generate entailing, contradicting, and neutral sentences. Annotators were instructed to judge the relation between sentences given that they describe the same event. Each pair is labeled as “entailment”, “neutral”, “contradiction” or “-”, where “-” indicates that an agreement could not be reached." }, { "dkey": "SentEval", "dval": "SentEval is a toolkit for evaluating the quality of universal sentence representations. SentEval encompasses a variety of tasks, including binary and multi-class classification, natural language inference and sentence similarity. The set of tasks was selected based on what appears to be the community consensus regarding the appropriate evaluations for universal sentence representations. The toolkit comes with scripts to download and preprocess datasets, and an easy interface to evaluate sentence encoders." }, { "dkey": "SST", "dval": "The Stanford Sentiment Treebank is a corpus with fully labeled parse trees that allows for a\ncomplete analysis of the compositional effects of\nsentiment in language. The corpus is based on\nthe dataset introduced by Pang and Lee (2005) and\nconsists of 11,855 single sentences extracted from\nmovie reviews. It was parsed with the Stanford\nparser and includes a total of 215,154 unique phrases\nfrom those parse trees, each annotated by 3 human judges.\n\nEach phrase is labelled as either negative, somewhat negative, neutral, somewhat positive or positive.\nThe corpus with all 5 labels is referred to as SST-5 or SST fine-grained. Binary classification experiments on full sentences (negative or somewhat negative vs somewhat positive or positive with neutral sentences discarded) refer to the dataset as SST-2 or SST binary." }, { "dkey": "WikiText-2", "dval": "The WikiText language modeling dataset is a collection of over 100 million tokens extracted from the set of verified Good and Featured articles on Wikipedia. The dataset is available under the Creative Commons Attribution-ShareAlike License.\n\nCompared to the preprocessed version of Penn Treebank (PTB), WikiText-2 is over 2 times larger and WikiText-103 is over 110 times larger. The WikiText dataset also features a far larger vocabulary and retains the original case, punctuation and numbers - all of which are removed in PTB. As it is composed of full articles, the dataset is well suited for models that can take advantage of long term dependencies." }, { "dkey": "BREAK", "dval": "Break is a question understanding dataset, aimed at training models to reason over complex questions. It features 83,978 natural language questions, annotated with a new meaning representation, Question Decomposition Meaning Representation (QDMR). Each example has the natural question along with its QDMR representation. Break contains human composed questions, sampled from 10 leading question-answering benchmarks over text, images and databases. This dataset was created by a team of NLP researchers at Tel Aviv University and Allen Institute for AI." }, { "dkey": "VoxPopuli", "dval": "VoxPopuli is a large-scale multilingual corpus providing 100K hours of unlabelled speech data in 23 languages. It is the largest open data to date for unsupervised representation learning as well as semi-supervised learning. VoxPopuli also contains 1.8K hours of transcribed speeches in 16 languages and their aligned oral interpretations into 5 other languages totaling 5.1K hours." }, { "dkey": "AtariARI", "dval": "The AtariARI (Atari Annotated RAM Interface) is an environment for representation learning. The Atari Arcade Learning Environment (ALE) does not explicitly expose any ground truth state information. However, ALE does expose the RAM state (128 bytes per timestep) which are used by the game programmer to store important state information such as the location of sprites, the state of the clock, or the current room the agent is in. To extract these variables, the dataset creators consulted commented disassemblies (or source code) of Atari 2600 games which were made available by Engelhardt and Jentzsch and CPUWIZ. The dataset creators were able to find and verify important state variables for a total of 22 games. Once this information was acquired, combining it with the ALE interface produced a wrapper that can automatically output a state label for every example frame generated from the game. The dataset creators make this available with an easy-to-use gym wrapper, which returns this information with no change to existing code using gym interfaces." }, { "dkey": "Icentia11K", "dval": "Public ECG dataset of continuous raw signals for representation learning containing 11 thousand patients and 2 billion labelled beats." }, { "dkey": "EmoBank", "dval": "EmoBank is a corpus of 10k English sentences balancing multiple genres, annotated with dimensional emotion metadata in the Valence-Arousal-Dominance (VAD) representation format. EmoBank excels with a bi-perspectival and bi-representational design." } ]

[ "SNIPS", "ConvAI2", "CLUECorpus2020", "MannequinChallenge", "UAVA", "YouTube-8M" ]

[ "ImageNet", "COCO" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "CLUECorpus2020", "dval": "CLUECorpus2020 is a large-scale corpus that can be used directly for self-supervised learning such as pre-training of a language model, or language generation. It has 100G raw corpus with 35 billion Chinese characters, which is retrieved from Common Crawl." }, { "dkey": "MannequinChallenge", "dval": "The MannequinChallenge Dataset (MQC) provides in-the-wild videos of people in static poses while a hand-held camera pans around the scene. The dataset consists of three splits for training, validation and testing." }, { "dkey": "UAVA", "dval": "The UAVA,<i>UAV-Assistant</i>, dataset is specifically designed for fostering applications which consider UAVs and humans as cooperative agents.\nWe employ a real-world 3D scanned dataset (<a href=\"https://niessner.github.io/Matterport/\">Matterport3D</a>), physically-based rendering, a gamified simulator for realistic drone navigation trajectory collection, to generate realistic multimodal data both from the user’s exocentric view of the drone, as well as the drone’s egocentric view." }, { "dkey": "YouTube-8M", "dval": "The YouTube-8M dataset is a large scale video dataset, which includes more than 7 million videos with 4716 classes labeled by the annotation system. The dataset consists of three parts: training set, validate set, and test set. In the training set, each class contains at least 100 training videos. Features of these videos are extracted by the state-of-the-art popular pre-trained models and released for public use. Each video contains audio and visual modality. Based on the visual information, videos are divided into 24 topics, such as sports, game, arts & entertainment, etc" } ]

[ "ANLI", "Glint360K", "Synscapes", "Dialogue State Tracking Challenge", "SICK", "MLMA Hate Speech", "LogiQA" ]

[ "RACE", "QNLI", "GLUE", "SQuAD" ]

[ { "dkey": "RACE", "dval": "The ReAding Comprehension dataset from Examinations (RACE) dataset is a machine reading comprehension dataset consisting of 27,933 passages and 97,867 questions from English exams, targeting Chinese students aged 12-18. RACE consists of two subsets, RACE-M and RACE-H, from middle school and high school exams, respectively. RACE-M has 28,293 questions and RACE-H has 69,574. Each question is associated with 4 candidate answers, one of which is correct. The data generation process of RACE differs from most machine reading comprehension datasets - instead of generating questions and answers by heuristics or crowd-sourcing, questions in RACE are specifically designed for testing human reading skills, and are created by domain experts." }, { "dkey": "QNLI", "dval": "The QNLI (Question-answering NLI) dataset is a Natural Language Inference dataset automatically derived from the Stanford Question Answering Dataset v1.1 (SQuAD). SQuAD v1.1 consists of question-paragraph pairs, where one of the sentences in the paragraph (drawn from Wikipedia) contains the answer to the corresponding question (written by an annotator). The dataset was converted into sentence pair classification by forming a pair between each question and each sentence in the corresponding context, and filtering out pairs with low lexical overlap between the question and the context sentence. The task is to determine whether the context sentence contains the answer to the question. This modified version of the original task removes the requirement that the model select the exact answer, but also removes the simplifying assumptions that the answer is always present in the input and that lexical overlap is a reliable cue. The QNLI dataset is part of GLEU benchmark." }, { "dkey": "GLUE", "dval": "General Language Understanding Evaluation (GLUE) benchmark is a collection of nine natural language understanding tasks, including single-sentence tasks CoLA and SST-2, similarity and paraphrasing tasks MRPC, STS-B and QQP, and natural language inference tasks MNLI, QNLI, RTE and WNLI." }, { "dkey": "SQuAD", "dval": "The Stanford Question Answering Dataset (SQuAD) is a collection of question-answer pairs derived from Wikipedia articles. In SQuAD, the correct answers of questions can be any sequence of tokens in the given text. Because the questions and answers are produced by humans through crowdsourcing, it is more diverse than some other question-answering datasets. SQuAD 1.1 contains 107,785 question-answer pairs on 536 articles. SQuAD2.0 (open-domain SQuAD, SQuAD-Open), the latest version, combines the 100,000 questions in SQuAD1.1 with over 50,000 un-answerable questions written adversarially by crowdworkers in forms that are similar to the answerable ones." }, { "dkey": "ANLI", "dval": "The Adversarial Natural Language Inference (ANLI, Nie et al.) is a new large-scale NLI benchmark dataset, collected via an iterative, adversarial human-and-model-in-the-loop procedure. Particular, the data is selected to be difficult to the state-of-the-art models, including BERT and RoBERTa." }, { "dkey": "Glint360K", "dval": "The largest and cleanest face recognition dataset Glint360K, \nwhich contains 17,091,657 images of 360,232 individuals, baseline models trained on Glint360K can easily achieve state-of-the-art performance." }, { "dkey": "Synscapes", "dval": "Synscapes is a synthetic dataset for street scene parsing created using photorealistic rendering techniques, and show state-of-the-art results for training and validation as well as new types of analysis." }, { "dkey": "Dialogue State Tracking Challenge", "dval": "The Dialog State Tracking Challenges 2 & 3 (DSTC2&3) were research challenge focused on improving the state of the art in tracking the state of spoken dialog systems. State tracking, sometimes called belief tracking, refers to accurately estimating the user's goal as a dialog progresses. Accurate state tracking is desirable because it provides robustness to errors in speech recognition, and helps reduce ambiguity inherent in language within a temporal process like dialog.\nIn these challenges, participants were given labelled corpora of dialogs to develop state tracking algorithms. The trackers were then evaluated on a common set of held-out dialogs, which were released, un-labelled, during a one week period.\n\nThe corpus was collected using Amazon Mechanical Turk, and consists of dialogs in two domains: restaurant information, and tourist information. Tourist information subsumes restaurant information, and includes bars, cafés etc. as well as multiple new slots. There were two rounds of evaluation using this data:\n\nDSTC 2 released a large number of training dialogs related to restaurant search. Compared to DSTC (which was in the bus timetables domain), DSTC 2 introduces changing user goals, tracking 'requested slots' as well as the new restaurants domain. Results from DSTC 2 were presented at SIGDIAL 2014.\nDSTC 3 addressed the problem of adaption to a new domain - tourist information. DSTC 3 releases a small amount of labelled data in the tourist information domain; participants will use this data plus the restaurant data from DSTC 2 for training.\nDialogs used for training are fully labelled; user transcriptions, user dialog-act semantics and dialog state are all annotated. (This corpus therefore is also suitable for studies in Spoken Language Understanding.)" }, { "dkey": "SICK", "dval": "The Sentences Involving Compositional Knowledge (SICK) dataset is a dataset for compositional distributional semantics. It includes a large number of sentence pairs that are rich in the lexical, syntactic and semantic phenomena. Each pair of sentences is annotated in two dimensions: relatedness and entailment. The relatedness score ranges from 1 to 5, and Pearson’s r is used for evaluation; the entailment relation is categorical, consisting of entailment, contradiction, and neutral. There are 4439 pairs in the train split, 495 in the trial split used for development and 4906 in the test split. The sentence pairs are generated from image and video caption datasets before being paired up using some algorithm." }, { "dkey": "MLMA Hate Speech", "dval": "A new multilingual multi-aspect hate speech analysis dataset and use it to test the current state-of-the-art multilingual multitask learning approaches." }, { "dkey": "LogiQA", "dval": "LogiQA consists of 8,678 QA instances, covering multiple types of deductive reasoning. Results show that state-of-the-art neural models perform by far worse than human ceiling. The dataset can also serve as a benchmark for reinvestigating logical AI under the deep learning NLP setting." } ]

[ "Cityscapes-VPS", "Cityscapes Panoptic Parts", "SNIPS", "Pascal Panoptic Parts", "ConvAI2", "I-HAZE" ]

[ "COCO", "Cityscapes" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "Cityscapes", "dval": "Cityscapes is a large-scale database which focuses on semantic understanding of urban street scenes. It provides semantic, instance-wise, and dense pixel annotations for 30 classes grouped into 8 categories (flat surfaces, humans, vehicles, constructions, objects, nature, sky, and void). The dataset consists of around 5000 fine annotated images and 20000 coarse annotated ones. Data was captured in 50 cities during several months, daytimes, and good weather conditions. It was originally recorded as video so the frames were manually selected to have the following features: large number of dynamic objects, varying scene layout, and varying background." }, { "dkey": "Cityscapes-VPS", "dval": "Cityscapes-VPS is a video extension of the Cityscapes validation split. It provides 2500-frame panoptic labels that temporally extend the 500 Cityscapes image-panoptic labels. There are total 3000-frame panoptic labels which correspond to 5, 10, 15, 20, 25, and 30th frames of each 500 videos, where all instance ids are associated over time. It not only supports video panoptic segmentation (VPS) task, but also provides super-set annotations for video semantic segmentation (VSS) and video instance segmentation (VIS) tasks." }, { "dkey": "Cityscapes Panoptic Parts", "dval": "The Cityscapes Panoptic Parts dataset introduces part-aware panoptic segmentation annotations for the Cityscapes dataset. It extends the original panoptic annotations for the Cityscapes dataset with part-level annotations for selected scene-level classes." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "Pascal Panoptic Parts", "dval": "The Pascal Panoptic Parts dataset consists of annotations for the part-aware panoptic segmentation task on the PASCAL VOC 2010 dataset. It is created by merging scene-level labels from PASCAL-Context with part-level labels from PASCAL-Part" }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "I-HAZE", "dval": "The I-Haze dataset contains 25 indoor hazy images (size 2833×4657 pixels) training. It has 5 hazy images for validation along with their corresponding ground truth images." } ]

[ "3D60", "INTERACTION Dataset", "AtariARI", "SWAG" ]

[ "COCO", "ModelNet" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "ModelNet", "dval": "The ModelNet40 dataset contains synthetic object point clouds. As the most widely used benchmark for point cloud analysis, ModelNet40 is popular because of its various categories, clean shapes, well-constructed dataset, etc. The original ModelNet40 consists of 12,311 CAD-generated meshes in 40 categories (such as airplane, car, plant, lamp), of which 9,843 are used for training while the rest 2,468 are reserved for testing. The corresponding point cloud data points are uniformly sampled from the mesh surfaces, and then further preprocessed by moving to the origin and scaling into a unit sphere." }, { "dkey": "3D60", "dval": "Collects high quality 360 datasets with ground truth depth annotations, by re-using recently released large scale 3D datasets and re-purposing them to 360 via rendering." }, { "dkey": "INTERACTION Dataset", "dval": "The INTERACTION dataset contains naturalistic motions of various traffic participants in a variety of highly interactive driving scenarios from different countries. The dataset can serve for many behavior-related research areas, such as \n\n\n1) intention/behavior/motion prediction, \n2) behavior cloning and imitation learning,\n3) behavior analysis and modeling,\n4) motion pattern and representation learning,\n5) interactive behavior extraction and categorization,\n6) social and human-like behavior generation,\n7) decision-making and planning algorithm development and verification,\n8) driving scenario/case generation, etc." }, { "dkey": "AtariARI", "dval": "The AtariARI (Atari Annotated RAM Interface) is an environment for representation learning. The Atari Arcade Learning Environment (ALE) does not explicitly expose any ground truth state information. However, ALE does expose the RAM state (128 bytes per timestep) which are used by the game programmer to store important state information such as the location of sprites, the state of the clock, or the current room the agent is in. To extract these variables, the dataset creators consulted commented disassemblies (or source code) of Atari 2600 games which were made available by Engelhardt and Jentzsch and CPUWIZ. The dataset creators were able to find and verify important state variables for a total of 22 games. Once this information was acquired, combining it with the ALE interface produced a wrapper that can automatically output a state label for every example frame generated from the game. The dataset creators make this available with an easy-to-use gym wrapper, which returns this information with no change to existing code using gym interfaces." }, { "dkey": "SWAG", "dval": "Given a partial description like \"she opened the hood of the car,\" humans can reason about the situation and anticipate what might come next (\"then, she examined the engine\"). SWAG (Situations With Adversarial Generations) is a large-scale dataset for this task of grounded commonsense inference, unifying natural language inference and physically grounded reasoning.\n\nThe dataset consists of 113k multiple choice questions about grounded situations. Each question is a video caption from LSMDC or ActivityNet Captions, with four answer choices about what might happen next in the scene. The correct answer is the (real) video caption for the next event in the video; the three incorrect answers are adversarially generated and human verified, so as to fool machines but not humans. The authors aim for SWAG to be a benchmark for evaluating grounded commonsense NLI and for learning representations." } ]

[ "DUT-OMRON", "Pascal Panoptic Parts", "GLUE", "UDIVA", "TyDiQA-GoldP", "T-LESS", "YouTube-100M" ]

[ "COCO", "ShapeNet" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "ShapeNet", "dval": "ShapeNet is a large scale repository for 3D CAD models developed by researchers from Stanford University, Princeton University and the Toyota Technological Institute at Chicago, USA. The repository contains over 300M models with 220,000 classified into 3,135 classes arranged using WordNet hypernym-hyponym relationships. ShapeNet Parts subset contains 31,693 meshes categorised into 16 common object classes (i.e. table, chair, plane etc.). Each shapes ground truth contains 2-5 parts (with a total of 50 part classes)." }, { "dkey": "DUT-OMRON", "dval": "The DUT-OMRON dataset is used for evaluation of Salient Object Detection task and it contains 5,168 high quality images. The images have one or more salient objects and relatively cluttered background." }, { "dkey": "Pascal Panoptic Parts", "dval": "The Pascal Panoptic Parts dataset consists of annotations for the part-aware panoptic segmentation task on the PASCAL VOC 2010 dataset. It is created by merging scene-level labels from PASCAL-Context with part-level labels from PASCAL-Part" }, { "dkey": "GLUE", "dval": "General Language Understanding Evaluation (GLUE) benchmark is a collection of nine natural language understanding tasks, including single-sentence tasks CoLA and SST-2, similarity and paraphrasing tasks MRPC, STS-B and QQP, and natural language inference tasks MNLI, QNLI, RTE and WNLI." }, { "dkey": "UDIVA", "dval": "UDIVA is a new non-acted dataset of face-to-face dyadic interactions, where interlocutors perform competitive and collaborative tasks with different behavior elicitation and cognitive workload. The dataset consists of 90.5 hours of dyadic interactions among 147 participants distributed in 188 sessions, recorded using multiple audiovisual and physiological sensors. Currently, it includes sociodemographic, self and peer-reported personality, internal state, and relationship profiling from participants." }, { "dkey": "TyDiQA-GoldP", "dval": "TyDiQA is the gold passage version of the Typologically Diverse Question Answering (TyDiWA) dataset, a benchmark for information-seeking question answering, which covers nine languages. The gold passage version is a simplified version of the primary task, which uses only the gold passage as context and excludes unanswerable questions. It is thus similar to XQuAD and MLQA, while being more challenging as questions have been written without seeing the answers, leading to 3× and 2× less lexical overlap compared to XQuAD and MLQA respectively." }, { "dkey": "T-LESS", "dval": "T-LESS is a dataset for estimating the 6D pose, i.e. translation and rotation, of texture-less rigid objects. The dataset features thirty industry-relevant objects with no significant texture and no discriminative color or reflectance properties. The objects exhibit symmetries and mutual similarities in shape and/or size. Compared to other datasets, a unique property is that some of the objects are parts of others. The dataset includes training and test images that were captured with three synchronized sensors, specifically a structured-light and a time-of-flight RGB-D sensor and a high-resolution RGB camera. There are approximately 39K training and 10K test images from each sensor. Additionally, two types of 3D models are provided for each object, i.e. a manually created CAD model and a semi-automatically reconstructed one. Training images depict individual objects against a black background. Test images originate from twenty test scenes having varying complexity, which increases from simple scenes with several isolated objects to very challenging ones with multiple instances of several objects and with a high amount of clutter and occlusion. The images were captured from a systematically sampled view sphere around the object/scene, and are annotated with accurate ground truth 6D poses of all modeled objects." }, { "dkey": "YouTube-100M", "dval": "The YouTube-100M data set consists of 100 million YouTube videos: 70M training videos, 10M evaluation videos, and 20M validation videos. Videos average 4.6 minutes each for a total of 5.4M training hours. Each of these videos is labeled with 1 or more topic identifiers from a set of 30,871 labels. There are an average of around 5 labels per video. The labels are assigned automatically based on a combination of metadata (title, description, comments, etc.), context, and image content for each video. The labels apply to the entire video and range from very generic (e.g. “Song”) to very specific (e.g. “Cormorant”).\nBeing machine generated, the labels are not 100% accurate and of the 30K labels, some are clearly acoustically relevant (“Trumpet”) and others are less so (“Web Page”). Videos often bear annotations with multiple degrees of specificity. For example, videos labeled with “Trumpet” are often labeled “Entertainment” as well, although no hierarchy is enforced." } ]

[ "BlendedMVS", "30MQA", "KITTI", "Make3D" ]

[ "DeepFashion", "CelebA" ]

[ { "dkey": "DeepFashion", "dval": "DeepFashion is a dataset containing around 800K diverse fashion images with their rich annotations (46 categories, 1,000 descriptive attributes, bounding boxes and landmark information) ranging from well-posed product images to real-world-like consumer photos." }, { "dkey": "CelebA", "dval": "CelebFaces Attributes dataset contains 202,599 face images of the size 178×218 from 10,177 celebrities, each annotated with 40 binary labels indicating facial attributes like hair color, gender and age." }, { "dkey": "BlendedMVS", "dval": "BlendedMVS is a novel large-scale dataset, to provide sufficient training ground truth for learning-based MVS. The dataset was created by applying a 3D reconstruction pipeline to recover high-quality textured meshes from images of well-selected scenes. Then, these mesh models were rendered to color images and depth maps." }, { "dkey": "30MQA", "dval": "An enormous question answer pair corpus produced by applying a novel neural network architecture on the knowledge base Freebase to transduce facts into natural language questions." }, { "dkey": "KITTI", "dval": "KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) is one of the most popular datasets for use in mobile robotics and autonomous driving. It consists of hours of traffic scenarios recorded with a variety of sensor modalities, including high-resolution RGB, grayscale stereo cameras, and a 3D laser scanner. Despite its popularity, the dataset itself does not contain ground truth for semantic segmentation. However, various researchers have manually annotated parts of the dataset to fit their necessities. Álvarez et al. generated ground truth for 323 images from the road detection challenge with three classes: road, vertical, and sky. Zhang et al. annotated 252 (140 for training and 112 for testing) acquisitions – RGB and Velodyne scans – from the tracking challenge for ten object categories: building, sky, road, vegetation, sidewalk, car, pedestrian, cyclist, sign/pole, and fence. Ros et al. labeled 170 training images and 46 testing images (from the visual odometry challenge) with 11 classes: building, tree, sky, car, sign, road, pedestrian, fence, pole, sidewalk, and bicyclist." }, { "dkey": "Make3D", "dval": "The Make3D dataset is a monocular Depth Estimation dataset that contains 400 single training RGB and depth map pairs, and 134 test samples. The RGB images have high resolution, while the depth maps are provided at low resolution." } ]

[ "MC-AFP", "VisualMRC", "RACE", "DRCD", "BiPaR" ]

[ "SNLI", "SQuAD" ]

[ { "dkey": "SNLI", "dval": "The SNLI dataset (Stanford Natural Language Inference) consists of 570k sentence-pairs manually labeled as entailment, contradiction, and neutral. Premises are image captions from Flickr30k, while hypotheses were generated by crowd-sourced annotators who were shown a premise and asked to generate entailing, contradicting, and neutral sentences. Annotators were instructed to judge the relation between sentences given that they describe the same event. Each pair is labeled as “entailment”, “neutral”, “contradiction” or “-”, where “-” indicates that an agreement could not be reached." }, { "dkey": "SQuAD", "dval": "The Stanford Question Answering Dataset (SQuAD) is a collection of question-answer pairs derived from Wikipedia articles. In SQuAD, the correct answers of questions can be any sequence of tokens in the given text. Because the questions and answers are produced by humans through crowdsourcing, it is more diverse than some other question-answering datasets. SQuAD 1.1 contains 107,785 question-answer pairs on 536 articles. SQuAD2.0 (open-domain SQuAD, SQuAD-Open), the latest version, combines the 100,000 questions in SQuAD1.1 with over 50,000 un-answerable questions written adversarially by crowdworkers in forms that are similar to the answerable ones." }, { "dkey": "MC-AFP", "dval": "A dataset of around 2 million examples for machine reading-comprehension." }, { "dkey": "VisualMRC", "dval": "VisualMRC is a visual machine reading comprehension dataset that proposes a task: given a question and a document image, a model produces an abstractive answer.\n\nYou can find more details, analyses, and baseline results in the paper, \nVisualMRC: Machine Reading Comprehension on Document Images, AAAI 2021.\n\nStatistics:\n10,197 images\n30,562 QA pairs\n10.53 average question tokens (tokenizing with NLTK tokenizer)\n9.53 average answer tokens (tokenizing wit NLTK tokenizer)\n151.46 average OCR tokens (tokenizing with NLTK tokenizer)" }, { "dkey": "RACE", "dval": "The ReAding Comprehension dataset from Examinations (RACE) dataset is a machine reading comprehension dataset consisting of 27,933 passages and 97,867 questions from English exams, targeting Chinese students aged 12-18. RACE consists of two subsets, RACE-M and RACE-H, from middle school and high school exams, respectively. RACE-M has 28,293 questions and RACE-H has 69,574. Each question is associated with 4 candidate answers, one of which is correct. The data generation process of RACE differs from most machine reading comprehension datasets - instead of generating questions and answers by heuristics or crowd-sourcing, questions in RACE are specifically designed for testing human reading skills, and are created by domain experts." }, { "dkey": "DRCD", "dval": "Delta Reading Comprehension Dataset (DRCD) is an open domain traditional Chinese machine reading comprehension (MRC) dataset. This dataset aimed to be a standard Chinese machine reading comprehension dataset, which can be a source dataset in transfer learning. The dataset contains 10,014 paragraphs from 2,108 Wikipedia articles and 30,000+ questions generated by annotators." }, { "dkey": "BiPaR", "dval": "BiPaR is a manually annotated bilingual parallel novel-style machine reading comprehension (MRC) dataset, developed to support monolingual, multilingual and cross-lingual reading comprehension on novels. The biggest difference between BiPaR and existing reading comprehension datasets is that each triple (Passage, Question, Answer) in BiPaR is written in parallel in two languages. BiPaR is diverse in prefixes of questions, answer types and relationships between questions and passages. Answering the questions requires reading comprehension skills of coreference resolution, multi-sentence reasoning, and understanding of implicit causality." } ]

[ "THEODORE", "CREMA-D", "UASOL", "ReCAM", "DOTmark", "GSL" ]

[ "SNLI", "MultiNLI" ]

[ { "dkey": "SNLI", "dval": "The SNLI dataset (Stanford Natural Language Inference) consists of 570k sentence-pairs manually labeled as entailment, contradiction, and neutral. Premises are image captions from Flickr30k, while hypotheses were generated by crowd-sourced annotators who were shown a premise and asked to generate entailing, contradicting, and neutral sentences. Annotators were instructed to judge the relation between sentences given that they describe the same event. Each pair is labeled as “entailment”, “neutral”, “contradiction” or “-”, where “-” indicates that an agreement could not be reached." }, { "dkey": "MultiNLI", "dval": "The Multi-Genre Natural Language Inference (MultiNLI) dataset has 433K sentence pairs. Its size and mode of collection are modeled closely like SNLI. MultiNLI offers ten distinct genres (Face-to-face, Telephone, 9/11, Travel, Letters, Oxford University Press, Slate, Verbatim, Goverment and Fiction) of written and spoken English data. There are matched dev/test sets which are derived from the same sources as those in the training set, and mismatched sets which do not closely resemble any seen at training time." }, { "dkey": "THEODORE", "dval": "Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset." }, { "dkey": "CREMA-D", "dval": "CREMA-D is an emotional multimodal actor data set of 7,442 original clips from 91 actors. These clips were from 48 male and 43 female actors between the ages of 20 and 74 coming from a variety of races and ethnicities (African America, Asian, Caucasian, Hispanic, and Unspecified).\n\nActors spoke from a selection of 12 sentences. The sentences were presented using one of six different emotions (Anger, Disgust, Fear, Happy, Neutral, and Sad) and four different emotion levels (Low, Medium, High, and Unspecified).\n\nParticipants rated the emotion and emotion levels based on the combined audiovisual presentation, the video alone, and the audio alone. Due to the large number of ratings needed, this effort was crowd-sourced and a total of 2443 participants each rated 90 unique clips, 30 audio, 30 visual, and 30 audio-visual. 95% of the clips have more than 7 ratings." }, { "dkey": "UASOL", "dval": "The UASOL an RGB-D stereo dataset, that contains 160902 frames, filmed at 33 different scenes, each with between 2 k and 10 k frames. The frames show different paths from the perspective of a pedestrian, including sidewalks, trails, roads, etc. The images were extracted from video files with 15 fps at HD2K resolution with a size of 2280 × 1282 pixels. The dataset also provides a GPS geolocalization tag for each second of the sequences and reflects different climatological conditions. It also involved up to 4 different persons filming the dataset at different moments of the day.\n\nWe propose a train, validation and test split to train the network. \nAdditionally, we introduce a subset of 676 pairs of RGB Stereo images and their respective depth, which we extracted randomly from the entire dataset. This given test set is introduced to make comparability possible between the different methods trained with the dataset." }, { "dkey": "ReCAM", "dval": "Tasks\nOur shared task has three subtasks. Subtask 1 and 2 focus on evaluating machine learning models' performance with regard to two definitions of abstractness (Spreen and Schulz, 1966; Changizi, 2008), which we call imperceptibility and nonspecificity, respectively. Subtask 3 aims to provide some insights to their relationships.\n\n• Subtask 1: ReCAM-Imperceptibility\n\nConcrete words refer to things, events, and properties that we can perceive directly with our senses (Spreen and Schulz, 1966; Coltheart 1981; Turney et al., 2011), e.g., donut, trees, and red. In contrast, abstract words refer to ideas and concepts that are distant from immediate perception. Examples include objective, culture, and economy. In subtask 1, the participanting systems are required to perform reading comprehension of abstract meaning for imperceptible concepts.\n\nBelow is an example. Given a passage and a question, your model needs to choose from the five candidates the best one for replacing @placeholder.\n\n• Subtask 2: ReCAM-Nonspecificity\n\nSubtask 2 focuses on a different type of definition. Compared to concrete concepts like groundhog and whale, hypernyms such as vertebrate are regarded as more abstract (Changizi, 2008). \n\n• Subtask 3: ReCAM-Intersection\nSubtask 3 aims to provide more insights to the relationship of the two views on abstractness, In this subtask, we test the performance of a system that is trained on one definition and evaluted on the other." }, { "dkey": "DOTmark", "dval": "DOTmark is a benchmark for discrete optimal transport, which is designed to serve as a neutral collection of problems, where discrete optimal transport methods can be tested, compared to one another, and brought to their limits on large-scale instances. It consists of a variety of grayscale images, in various resolutions and classes, such as several types of randomly generated images, classical test images and real data from microscopy." }, { "dkey": "GSL", "dval": "Dataset Description\nThe Greek Sign Language (GSL) is a large-scale RGB+D dataset, suitable for Sign Language Recognition (SLR) and Sign Language Translation (SLT). The video captures are conducted using an Intel RealSense D435 RGB+D camera at a rate of 30 fps. Both the RGB and the depth streams are acquired in the same spatial resolution of 848×480 pixels. To increase variability in the videos, the camera position and orientation is slightly altered within subsequent recordings. Seven different signers are employed to perform 5 individual and commonly met scenarios in different public services. The average length of each scenario is twenty sentences.\n\nThe dataset contains 10,290 sentence instances, 40,785 gloss instances, 310 unique glosses (vocabulary size) and 331 unique sentences, with 4.23 glosses per sentence on average. Each signer is asked to perform the pre-defined dialogues five consecutive times. In all cases, the simulation considers a deaf person communicating with a single public service employee. The involved signer performs the sequence of glosses of both agents in the discussion. For the annotation of each gloss sequence, GSL linguistic experts are involved. The given annotations are at individual gloss and gloss sequence level. A translation of the gloss sentences to spoken Greek is also provided.\n\nEvaluation\nThe GSL dataset includes the 3 evaluation setups:\n\n\n\nSigner-dependent continuous sign language recognition (GSL SD) – roughly 80% of videos are used for training, corresponding to 8,189 instances. The rest 1,063 (10%) were kept for validation and 1,043 (10%) for testing.\n\n\n\nSigner-independent continuous sign language recognition (GSL SI) – the selected test gloss sequences are not used in the training set, while all the individual glosses exist in the training set. In GSL SI, the recordings of one signer are left out for validation and testing (588 and 881 instances, respectively). The rest 8821 instances are utilized for training.\n\n\n\nIsolated gloss sign language recognition (GSL isol.) – The validation set consists of 2,231 gloss instances, the test set 3,500, while the remaining 34,995 are used for training. All 310 unique glosses are seen in the training set.\n\n\n\nFor more info and results, advice our paper\n\nPaper Abstract: A Comprehensive Study on Sign Language Recognition Methods, Adaloglou et al. 2020\nIn this paper, a comparative experimental assessment of computer vision-based methods for sign language recognition is conducted. By implementing the most recent deep neural network methods in this field, a thorough evaluation on multiple publicly available datasets is performed. The aim of the present study is to provide insights on sign language recognition, focusing on mapping non-segmented video streams to glosses. For this task, two new sequence training criteria, known from the fields of speech and scene text recognition, are introduced. Furthermore, a\nplethora of pretraining schemes are thoroughly discussed. Finally, a new RGB+D dataset for the Greek sign language is created. To the best of our knowledge, this is the first sign language dataset where sentence and gloss level annotations are provided for every video capture.\n\nArxiv link" } ]

[ "Cityscapes-VPS", "Pascal Panoptic Parts", "Cityscapes Panoptic Parts", "SNIPS", "ConvAI2", "I-HAZE" ]

[ "COCO", "Cityscapes" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "Cityscapes", "dval": "Cityscapes is a large-scale database which focuses on semantic understanding of urban street scenes. It provides semantic, instance-wise, and dense pixel annotations for 30 classes grouped into 8 categories (flat surfaces, humans, vehicles, constructions, objects, nature, sky, and void). The dataset consists of around 5000 fine annotated images and 20000 coarse annotated ones. Data was captured in 50 cities during several months, daytimes, and good weather conditions. It was originally recorded as video so the frames were manually selected to have the following features: large number of dynamic objects, varying scene layout, and varying background." }, { "dkey": "Cityscapes-VPS", "dval": "Cityscapes-VPS is a video extension of the Cityscapes validation split. It provides 2500-frame panoptic labels that temporally extend the 500 Cityscapes image-panoptic labels. There are total 3000-frame panoptic labels which correspond to 5, 10, 15, 20, 25, and 30th frames of each 500 videos, where all instance ids are associated over time. It not only supports video panoptic segmentation (VPS) task, but also provides super-set annotations for video semantic segmentation (VSS) and video instance segmentation (VIS) tasks." }, { "dkey": "Pascal Panoptic Parts", "dval": "The Pascal Panoptic Parts dataset consists of annotations for the part-aware panoptic segmentation task on the PASCAL VOC 2010 dataset. It is created by merging scene-level labels from PASCAL-Context with part-level labels from PASCAL-Part" }, { "dkey": "Cityscapes Panoptic Parts", "dval": "The Cityscapes Panoptic Parts dataset introduces part-aware panoptic segmentation annotations for the Cityscapes dataset. It extends the original panoptic annotations for the Cityscapes dataset with part-level annotations for selected scene-level classes." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "I-HAZE", "dval": "The I-Haze dataset contains 25 indoor hazy images (size 2833×4657 pixels) training. It has 5 hazy images for validation along with their corresponding ground truth images." } ]

[ "STL-10", "REDDIT-12K", "MVTecAD", "CC100", "SuperGLUE" ]

[ "ImageNet", "Oxford5k" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "Oxford5k", "dval": "Oxford5K is the Oxford Buildings Dataset, which contains 5062 images collected from Flickr. It offers a set of 55 queries for 11 landmark buildings, five for each landmark." }, { "dkey": "STL-10", "dval": "The STL-10 is an image dataset derived from ImageNet and popularly used to evaluate algorithms of unsupervised feature learning or self-taught learning. Besides 100,000 unlabeled images, it contains 13,000 labeled images from 10 object classes (such as birds, cats, trucks), among which 5,000 images are partitioned for training while the remaining 8,000 images for testing. All the images are color images with 96×96 pixels in size." }, { "dkey": "REDDIT-12K", "dval": "Reddit12k contains 11929 graphs each corresponding to an online discussion thread where nodes represent users, and an edge represents the fact that one of the two users responded to the comment of the other user. There is 1 of 11 graph labels associated with each of these 11929 discussion graphs, representing the category of the community." }, { "dkey": "MVTecAD", "dval": "MVTec AD is a dataset for benchmarking anomaly detection methods with a focus on industrial inspection. It contains over 5000 high-resolution images divided into fifteen different object and texture categories. Each category comprises a set of defect-free training images and a test set of images with various kinds of defects as well as images without defects.\n\nThere are two common metrics: Detection AUROC and Segmentation (or pixelwise) AUROC\n\nDetection (or, classification) methods output single float (anomaly score) per input test image. \n\nSegmentation methods output anomaly probability for each pixel. \n\"To assess segmentation performance, we evaluate the relative per-region overlap of the segmentation with the ground truth. To get an additional performance measure that is independent of the determined threshold, we compute the area under the receiver operating characteristic curve (ROC AUC). We define the true positive rate as the percentage of pixels that were correctly classified as anomalous\" [1]\nLater segmentation metric was improved to balance regions with small and large area, see PRO-AUC and other in [2]\n\n[1] Paul Bergmann et al, \"MVTec AD — A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection\"\n[2] Bergmann, P., Batzner, K., Fauser, M. et al. The MVTec Anomaly Detection Dataset: A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection. Int J Comput Vis (2021). https://doi.org/10.1007/s11263-020-01400-4" }, { "dkey": "CC100", "dval": "This corpus comprises of monolingual data for 100+ languages and also includes data for romanized languages. This was constructed using the urls and paragraph indices provided by the CC-Net repository by processing January-December 2018 Commoncrawl snapshots. Each file comprises of documents separated by double-newlines and paragraphs within the same document separated by a newline. The data is generated using the open source CC-Net repository." }, { "dkey": "SuperGLUE", "dval": "SuperGLUE is a benchmark dataset designed to pose a more rigorous test of language understanding than GLUE. SuperGLUE has the same high-level motivation as GLUE: to provide a simple, hard-to-game measure of progress toward general-purpose language understanding technologies for English. SuperGLUE follows the basic design of GLUE: It consists of a public leaderboard built around eight language understanding tasks, drawing on existing data, accompanied by a single-number\nperformance metric, and an analysis toolkit. However, it improves upon GLUE in several ways:\n\n\nMore challenging tasks: SuperGLUE retains the two hardest tasks in GLUE. The remaining tasks were identified from those submitted to an open call for task proposals and were selected based on difficulty for current NLP approaches.\nMore diverse task formats: The task formats in GLUE are limited to sentence- and sentence-pair classification. The authors expand the set of task formats in SuperGLUE to include\ncoreference resolution and question answering (QA).\nComprehensive human baselines: the authors include human performance estimates for all benchmark tasks, which verify that substantial headroom exists between a strong BERT-based baseline and human performance.\nImproved code support: SuperGLUE is distributed with a new, modular toolkit for work on pretraining, multi-task learning, and transfer learning in NLP, built around standard tools including PyTorch (Paszke et al., 2017) and AllenNLP (Gardner et al., 2017).\nRefined usage rules: The conditions for inclusion on the SuperGLUE leaderboard were revamped to ensure fair competition, an informative leaderboard, and full credit\nassignment to data and task creators." } ]

[ "ACDC", "ConvAI2", "WikiReading", "ROCStories", "THEODORE" ]

[ "KITTI", "Cityscapes" ]

[ { "dkey": "KITTI", "dval": "KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) is one of the most popular datasets for use in mobile robotics and autonomous driving. It consists of hours of traffic scenarios recorded with a variety of sensor modalities, including high-resolution RGB, grayscale stereo cameras, and a 3D laser scanner. Despite its popularity, the dataset itself does not contain ground truth for semantic segmentation. However, various researchers have manually annotated parts of the dataset to fit their necessities. Álvarez et al. generated ground truth for 323 images from the road detection challenge with three classes: road, vertical, and sky. Zhang et al. annotated 252 (140 for training and 112 for testing) acquisitions – RGB and Velodyne scans – from the tracking challenge for ten object categories: building, sky, road, vegetation, sidewalk, car, pedestrian, cyclist, sign/pole, and fence. Ros et al. labeled 170 training images and 46 testing images (from the visual odometry challenge) with 11 classes: building, tree, sky, car, sign, road, pedestrian, fence, pole, sidewalk, and bicyclist." }, { "dkey": "Cityscapes", "dval": "Cityscapes is a large-scale database which focuses on semantic understanding of urban street scenes. It provides semantic, instance-wise, and dense pixel annotations for 30 classes grouped into 8 categories (flat surfaces, humans, vehicles, constructions, objects, nature, sky, and void). The dataset consists of around 5000 fine annotated images and 20000 coarse annotated ones. Data was captured in 50 cities during several months, daytimes, and good weather conditions. It was originally recorded as video so the frames were manually selected to have the following features: large number of dynamic objects, varying scene layout, and varying background." }, { "dkey": "ACDC", "dval": "The goal of the Automated Cardiac Diagnosis Challenge (ACDC) challenge is to:\n\n\ncompare the performance of automatic methods on the segmentation of the left ventricular endocardium and epicardium as the right ventricular endocardium for both end diastolic and end systolic phase instances;\ncompare the performance of automatic methods for the classification of the examinations in five classes (normal case, heart failure with infarction, dilated cardiomyopathy, hypertrophic cardiomyopathy, abnormal right ventricle).\n\nThe overall ACDC dataset was created from real clinical exams acquired at the University Hospital of Dijon. Acquired data were fully anonymized and handled within the regulations set by the local ethical committee of the Hospital of Dijon (France). Our dataset covers several well-defined pathologies with enough cases to (1) properly train machine learning methods and (2) clearly assess the variations of the main physiological parameters obtained from cine-MRI (in particular diastolic volume and ejection fraction). The dataset is composed of 150 exams (all from different patients) divided into 5 evenly distributed subgroups (4 pathological plus 1 healthy subject groups) as described below. Furthermore, each patient comes with the following additional information : weight, height, as well as the diastolic and systolic phase instants.\n\nThe database is made available to participants through two datasets from the dedicated online evaluation website after a personal registration: i) a training dataset of 100 patients along with the corresponding manual references based on the analysis of one clinical expert; ii) a testing dataset composed of 50 new patients, without manual annotations but with the patient information given above. The raw input images are provided through the Nifti format." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "WikiReading", "dval": "WikiReading is a large-scale natural language understanding task and publicly-available dataset with 18 million instances. The task is to predict textual values from the structured knowledge base Wikidata by reading the text of the corresponding Wikipedia articles. The task contains a rich variety of challenging classification and extraction sub-tasks, making it well-suited for end-to-end models such as deep neural networks (DNNs)." }, { "dkey": "ROCStories", "dval": "ROCStories is a collection of commonsense short stories. The corpus consists of 100,000 five-sentence stories. Each story logically follows everyday topics created by Amazon Mechanical Turk workers. These stories contain a variety of commonsense causal and temporal relations between everyday events. Writers also develop an additional 3,742 Story Cloze Test stories which contain a four-sentence-long body and two candidate endings. The endings were collected by asking Mechanical Turk workers to write both a right ending and a wrong ending after eliminating original endings of given short stories. Both endings were required to make logical sense and include at least one character from the main story line. The published ROCStories dataset is constructed with ROCStories as a training set that includes 98,162 stories that exclude candidate wrong endings, an evaluation set, and a test set, which have the same structure (1 body + 2 candidate endings) and a size of 1,871." }, { "dkey": "THEODORE", "dval": "Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset." } ]

[ "GSL", "CUB-200-2011", "ETH Py150 Open", "MVTecAD" ]

[ "SNLI", "MRPC", "GLUE" ]

[ { "dkey": "SNLI", "dval": "The SNLI dataset (Stanford Natural Language Inference) consists of 570k sentence-pairs manually labeled as entailment, contradiction, and neutral. Premises are image captions from Flickr30k, while hypotheses were generated by crowd-sourced annotators who were shown a premise and asked to generate entailing, contradicting, and neutral sentences. Annotators were instructed to judge the relation between sentences given that they describe the same event. Each pair is labeled as “entailment”, “neutral”, “contradiction” or “-”, where “-” indicates that an agreement could not be reached." }, { "dkey": "MRPC", "dval": "Microsoft Research Paraphrase Corpus (MRPC) is a corpus consists of 5,801 sentence pairs collected from newswire articles. Each pair is labelled if it is a paraphrase or not by human annotators. The whole set is divided into a training subset (4,076 sentence pairs of which 2,753 are paraphrases) and a test subset (1,725 pairs of which 1,147 are paraphrases)." }, { "dkey": "GLUE", "dval": "General Language Understanding Evaluation (GLUE) benchmark is a collection of nine natural language understanding tasks, including single-sentence tasks CoLA and SST-2, similarity and paraphrasing tasks MRPC, STS-B and QQP, and natural language inference tasks MNLI, QNLI, RTE and WNLI." }, { "dkey": "GSL", "dval": "Dataset Description\nThe Greek Sign Language (GSL) is a large-scale RGB+D dataset, suitable for Sign Language Recognition (SLR) and Sign Language Translation (SLT). The video captures are conducted using an Intel RealSense D435 RGB+D camera at a rate of 30 fps. Both the RGB and the depth streams are acquired in the same spatial resolution of 848×480 pixels. To increase variability in the videos, the camera position and orientation is slightly altered within subsequent recordings. Seven different signers are employed to perform 5 individual and commonly met scenarios in different public services. The average length of each scenario is twenty sentences.\n\nThe dataset contains 10,290 sentence instances, 40,785 gloss instances, 310 unique glosses (vocabulary size) and 331 unique sentences, with 4.23 glosses per sentence on average. Each signer is asked to perform the pre-defined dialogues five consecutive times. In all cases, the simulation considers a deaf person communicating with a single public service employee. The involved signer performs the sequence of glosses of both agents in the discussion. For the annotation of each gloss sequence, GSL linguistic experts are involved. The given annotations are at individual gloss and gloss sequence level. A translation of the gloss sentences to spoken Greek is also provided.\n\nEvaluation\nThe GSL dataset includes the 3 evaluation setups:\n\n\n\nSigner-dependent continuous sign language recognition (GSL SD) – roughly 80% of videos are used for training, corresponding to 8,189 instances. The rest 1,063 (10%) were kept for validation and 1,043 (10%) for testing.\n\n\n\nSigner-independent continuous sign language recognition (GSL SI) – the selected test gloss sequences are not used in the training set, while all the individual glosses exist in the training set. In GSL SI, the recordings of one signer are left out for validation and testing (588 and 881 instances, respectively). The rest 8821 instances are utilized for training.\n\n\n\nIsolated gloss sign language recognition (GSL isol.) – The validation set consists of 2,231 gloss instances, the test set 3,500, while the remaining 34,995 are used for training. All 310 unique glosses are seen in the training set.\n\n\n\nFor more info and results, advice our paper\n\nPaper Abstract: A Comprehensive Study on Sign Language Recognition Methods, Adaloglou et al. 2020\nIn this paper, a comparative experimental assessment of computer vision-based methods for sign language recognition is conducted. By implementing the most recent deep neural network methods in this field, a thorough evaluation on multiple publicly available datasets is performed. The aim of the present study is to provide insights on sign language recognition, focusing on mapping non-segmented video streams to glosses. For this task, two new sequence training criteria, known from the fields of speech and scene text recognition, are introduced. Furthermore, a\nplethora of pretraining schemes are thoroughly discussed. Finally, a new RGB+D dataset for the Greek sign language is created. To the best of our knowledge, this is the first sign language dataset where sentence and gloss level annotations are provided for every video capture.\n\nArxiv link" }, { "dkey": "CUB-200-2011", "dval": "The Caltech-UCSD Birds-200-2011 (CUB-200-2011) dataset is the most widely-used dataset for fine-grained visual categorization task. It contains 11,788 images of 200 subcategories belonging to birds, 5,994 for training and 5,794 for testing. Each image has detailed annotations: 1 subcategory label, 15 part locations, 312 binary attributes and 1 bounding box. The textual information comes from Reed et al.. They expand the CUB-200-2011 dataset by collecting fine-grained natural language descriptions. Ten single-sentence descriptions are collected for each image. The natural language descriptions are collected through the Amazon Mechanical Turk (AMT) platform, and are required at least 10 words, without any information of subcategories and actions." }, { "dkey": "ETH Py150 Open", "dval": "A massive, deduplicated corpus of 7.4M Python files from GitHub." }, { "dkey": "MVTecAD", "dval": "MVTec AD is a dataset for benchmarking anomaly detection methods with a focus on industrial inspection. It contains over 5000 high-resolution images divided into fifteen different object and texture categories. Each category comprises a set of defect-free training images and a test set of images with various kinds of defects as well as images without defects.\n\nThere are two common metrics: Detection AUROC and Segmentation (or pixelwise) AUROC\n\nDetection (or, classification) methods output single float (anomaly score) per input test image. \n\nSegmentation methods output anomaly probability for each pixel. \n\"To assess segmentation performance, we evaluate the relative per-region overlap of the segmentation with the ground truth. To get an additional performance measure that is independent of the determined threshold, we compute the area under the receiver operating characteristic curve (ROC AUC). We define the true positive rate as the percentage of pixels that were correctly classified as anomalous\" [1]\nLater segmentation metric was improved to balance regions with small and large area, see PRO-AUC and other in [2]\n\n[1] Paul Bergmann et al, \"MVTec AD — A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection\"\n[2] Bergmann, P., Batzner, K., Fauser, M. et al. The MVTec Anomaly Detection Dataset: A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection. Int J Comput Vis (2021). https://doi.org/10.1007/s11263-020-01400-4" } ]

[ "DukeMTMC-attribute", "JAAD", "TITAN", "MultiviewX", "VisDrone", "ISBDA" ]

[ "PETA", "MARS" ]

[ { "dkey": "PETA", "dval": "The PEdesTrian Attribute dataset (PETA) is a dataset fore recognizing pedestrian attributes, such as gender and clothing style, at a far distance. It is of interest in video surveillance scenarios where face and body close-shots and hardly available. It consists of 19,000 pedestrian images with 65 attributes (61 binary and 4 multi-class). Those images contain 8705 persons." }, { "dkey": "MARS", "dval": "MARS (Motion Analysis and Re-identification Set) is a large scale video based person reidentification dataset, an extension of the Market-1501 dataset. It has been collected from six near-synchronized cameras. It consists of 1,261 different pedestrians, who are captured by at least 2 cameras. The variations in poses, colors and illuminations of pedestrians, as well as the poor image quality, make it very difficult to yield high matching accuracy. Moreover, the dataset contains 3,248 distractors in order to make it more realistic. Deformable Part Model and GMMCP tracker were used to automatically generate the tracklets (mostly 25-50 frames long)." }, { "dkey": "DukeMTMC-attribute", "dval": "The images in DukeMTMC-attribute dataset comes from Duke University. There are 1812 identities and 34183 annotated bounding boxes in the DukeMTMC-attribute dataset. This dataset contains 702 identities for training and 1110 identities for testing, corresponding to 16522 and 17661 images respectively. The attributes are annotated in the identity level, every image in this dataset is annotated with 23 attributes.\n\nNOTE: This dataset has been retracted." }, { "dkey": "JAAD", "dval": "JAAD is a dataset for studying joint attention in the context of autonomous driving. The focus is on pedestrian and driver behaviors at the point of crossing and factors that influence them. To this end, JAAD dataset provides a richly annotated collection of 346 short video clips (5-10 sec long) extracted from over 240 hours of driving footage. These videos filmed in several locations in North America and Eastern Europe represent scenes typical for everyday urban driving in various weather conditions.\n\nBounding boxes with occlusion tags are provided for all pedestrians making this dataset suitable for pedestrian detection.\n\nBehavior annotations specify behaviors for pedestrians that interact with or require attention of the driver. For each video there are several tags (weather, locations, etc.) and timestamped behavior labels from a fixed list (e.g. stopped, walking, looking, etc.). In addition, a list of demographic attributes is provided for each pedestrian (e.g. age, gender, direction of motion, etc.) as well as a list of visible traffic scene elements (e.g. stop sign, traffic signal, etc.) for each frame.\n\nPaper: Are They Going to Cross? A Benchmark Dataset and Baseline for Pedestrian Crosswalk Behavior" }, { "dkey": "TITAN", "dval": "TITAN consists of 700 labeled video-clips (with odometry) captured from a moving vehicle on highly interactive urban traffic scenes in Tokyo. The dataset includes 50 labels including vehicle states and actions, pedestrian age groups, and targeted pedestrian action attributes that are organized hierarchically corresponding to atomic, simple/complex-contextual, transportive, and communicative actions." }, { "dkey": "MultiviewX", "dval": "MultiviewX is a synthetic Multiview pedestrian detection dataset. It is build using pedestrian models from PersonX, in Unity.\nThe MultiviewX dataset covers a square of 16 meters by 25 meters. The ground plane is quantized into a 640x1000 grid. There are 6 cameras with overlapping field-of-view in the MultiviewX dataset, each of which outputs a 1080x1920 resolution image. On average, 4.41 cameras are covering the same location." }, { "dkey": "VisDrone", "dval": "VisDrone is a large-scale benchmark with carefully annotated ground-truth for various important computer vision tasks, to make vision meet drones. The VisDrone2019 dataset is collected by the AISKYEYE team at Lab of Machine Learning and Data Mining, Tianjin University, China. The benchmark dataset consists of 288 video clips formed by 261,908 frames and 10,209 static images, captured by various drone-mounted cameras, covering a wide range of aspects including location (taken from 14 different cities separated by thousands of kilometers in China), environment (urban and country), objects (pedestrian, vehicles, bicycles, etc.), and density (sparse and crowded scenes). Note that, the dataset was collected using various drone platforms (i.e., drones with different models), in different scenarios, and under various weather and lighting conditions. These frames are manually annotated with more than 2.6 million bounding boxes of targets of frequent interests, such as pedestrians, cars, bicycles, and tricycles. Some important attributes including scene visibility, object class and occlusion, are also provided for better data utilization." }, { "dkey": "ISBDA", "dval": "Consists of user-generated aerial videos from social media with annotations of instance-level building damage masks. This provides the first benchmark for quantitative evaluation of models to assess building damage using aerial videos." } ]

[ "300-VW", "AFLW2000-3D", "AFEW-VA", "LS3D-W", "Oulu-CASIA" ]

[ "COFW", "AFW" ]

[ { "dkey": "COFW", "dval": "The Caltech Occluded Faces in the Wild (COFW) dataset is designed to present faces in real-world conditions. Faces show large variations in shape and occlusions due to differences in pose, expression, use of accessories such as sunglasses and hats and interactions with objects (e.g. food, hands, microphones,
etc.). All images were hand annotated using the same 29 landmarks as in LFPW. Both the landmark positions as well as their occluded/unoccluded state were annotated. The faces are occluded to different degrees, with large variations in the type of occlusions encountered. COFW has an average occlusion of over 23." }, { "dkey": "AFW", "dval": "AFW (Annotated Faces in the Wild) is a face detection dataset that contains 205 images with 468 faces. Each face image is labeled with at most 6 landmarks with visibility labels, as well as a bounding box." }, { "dkey": "300-VW", "dval": "300 Videos in the Wild (300-VW) is a dataset for evaluating facial landmark tracking algorithms in the wild. The dataset authors collected a large number of long facial videos recorded in the wild. Each video has duration of ~1 minute (at 25-30 fps). All frames have been annotated with regards to the same mark-up (i.e. set of facial landmarks) used in the 300 W competition as well (a total of 68 landmarks). The dataset includes 114 videos (circa 1 min each)." }, { "dkey": "AFLW2000-3D", "dval": "AFLW2000-3D is a dataset of 2000 images that have been annotated with image-level 68-point 3D facial landmarks. This dataset is used for evaluation of 3D facial landmark detection models. The head poses are very diverse and often hard to be detected by a CNN-based face detector." }, { "dkey": "AFEW-VA", "dval": "The AFEW-VA databaset is a collection of highly accurate per-frame annotations levels of valence and arousal, along with per-frame annotations of 68 facial landmarks for 600 challenging video clips. These clips are extracted from feature films and were also annotated in terms of discrete emotion categories in the form of the AFEW database (that can be obtained there)." }, { "dkey": "LS3D-W", "dval": "A 3D facial landmark dataset of around 230,000 images." }, { "dkey": "Oulu-CASIA", "dval": "The Oulu-CASIA NIR&VIS facial expression database consists of six expressions (surprise, happiness, sadness, anger, fear and disgust) from 80 people between 23 and 58 years old. 73.8% of the subjects are males. The subjects were asked to sit on a chair in the observation room in a way that he/ she is in front of camera. Camera-face distance is about 60 cm. Subjects were asked to make a facial expression according to an expression example shown in picture sequences. The imaging hardware works at the rate of 25 frames per second and the image resolution is 320 × 240 pixels." } ]

[ "ConvAI2", "MultiRC", "Libri-Adapt", "EPIC-KITCHENS-100", "Dialogue State Tracking Challenge" ]

[ "BP4D", "DISFA" ]

[ { "dkey": "BP4D", "dval": "The BP4D-Spontaneous dataset is a 3D video database of spontaneous facial expressions in a diverse group of young adults. Well-validated emotion inductions were used to elicit expressions of emotion and paralinguistic communication. Frame-level ground-truth for facial actions was obtained using the Facial Action Coding System. Facial features were tracked in both 2D and 3D domains using both person-specific and generic approaches.\nThe database includes forty-one participants (23 women, 18 men). They were 18 – 29 years of age; 11 were Asian, 6 were African-American, 4 were Hispanic, and 20 were Euro-American. An emotion elicitation protocol was designed to elicit emotions of participants effectively. Eight tasks were covered with an interview process and a series of activities to elicit eight emotions.\nThe database is structured by participants. Each participant is associated with 8 tasks. For each task, there are both 3D and 2D videos. As well, the Metadata include manually annotated action units (FACS AU), automatically tracked head pose, and 2D/3D facial landmarks. The database is in the size of about 2.6TB (without compression)." }, { "dkey": "DISFA", "dval": "The Denver Intensity of Spontaneous Facial Action (DISFA) dataset consists of 27 videos of 4844 frames each, with 130,788 images in total. Action unit annotations are on different levels of intensity, which are ignored in the following experiments and action units are either set or unset. DISFA was selected from a wider range of databases popular in the field of facial expression recognition because of the high number of smiles, i.e. action unit 12. In detail, 30,792 have this action unit set, 82,176 images have some action unit(s) set and 48,612 images have no action unit(s) set at all." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "MultiRC", "dval": "MultiRC (Multi-Sentence Reading Comprehension) is a dataset of short paragraphs and multi-sentence questions, i.e., questions that can be answered by combining information from multiple sentences of the paragraph.\nThe dataset was designed with three key challenges in mind:\n* The number of correct answer-options for each question is not pre-specified. This removes the over-reliance on answer-options and forces them to decide on the correctness of each candidate answer independently of others. In other words, the task is not to simply identify the best answer-option, but to evaluate the correctness of each answer-option individually.\n* The correct answer(s) is not required to be a span in the text.\n* The paragraphs in the dataset have diverse provenance by being extracted from 7 different domains such as news, fiction, historical text etc., and hence are expected to be more diverse in their contents as compared to single-domain datasets.\nThe entire corpus consists of around 10K questions (including about 6K multiple-sentence questions). The 60% of the data is released as training and development data. The rest of the data is saved for evaluation and every few months a new unseen additional data is included for evaluation to prevent unintentional overfitting over time." }, { "dkey": "Libri-Adapt", "dval": "Libri-Adapt aims to support unsupervised domain adaptation research on speech recognition models." }, { "dkey": "EPIC-KITCHENS-100", "dval": "This paper introduces the pipeline to scale the largest dataset in egocentric vision EPIC-KITCHENS. The effort culminates in EPIC-KITCHENS-100, a collection of 100 hours, 20M frames, 90K actions in 700 variable-length videos, capturing long-term unscripted activities in 45 environments, using head-mounted cameras. Compared to its previous version (EPIC-KITCHENS-55), EPIC-KITCHENS-100 has been annotated using a novel pipeline that allows denser (54% more actions per minute) and more complete annotations of fine-grained actions (+128% more action segments). This collection also enables evaluating the \"test of time\" - i.e. whether models trained on data collected in 2018 can generalise to new footage collected under the same hypotheses albeit \"two years on\".\nThe dataset is aligned with 6 challenges: action recognition (full and weak supervision), action detection, action anticipation, cross-modal retrieval (from captions), as well as unsupervised domain adaptation for action recognition. For each challenge, we define the task, provide baselines and evaluation metrics." }, { "dkey": "Dialogue State Tracking Challenge", "dval": "The Dialog State Tracking Challenges 2 & 3 (DSTC2&3) were research challenge focused on improving the state of the art in tracking the state of spoken dialog systems. State tracking, sometimes called belief tracking, refers to accurately estimating the user's goal as a dialog progresses. Accurate state tracking is desirable because it provides robustness to errors in speech recognition, and helps reduce ambiguity inherent in language within a temporal process like dialog.\nIn these challenges, participants were given labelled corpora of dialogs to develop state tracking algorithms. The trackers were then evaluated on a common set of held-out dialogs, which were released, un-labelled, during a one week period.\n\nThe corpus was collected using Amazon Mechanical Turk, and consists of dialogs in two domains: restaurant information, and tourist information. Tourist information subsumes restaurant information, and includes bars, cafés etc. as well as multiple new slots. There were two rounds of evaluation using this data:\n\nDSTC 2 released a large number of training dialogs related to restaurant search. Compared to DSTC (which was in the bus timetables domain), DSTC 2 introduces changing user goals, tracking 'requested slots' as well as the new restaurants domain. Results from DSTC 2 were presented at SIGDIAL 2014.\nDSTC 3 addressed the problem of adaption to a new domain - tourist information. DSTC 3 releases a small amount of labelled data in the tourist information domain; participants will use this data plus the restaurant data from DSTC 2 for training.\nDialogs used for training are fully labelled; user transcriptions, user dialog-act semantics and dialog state are all annotated. (This corpus therefore is also suitable for studies in Spoken Language Understanding.)" } ]

[ "A2D", "TrashCan", "CIHP", "Cityscapes-VPS", "TTPLA" ]

[ "SBD", "Cityscapes" ]

[ { "dkey": "SBD", "dval": "The Semantic Boundaries Dataset (SBD) is a dataset for predicting pixels on the boundary of the object (as opposed to the inside of the object with semantic segmentation). The dataset consists of 11318 images from the trainval set of the PASCAL VOC2011 challenge, divided into 8498 training and 2820 test images. This dataset has object instance boundaries with accurate figure/ground masks that are also labeled with one of 20 Pascal VOC classes." }, { "dkey": "Cityscapes", "dval": "Cityscapes is a large-scale database which focuses on semantic understanding of urban street scenes. It provides semantic, instance-wise, and dense pixel annotations for 30 classes grouped into 8 categories (flat surfaces, humans, vehicles, constructions, objects, nature, sky, and void). The dataset consists of around 5000 fine annotated images and 20000 coarse annotated ones. Data was captured in 50 cities during several months, daytimes, and good weather conditions. It was originally recorded as video so the frames were manually selected to have the following features: large number of dynamic objects, varying scene layout, and varying background." }, { "dkey": "A2D", "dval": "A2D (Actor-Action Dataset) is a dataset for simultaneously inferring actors and actions in videos. A2D has seven actor classes (adult, baby, ball, bird, car, cat, and dog) and eight action classes (climb, crawl, eat, fly, jump, roll, run, and walk) not including the no-action class, which we also consider. The A2D has 3,782 videos with at least 99 instances per valid actor-action tuple and videos are labeled with both pixel-level actors and actions for sampled frames. The A2D dataset serves as a large-scale testbed for various vision problems: video-level single- and multiple-label actor-action recognition, instance-level object segmentation/co-segmentation, as well as pixel-level actor-action semantic segmentation to name a few." }, { "dkey": "TrashCan", "dval": "The TrashCan dataset is an instance-segmentation dataset of underwater trash. It is comprised of annotated images (7,212 images) which contain observations of trash, ROVs, and a wide variety of undersea flora and fauna. The annotations in this dataset take the format of instance segmentation annotations: bitmaps containing a mask marking which pixels in the image contain each object. The imagery in TrashCan is sourced from the J-EDI (JAMSTEC E-Library of Deep-sea Images) dataset, curated by the Japan Agency of Marine Earth Science and Technology (JAMSTEC)." }, { "dkey": "CIHP", "dval": "The Crowd Instance-level Human Parsing (CIHP) dataset has 38,280 diverse human images. Each image in CIHP is labeled with pixel-wise annotations on 20 categories and instance-level identification. The dataset can be used for the human part segmentation task." }, { "dkey": "Cityscapes-VPS", "dval": "Cityscapes-VPS is a video extension of the Cityscapes validation split. It provides 2500-frame panoptic labels that temporally extend the 500 Cityscapes image-panoptic labels. There are total 3000-frame panoptic labels which correspond to 5, 10, 15, 20, 25, and 30th frames of each 500 videos, where all instance ids are associated over time. It not only supports video panoptic segmentation (VPS) task, but also provides super-set annotations for video semantic segmentation (VSS) and video instance segmentation (VIS) tasks." }, { "dkey": "TTPLA", "dval": "TTPLA is a public dataset which is a collection of aerial images on Transmission Towers (TTs) and Power Lines (PLs). It can be used for detection and segmentation of transmission towers and power lines. It consists of 1,100 images with the resolution of 3,840×2,160 pixels, as well as manually labelled 8,987 instances of TTs and PLs." } ]

[ "FC100", "PASCAL-5i", "Meta-Dataset", "FewRel", "MetaLWOz", "FSOD", "ConvAI2" ]

[ "ImageNet", "PASCAL3D+" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "PASCAL3D+", "dval": "The Pascal3D+ multi-view dataset consists of images in the wild, i.e., images of object categories exhibiting high variability, captured under uncontrolled settings, in cluttered scenes and under many different poses. Pascal3D+ contains 12 categories of rigid objects selected from the PASCAL VOC 2012 dataset. These objects are annotated with pose information (azimuth, elevation and distance to camera). Pascal3D+ also adds pose annotated images of these 12 categories from the ImageNet dataset." }, { "dkey": "FC100", "dval": "The FC100 dataset (Fewshot-CIFAR100) is a newly split dataset based on CIFAR-100 for few-shot learning. It contains 20 high-level categories which are divided into 12, 4, 4 categories for training, validation and test. There are 60, 20, 20 low-level classes in the corresponding split containing 600 images of size 32 × 32 per class. Smaller image size makes it more challenging for few-shot learning." }, { "dkey": "PASCAL-5i", "dval": "PASCAL-5i is a dataset used to evaluate few-shot segmentation. The dataset is sub-divided into 4 folds each containing 5 classes. A fold contains labelled samples from 5 classes that are used for evaluating the few-shot learning method. The rest 15 classes are used for training." }, { "dkey": "Meta-Dataset", "dval": "The Meta-Dataset benchmark is a large few-shot learning benchmark and consists of multiple datasets of different data distributions. It does not restrict few-shot tasks to have fixed ways and shots, thus representing a more realistic scenario. It consists of 10 datasets from diverse domains: \n\n\nILSVRC-2012 (the ImageNet dataset, consisting of natural images with 1000 categories)\nOmniglot (hand-written characters, 1623 classes)\nAircraft (dataset of aircraft images, 100 classes)\nCUB-200-2011 (dataset of Birds, 200 classes)\nDescribable Textures (different kinds of texture images with 43 categories)\nQuick Draw (black and white sketches of 345 different categories)\nFungi (a large dataset of mushrooms with 1500 categories)\nVGG Flower (dataset of flower images with 102 categories), \nTraffic Signs (German traffic sign images with 43 classes)\nMSCOCO (images collected from Flickr, 80 classes). \n\nAll datasets except Traffic signs and MSCOCO have a training, validation and test split (proportioned roughly into 70%, 15%, 15%). The datasets Traffic Signs and MSCOCO are reserved for testing only." }, { "dkey": "FewRel", "dval": "The FewRel (Few-Shot Relation Classification Dataset) contains 100 relations and 70,000 instances from Wikipedia. The dataset is divided into three subsets: training set (64 relations), validation set (16 relations) and test set (20 relations)." }, { "dkey": "MetaLWOz", "dval": "Collected by leveraging background knowledge from a larger, more highly represented dialogue source." }, { "dkey": "FSOD", "dval": "Few-Shot Object Detection Dataset (FSOD) is a high-diverse dataset specifically designed for few-shot object detection and intrinsically designed to evaluate thegenerality of a model on novel categories." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." } ]

[ "AFLW2000-3D", "People Snapshot Dataset", "Hollywood 3D dataset", "Completion3D", "Kitchen Scenes" ]

[ "nuScenes", "KITTI" ]

[ { "dkey": "nuScenes", "dval": "The nuScenes dataset is a large-scale autonomous driving dataset. The dataset has 3D bounding boxes for 1000 scenes collected in Boston and Singapore. Each scene is 20 seconds long and annotated at 2Hz. This results in a total of 28130 samples for training, 6019 samples for validation and 6008 samples for testing. The dataset has the full autonomous vehicle data suite: 32-beam LiDAR, 6 cameras and radars with complete 360° coverage. The 3D object detection challenge evaluates the performance on 10 classes: cars, trucks, buses, trailers, construction vehicles, pedestrians, motorcycles, bicycles, traffic cones and barriers." }, { "dkey": "KITTI", "dval": "KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) is one of the most popular datasets for use in mobile robotics and autonomous driving. It consists of hours of traffic scenarios recorded with a variety of sensor modalities, including high-resolution RGB, grayscale stereo cameras, and a 3D laser scanner. Despite its popularity, the dataset itself does not contain ground truth for semantic segmentation. However, various researchers have manually annotated parts of the dataset to fit their necessities. Álvarez et al. generated ground truth for 323 images from the road detection challenge with three classes: road, vertical, and sky. Zhang et al. annotated 252 (140 for training and 112 for testing) acquisitions – RGB and Velodyne scans – from the tracking challenge for ten object categories: building, sky, road, vegetation, sidewalk, car, pedestrian, cyclist, sign/pole, and fence. Ros et al. labeled 170 training images and 46 testing images (from the visual odometry challenge) with 11 classes: building, tree, sky, car, sign, road, pedestrian, fence, pole, sidewalk, and bicyclist." }, { "dkey": "AFLW2000-3D", "dval": "AFLW2000-3D is a dataset of 2000 images that have been annotated with image-level 68-point 3D facial landmarks. This dataset is used for evaluation of 3D facial landmark detection models. The head poses are very diverse and often hard to be detected by a CNN-based face detector." }, { "dkey": "People Snapshot Dataset", "dval": "Enables detailed human body model reconstruction in clothing from a single monocular RGB video without requiring a pre scanned template or manually clicked points." }, { "dkey": "Hollywood 3D dataset", "dval": "A dataset for benchmarking action recognition algorithms in natural environments, while making use of 3D information. The dataset contains around 650 video clips, across 14 classes. In addition, two state of the art action recognition algorithms are extended to make use of the 3D data, and five new interest point detection strategies are also proposed, that extend to the 3D data." }, { "dkey": "Completion3D", "dval": "The Completion3D benchmark is a dataset for evaluating state-of-the-art 3D Object Point Cloud Completion methods. Ggiven a partial 3D object point cloud the goal is to infer a complete 3D point cloud for the object." }, { "dkey": "Kitchen Scenes", "dval": "Kitchen Scenes is a multi-view RGB-D dataset of nine kitchen scenes, each containing several objects in realistic cluttered environments including a subset of objects from the BigBird dataset. The viewpoints of the scenes are densely sampled and objects in the scenes are annotated with bounding boxes and in the 3D point cloud." } ]

[ "MELD", "ExpW", "SoF", "SemEval 2014 Task 4 Sub Task 2", "300W" ]

[ "FER2013", "AffectNet" ]

[ { "dkey": "FER2013", "dval": "Fer2013 contains approximately 30,000 facial RGB images of different expressions with size restricted to 48×48, and the main labels of it can be divided into 7 types: 0=Angry, 1=Disgust, 2=Fear, 3=Happy, 4=Sad, 5=Surprise, 6=Neutral. The Disgust expression has the minimal number of images – 600, while other labels have nearly 5,000 samples each." }, { "dkey": "AffectNet", "dval": "AffectNet is a large facial expression dataset with around 0.4 million images manually labeled for the presence of eight (neutral, happy, angry, sad, fear, surprise, disgust, contempt) facial expressions along with the intensity of valence and arousal." }, { "dkey": "MELD", "dval": "Multimodal EmotionLines Dataset (MELD) has been created by enhancing and extending EmotionLines dataset. MELD contains the same dialogue instances available in EmotionLines, but it also encompasses audio and visual modality along with text. MELD has more than 1400 dialogues and 13000 utterances from Friends TV series. Multiple speakers participated in the dialogues. Each utterance in a dialogue has been labeled by any of these seven emotions -- Anger, Disgust, Sadness, Joy, Neutral, Surprise and Fear. MELD also has sentiment (positive, negative and neutral) annotation for each utterance." }, { "dkey": "ExpW", "dval": "The Expression in-the-Wild (ExpW) dataset is for facial expression recognition and contains 91,793 faces manually labeled with expressions. Each of the face images is annotated as one of the seven basic expression categories: “angry”, “disgust”, “fear”, “happy”, “sad”, “surprise”, or “neutral”." }, { "dkey": "SoF", "dval": "The Specs on Faces (SoF) dataset, a collection of 42,592 (2,662×16) images for 112 persons (66 males and 46 females) who wear glasses under different illumination conditions. The dataset is FREE for reasonable academic fair use. The dataset presents a new challenge regarding face detection and recognition. It is focused on two challenges: harsh illumination environments and face occlusions, which highly affect face detection, recognition, and classification. The glasses are the common natural occlusion in all images of the dataset. However, there are two more synthetic occlusions (nose and mouth) added to each image. Moreover, three image filters, that may evade face detectors and facial recognition systems, were applied to each image. All generated images are categorized into three levels of difficulty (easy, medium, and hard). That enlarges the number of images to be 42,592 images (26,112 male images and 16,480 female images). There is metadata for each image that contains many information such as: the subject ID, facial landmarks, face and glasses rectangles, gender and age labels, year that the photo was taken, facial emotion, glasses type, and more." }, { "dkey": "SemEval 2014 Task 4 Sub Task 2", "dval": "Sentiment analysis is increasingly viewed as a vital task both from an academic and a commercial standpoint. The majority of current approaches, however, attempt to detect the overall polarity of a sentence, paragraph, or text span, regardless of the entities mentioned (e.g., laptops, restaurants) and their aspects (e.g., battery, screen; food, service). By contrast, this task is concerned with aspect based sentiment analysis (ABSA), where the goal is to identify the aspects of given target entities and the sentiment expressed towards each aspect. Datasets consisting of customer reviews with human-authored annotations identifying the mentioned aspects of the target entities and the sentiment polarity of each aspect will be provided.\n\nSubtask 2: Aspect term polarity\n\nFor a given set of aspect terms within a sentence, determine whether the polarity of each aspect term is positive, negative, neutral or conflict (i.e., both positive and negative).\n\nFor example:\n\n“I loved their fajitas” → {fajitas: positive}\n“I hated their fajitas, but their salads were great” → {fajitas: negative, salads: positive}\n“The fajitas are their first plate” → {fajitas: neutral}\n“The fajitas were great to taste, but not to see” → {fajitas: conflict}" }, { "dkey": "300W", "dval": "The 300-W is a face dataset that consists of 300 Indoor and 300 Outdoor in-the-wild images. It covers a large variation of identity, expression, illumination conditions, pose, occlusion and face size. The images were downloaded from google.com by making queries such as “party”, “conference”, “protests”, “football” and “celebrities”. Compared to the rest of in-the-wild datasets, the 300-W database contains a larger percentage of partially-occluded images and covers more expressions than the common “neutral” or “smile”, such as “surprise” or “scream”.\nImages were annotated with the 68-point mark-up using a semi-automatic methodology. The images of the database were carefully selected so that they represent a characteristic sample of challenging but natural face instances under totally unconstrained conditions. Thus, methods that achieve accurate performance on the 300-W database can demonstrate the same accuracy in most realistic cases.\nMany images of the database contain more than one annotated faces (293 images with 1 face, 53 images with 2 faces and 53 images with [3, 7] faces). Consequently, the database consists of 600 annotated face instances, but 399 unique images. Finally, there is a large variety of face sizes. Specifically, 49.3% of the faces have size in the range [48.6k, 2.0M] and the overall mean size is 85k (about 292 × 292) pixels." } ]

[ "CIFAR10-DVS", "PG-19", "Office-Caltech-10", "arXiv Summarization Dataset", "MUSDB18", "Sentence Compression", "IMPPRES" ]

[ "ImageNet", "COCO" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "CIFAR10-DVS", "dval": "CIFAR10-DVS is an event-stream dataset for object classification. 10,000 frame-based images that come from CIFAR-10 dataset are converted into 10,000 event streams with an event-based sensor, whose resolution is 128×128 pixels. The dataset has an intermediate difficulty with 10 different classes. The repeated closed-loop smooth (RCLS) movement of frame-based images is adopted to implement the conversion. Due to the transformation, they produce rich local intensity changes in continuous time which are quantized by each pixel of the event-based camera." }, { "dkey": "PG-19", "dval": "A new open-vocabulary language modelling benchmark derived from books." }, { "dkey": "Office-Caltech-10", "dval": "Office-Caltech-10 a standard benchmark for domain adaptation, which consists of Office 10 and Caltech 10 datasets. It contains the 10 overlapping categories between the Office dataset and Caltech256 dataset. SURF BoW historgram features, vector quantized to 800 dimensions are also available for this dataset." }, { "dkey": "arXiv Summarization Dataset", "dval": "This is a dataset for evaluating summarisation methods for research papers." }, { "dkey": "MUSDB18", "dval": "The MUSDB18 is a dataset of 150 full lengths music tracks (~10h duration) of different genres along with their isolated drums, bass, vocals and others stems.\n\nThe dataset is split into training and test sets with 100 and 50 songs, respectively. All signals are stereophonic and encoded at 44.1kHz." }, { "dkey": "Sentence Compression", "dval": "Sentence Compression is a dataset where the syntactic trees of the compressions are subtrees of their uncompressed counterparts, and hence where supervised systems which require a structural alignment between the input and output can be successfully trained." }, { "dkey": "IMPPRES", "dval": "An IMPlicature and PRESupposition diagnostic dataset (IMPPRES), consisting of >25k semiautomatically generated sentence pairs illustrating well-studied pragmatic inference types." } ]

[ "LPW", "WiderPerson", "Botswana", "OpenBookQA" ]

[ "CrowdHuman", "CityPersons" ]

[ { "dkey": "CrowdHuman", "dval": "CrowdHuman is a large and rich-annotated human detection dataset, which contains 15,000, 4,370 and 5,000 images collected from the Internet for training, validation and testing respectively. The number is more than 10× boosted compared with previous challenging pedestrian detection dataset like CityPersons. The total number of persons is also noticeably larger than the others with ∼340k person and ∼99k ignore region annotations in the CrowdHuman training subset." }, { "dkey": "CityPersons", "dval": "The CityPersons dataset is a subset of Cityscapes which only consists of person annotations. There are 2975 images for training, 500 and 1575 images for validation and testing. The average of the number of pedestrians in an image is 7. The visible-region and full-body annotations are provided." }, { "dkey": "LPW", "dval": "Labeled Pedestrian in the Wild (LPW) is a pedestrian detection dataset that contains 2,731 pedestrians in three different scenes where each annotated identity is captured by from 2 to 4 cameras. The LPW features a notable scale of 7,694 tracklets with over 590,000 images as well as the cleanliness of its tracklets. It distinguishes from existing datasets in three aspects: large scale with cleanliness, automatically detected bounding boxes and far more crowded scenes with greater age span. This dataset provides a more realistic and challenging benchmark, which facilitates the further exploration of more powerful algorithms." }, { "dkey": "WiderPerson", "dval": "WiderPerson contains a total of 13,382 images with 399,786 annotations, i.e., 29.87 annotations per image, which means this dataset contains dense pedestrians with various kinds of occlusions. Hence, pedestrians in the proposed dataset are extremely challenging due to large variations in the scenario and occlusion, which is suitable to evaluate pedestrian detectors in the wild." }, { "dkey": "Botswana", "dval": "Botswana is a hyperspectral image classification dataset. The NASA EO-1 satellite acquired a sequence of data over the Okavango Delta, Botswana in 2001-2004. The Hyperion sensor on EO-1 acquires data at 30 m pixel resolution over a 7.7 km strip in 242 bands covering the 400-2500 nm portion of the spectrum in 10 nm windows. Preprocessing of the data was performed by the UT Center for Space Research to mitigate the effects of bad detectors, inter-detector miscalibration, and intermittent anomalies. Uncalibrated and noisy bands that cover water absorption features were removed, and the remaining 145 bands were included as candidate features: [10-55, 82-97, 102-119, 134-164, 187-220]. The data analyzed in this study, acquired May 31, 2001, consist of observations from 14 identified classes representing the land cover types in seasonal swamps, occasional swamps, and drier woodlands located in the distal portion of the Delta.\n\nSource: M Graña, MA Veganzons, B Ayerdi" }, { "dkey": "OpenBookQA", "dval": "OpenBookQA is a new kind of question-answering dataset modeled after open book exams for assessing human understanding of a subject. It consists of 5,957 multiple-choice elementary-level science questions (4,957 train, 500 dev, 500 test), which probe the understanding of a small “book” of 1,326 core science facts and the application of these facts to novel situations. For training, the dataset includes a mapping from each question to the core science fact it was designed to probe. Answering OpenBookQA questions requires additional broad common knowledge, not contained in the book. The questions, by design, are answered incorrectly by both a retrieval-based algorithm and a word co-occurrence algorithm.\nAdditionally, the dataset includes a collection of 5,167 crowd-sourced common knowledge facts, and an expanded version of the train/dev/test questions where each question is associated with its originating core fact, a human accuracy score, a clarity score, and an anonymized crowd-worker ID." } ]

[ "THEODORE", "DDD20", "EyeCar", "DeeperForensics-1.0", "DIPS", "E2E" ]

[ "DAVIS", "COCO" ]

[ { "dkey": "DAVIS", "dval": "The Densely Annotation Video Segmentation dataset (DAVIS) is a high quality and high resolution densely annotated video segmentation dataset under two resolutions, 480p and 1080p. There are 50 video sequences with 3455 densely annotated frames in pixel level. 30 videos with 2079 frames are for training and 20 videos with 1376 frames are for validation." }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "THEODORE", "dval": "Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset." }, { "dkey": "DDD20", "dval": "The dataset was captured with a DAVIS camera that concurrently streams both dynamic vision sensor (DVS) brightness change events and active pixel sensor (APS) intensity frames. DDD20 is the longest event camera end-to-end driving dataset to date with 51h of DAVIS event+frame camera and vehicle human control data collected from 4000km of highway and urban driving under a variety of lighting conditions." }, { "dkey": "EyeCar", "dval": "EyeCar is a dataset of driving videos of vehicles involved in rear-end collisions paired with eye fixation data captured from human subjects. It contains 21 front-view videos that were captured in various traffic, weather, and day light conditions. Each video is 30sec in length and contains typical driving tasks (e.g., lanekeeping, merging-in, and braking) ending to rear-end collisions." }, { "dkey": "DeeperForensics-1.0", "dval": "DeeperForensics-1.0 represents the largest face forgery detection dataset by far, with 60,000 videos constituted by a total of 17.6 million frames, 10 times larger than existing datasets of the same kind. The full dataset includes 48,475 source videos and 11,000 manipulated videos. The source videos are collected on 100 paid and consented actors from 26 countries, and the manipulated videos are generated by a newly proposed many-to-many end-to-end face swapping method, DF-VAE. 7 types of real-world perturbations at 5 intensity levels are employed to ensure a larger scale and higher diversity." }, { "dkey": "DIPS", "dval": "Contains biases but is two orders of magnitude larger than those used previously." }, { "dkey": "E2E", "dval": "End-to-End NLG Challenge (E2E) aims to assess whether recent end-to-end NLG systems can generate more complex output by learning from datasets containing higher lexical richness, syntactic complexity and diverse discourse phenomena." } ]

[ "Make3D", "7-Scenes", "CrowdFlow", "JHMDB", "Virtual KITTI", "GSL" ]

[ "SUNCG", "SceneNN", "ImageNet", "ScanNet", "Matterport3D" ]

[ { "dkey": "SUNCG", "dval": "SUNCG is a large-scale dataset of synthetic 3D scenes with dense volumetric annotations.\n\nThe dataset is currently not available." }, { "dkey": "SceneNN", "dval": "SceneNN is an RGB-D scene dataset consisting of more than 100 indoor scenes. The scenes are captured at various places, e.g., offices, dormitory, classrooms, pantry, etc., from University of Massachusetts Boston and Singapore University of Technology and Design.\nAll scenes are reconstructed into triangle meshes and have per-vertex and per-pixel annotation. The dataset is additionally enriched with fine-grained information such as axis-aligned bounding boxes, oriented bounding boxes, and object poses." }, { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "ScanNet", "dval": "ScanNet is an instance-level indoor RGB-D dataset that includes both 2D and 3D data. It is a collection of labeled voxels rather than points or objects. Up to now, ScanNet v2, the newest version of ScanNet, has collected 1513 annotated scans with an approximate 90% surface coverage. In the semantic segmentation task, this dataset is marked in 20 classes of annotated 3D voxelized objects." }, { "dkey": "Matterport3D", "dval": "The Matterport3D dataset is a large RGB-D dataset for scene understanding in indoor environments. It contains 10,800 panoramic views inside 90 real building-scale scenes, constructed from 194,400 RGB-D images. Each scene is a residential building consisting of multiple rooms and floor levels, and is annotated with surface construction, camera poses, and semantic segmentation." }, { "dkey": "Make3D", "dval": "The Make3D dataset is a monocular Depth Estimation dataset that contains 400 single training RGB and depth map pairs, and 134 test samples. The RGB images have high resolution, while the depth maps are provided at low resolution." }, { "dkey": "7-Scenes", "dval": "The 7-Scenes dataset is a collection of tracked RGB-D camera frames. The dataset may be used for evaluation of methods for different applications such as dense tracking and mapping and relocalization techniques.\nAll scenes were recorded from a handheld Kinect RGB-D camera at 640×480 resolution. The dataset creators use an implementation of the KinectFusion system to obtain the ‘ground truth’ camera tracks, and a dense 3D model. Several sequences were recorded per scene by different users, and split into distinct training and testing sequence sets." }, { "dkey": "CrowdFlow", "dval": "The TUB CrowdFlow is a synthetic dataset that contains 10 sequences showing 5 scenes. Each scene is rendered twice: with a static point of view and a dynamic camera to simulate drone/UAV based surveillance. The scenes are render using Unreal Engine at HD resolution (1280x720) at 25 fps, which is typical for current commercial CCTV surveillance systems. The total number of frames is 3200.\n\nEach sequence has the following ground-truth data:\n\n\nOptical flow fields\nPerson trajectories (up to 1451)\nDense pixel trajectories" }, { "dkey": "JHMDB", "dval": "JHMDB is an action recognition dataset that consists of 960 video sequences belonging to 21 actions. It is a subset of the larger HMDB51 dataset collected from digitized movies and YouTube videos. The dataset contains video and annotation for puppet flow per frame (approximated optimal flow on the person), puppet mask per frame, joint positions per frame, action label per clip and meta label per clip (camera motion, visible body parts, camera viewpoint, number of people, video quality)." }, { "dkey": "Virtual KITTI", "dval": "Virtual KITTI is a photo-realistic synthetic video dataset designed to learn and evaluate computer vision models for several video understanding tasks: object detection and multi-object tracking, scene-level and instance-level semantic segmentation, optical flow, and depth estimation.\n\nVirtual KITTI contains 50 high-resolution monocular videos (21,260 frames) generated from five different virtual worlds in urban settings under different imaging and weather conditions. These worlds were created using the Unity game engine and a novel real-to-virtual cloning method. These photo-realistic synthetic videos are automatically, exactly, and fully annotated for 2D and 3D multi-object tracking and at the pixel level with category, instance, flow, and depth labels (cf. below for download links)." }, { "dkey": "GSL", "dval": "Dataset Description\nThe Greek Sign Language (GSL) is a large-scale RGB+D dataset, suitable for Sign Language Recognition (SLR) and Sign Language Translation (SLT). The video captures are conducted using an Intel RealSense D435 RGB+D camera at a rate of 30 fps. Both the RGB and the depth streams are acquired in the same spatial resolution of 848×480 pixels. To increase variability in the videos, the camera position and orientation is slightly altered within subsequent recordings. Seven different signers are employed to perform 5 individual and commonly met scenarios in different public services. The average length of each scenario is twenty sentences.\n\nThe dataset contains 10,290 sentence instances, 40,785 gloss instances, 310 unique glosses (vocabulary size) and 331 unique sentences, with 4.23 glosses per sentence on average. Each signer is asked to perform the pre-defined dialogues five consecutive times. In all cases, the simulation considers a deaf person communicating with a single public service employee. The involved signer performs the sequence of glosses of both agents in the discussion. For the annotation of each gloss sequence, GSL linguistic experts are involved. The given annotations are at individual gloss and gloss sequence level. A translation of the gloss sentences to spoken Greek is also provided.\n\nEvaluation\nThe GSL dataset includes the 3 evaluation setups:\n\n\n\nSigner-dependent continuous sign language recognition (GSL SD) – roughly 80% of videos are used for training, corresponding to 8,189 instances. The rest 1,063 (10%) were kept for validation and 1,043 (10%) for testing.\n\n\n\nSigner-independent continuous sign language recognition (GSL SI) – the selected test gloss sequences are not used in the training set, while all the individual glosses exist in the training set. In GSL SI, the recordings of one signer are left out for validation and testing (588 and 881 instances, respectively). The rest 8821 instances are utilized for training.\n\n\n\nIsolated gloss sign language recognition (GSL isol.) – The validation set consists of 2,231 gloss instances, the test set 3,500, while the remaining 34,995 are used for training. All 310 unique glosses are seen in the training set.\n\n\n\nFor more info and results, advice our paper\n\nPaper Abstract: A Comprehensive Study on Sign Language Recognition Methods, Adaloglou et al. 2020\nIn this paper, a comparative experimental assessment of computer vision-based methods for sign language recognition is conducted. By implementing the most recent deep neural network methods in this field, a thorough evaluation on multiple publicly available datasets is performed. The aim of the present study is to provide insights on sign language recognition, focusing on mapping non-segmented video streams to glosses. For this task, two new sequence training criteria, known from the fields of speech and scene text recognition, are introduced. Furthermore, a\nplethora of pretraining schemes are thoroughly discussed. Finally, a new RGB+D dataset for the Greek sign language is created. To the best of our knowledge, this is the first sign language dataset where sentence and gloss level annotations are provided for every video capture.\n\nArxiv link" } ]

[ "ROSE", "RITE", "ORVS", "CHASE_DB1", "ADAM" ]

[ "STARE", "DRIVE" ]

[ { "dkey": "STARE", "dval": "The STARE (Structured Analysis of the Retina) dataset is a dataset for retinal vessel segmentation. It contains 20 equal-sized (700×605) color fundus images. For each image, two groups of annotations are provided.." }, { "dkey": "DRIVE", "dval": "The Digital Retinal Images for Vessel Extraction (DRIVE) dataset is a dataset for retinal vessel segmentation. It consists of a total of JPEG 40 color fundus images; including 7 abnormal pathology cases. The images were obtained from a diabetic retinopathy screening program in the Netherlands. The images were acquired using Canon CR5 non-mydriatic 3CCD camera with FOV equals to 45 degrees. Each image resolution is 584*565 pixels with eight bits per color channel (3 channels). \n\nThe set of 40 images was equally divided into 20 images for the training set and 20 images for the testing set. Inside both sets, for each image, there is circular field of view (FOV) mask of diameter that is approximately 540 pixels. Inside training set, for each image, one manual segmentation by an ophthalmological expert has been applied. Inside testing set, for each image, two manual segmentations have been applied by two different observers, where the first observer segmentation is accepted as the ground-truth for performance evaluation." }, { "dkey": "ROSE", "dval": "Retinal OCTA SEgmentation dataset (ROSE) consists of 229 OCTA images with vessel annotations at either centerline-level or pixel level." }, { "dkey": "RITE", "dval": "The RITE (Retinal Images vessel Tree Extraction) is a database that enables comparative studies on segmentation or classification of arteries and veins on retinal fundus images, which is established based on the public available DRIVE database (Digital Retinal Images for Vessel Extraction).\n\nRITE contains 40 sets of images, equally separated into a training subset and a test subset, the same as DRIVE. The two subsets are built from the corresponding two subsets in DRIVE. For each set, there is a fundus photograph, a vessel reference standard, and a Arteries/Veins (A/V) reference standard. \n\n\nThe fundus photograph is inherited from DRIVE. \nFor the training set, the vessel reference standard is a modified version of 1st_manual from DRIVE. \nFor the test set, the vessel reference standard is 2nd_manual from DRIVE. \nFor the A/V reference standard, four types of vessels are labelled using four colors based on the vessel reference standard. \nArteries are labelled in red; veins are labelled in blue; the overlapping of arteries and veins are labelled in green; the vessels which are uncertain are labelled in white. \nThe fundus photograph is in tif format. And the vessel reference standard and the A/V reference standard are in png format. \n\nThe dataset is described in more detail in our paper, which you will cite if you use the dataset in any way: \n\nHu Q, Abràmoff MD, Garvin MK. Automated separation of binary overlapping trees in low-contrast color retinal images. Med Image Comput Comput Assist Interv. 2013;16(Pt 2):436-43. PubMed PMID: 24579170 https://doi.org/10.1007/978-3-642-40763-5_54" }, { "dkey": "ORVS", "dval": "The ORVS dataset has been newly established as a collaboration between the computer science and visual-science departments at the University of Calgary.\n\nThis dataset contains 49 images (42 training and seven testing images) collected from a clinic in Calgary-Canada. All images were acquired with a Zeiss Visucam 200 with 30 degrees field of view (FOV). The image size is 1444×1444 with 24 bits per pixel. Images and are stored in JPEG format with low compression, which is common in ophthalmology practice. All images were manually traced by an expert who a has been working in the field of retinal-image analysis and went through training. The expert was asked to label all pixels belonging to retinal vessels. The Windows Paint 3D tool was used to manually label the images." }, { "dkey": "CHASE_DB1", "dval": "CHASE_DB1 is a dataset for retinal vessel segmentation which contains 28 color retina images with the size of 999×960 pixels which are collected from both left and right eyes of 14 school children. Each image is annotated by two independent human experts." }, { "dkey": "ADAM", "dval": "ADAM is organized as a half day Challenge, a Satellite Event of the ISBI 2020 conference in Iowa City, Iowa, USA.\n\nThe ADAM challenge focuses on the investigation and development of algorithms associated with the diagnosis of Age-related Macular degeneration (AMD) and segmentation of lesions in fundus photos from AMD patients. The goal of the challenge is to evaluate and compare automated algorithms for the detection of AMD on a common dataset of retinal fundus images. We invite the medical image analysis community to participate by developing and testing existing and novel automated fundus classification and segmentation methods.\n\nInstructions: \nADAM: Automatic Detection challenge on Age-related Macular degeneration\n\nLink: https://amd.grand-challenge.org\n\nAge-related macular degeneration, abbreviated as AMD, is a degenerative disorder in the macular region. It mainly occurs in people older than 45 years old and its incidence rate is even higher than diabetic retinopathy in the elderly. \n\nThe etiology of AMD is not fully understood, which could be related to multiple factors, including genetics, chronic photodestruction effect, and nutritional disorder. AMD is classified into Dry AMD and Wet AMD. Dry AMD (also called nonexudative AMD) is not neovascular. It is characterized by progressive atrophy of retinal pigment epithelium (RPE). In the late stage, drusen and the large area of atrophy could be observed under ophthalmoscopy. Wet AMD (also called neovascular or exudative AMD), is characterized by active neovascularization under RPE, subsequently causing exudation, hemorrhage, and scarring, and will eventually cause irreversible damage to the photoreceptors and rapid vision loss if left untreated.\n\nAn early diagnosis of AMD is crucial to treatment and prognosis. Fundus photo is one of the basic examinations. The current dataset is composed of AMD and non-AMD (myopia, normal control, etc.) photos. Typical signs of AMD that can be found in these photos include drusen, exudation, hemorrhage, etc. \n\nThe ADAM challenge has 4 tasks:\n\nTask 1: Classification of AMD and non-AMD fundus images.\n\nTask 2: Detection and segmentation of optic disc.\n\nTask 3: Localization of fovea.\n\nTask 4: Detection and Segmentation of lesions from fundus images." } ]

[ "SBU Captions Dataset", "BDD100K", "Localized Narratives", "THEODORE", "LS3D-W" ]

[ "Oxford105k", "Oxford5k" ]

[ { "dkey": "Oxford105k", "dval": "Oxford105k is the combination of the Oxford5k dataset and 99782 negative images crawled from Flickr using 145 most popular tags. This dataset is used to evaluate search performance for object retrieval (reported as mAP) on a large scale." }, { "dkey": "Oxford5k", "dval": "Oxford5K is the Oxford Buildings Dataset, which contains 5062 images collected from Flickr. It offers a set of 55 queries for 11 landmark buildings, five for each landmark." }, { "dkey": "SBU Captions Dataset", "dval": "A collection that allows researchers to approach the extremely challenging problem of description generation using relatively simple non-parametric methods and produces surprisingly effective results." }, { "dkey": "BDD100K", "dval": "Datasets drive vision progress, yet existing driving datasets are impoverished in terms of visual content and supported tasks to study multitask learning for autonomous driving. Researchers are usually constrained to study a small set of problems on one dataset, while real-world computer vision applications require performing tasks of various complexities. We construct BDD100K, the largest driving video dataset with 100K videos and 10 tasks to evaluate the exciting progress of image recognition algorithms on autonomous driving. The dataset possesses geographic, environmental, and weather diversity, which is useful for training models that are less likely to be surprised by new conditions. Based on this diverse dataset, we build a benchmark for heterogeneous multitask learning and study how to solve the tasks together. Our experiments show that special training strategies are needed for existing models to perform such heterogeneous tasks. BDD100K opens the door for future studies in this important venue. More detail is at the dataset home page." }, { "dkey": "Localized Narratives", "dval": "We propose Localized Narratives, a new form of multimodal image annotations connecting vision and language. We ask annotators to describe an image with their voice while simultaneously hovering their mouse over the region they are describing. Since the voice and the mouse pointer are synchronized, we can localize every single word in the description. This dense visual grounding takes the form of a mouse trace segment per word and is unique to our data. We annotated 849k images with Localized Narratives: the whole COCO, Flickr30k, and ADE20K datasets, and 671k images of Open Images, all of which we make publicly available. We provide an extensive analysis of these annotations showing they are diverse, accurate, and efficient to produce. We also demonstrate their utility on the application of controlled image captioning." }, { "dkey": "THEODORE", "dval": "Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset." }, { "dkey": "LS3D-W", "dval": "A 3D facial landmark dataset of around 230,000 images." } ]

[ "VRD", "QA-SRL", "Tasty Videos", "RareAct", "LAD", "decaNLP" ]

[ "UCF101", "ActivityNet", "Charades" ]

[ { "dkey": "UCF101", "dval": "UCF101 dataset is an extension of UCF50 and consists of 13,320 video clips, which are classified into 101 categories. These 101 categories can be classified into 5 types (Body motion, Human-human interactions, Human-object interactions, Playing musical instruments and Sports). The total length of these video clips is over 27 hours. All the videos are collected from YouTube and have a fixed frame rate of 25 FPS with the resolution of 320 × 240." }, { "dkey": "ActivityNet", "dval": "The ActivityNet dataset contains 200 different types of activities and a total of 849 hours of videos collected from YouTube. ActivityNet is the largest benchmark for temporal activity detection to date in terms of both the number of activity categories and number of videos, making the task particularly challenging. Version 1.3 of the dataset contains 19994 untrimmed videos in total and is divided into three disjoint subsets, training, validation, and testing by a ratio of 2:1:1. On average, each activity category has 137 untrimmed videos. Each video on average has 1.41 activities which are annotated with temporal boundaries. The ground-truth annotations of test videos are not public." }, { "dkey": "Charades", "dval": "The Charades dataset is composed of 9,848 videos of daily indoors activities with an average length of 30 seconds, involving interactions with 46 objects classes in 15 types of indoor scenes and containing a vocabulary of 30 verbs leading to 157 action classes. Each video in this dataset is annotated by multiple free-text descriptions, action labels, action intervals and classes of interacting objects. 267 different users were presented with a sentence, which includes objects and actions from a fixed vocabulary, and they recorded a video acting out the sentence. In total, the dataset contains 66,500 temporal annotations for 157 action classes, 41,104 labels for 46 object classes, and 27,847 textual descriptions of the videos. In the standard split there are7,986 training video and 1,863 validation video." }, { "dkey": "VRD", "dval": "The Visual Relationship Dataset (VRD) contains 4000 images for training and 1000 for testing annotated with visual relationships. Bounding boxes are annotated with a label containing 100 unary predicates. These labels refer to animals, vehicles, clothes and generic objects. Pairs of bounding boxes are annotated with a label containing 70 binary predicates. These labels refer to actions, prepositions, spatial relations, comparatives or preposition phrases. The dataset has 37993 instances of visual relationships and 6672 types of relationships. 1877 instances of relationships occur only in the test set and they are used to evaluate the zero-shot learning scenario." }, { "dkey": "QA-SRL", "dval": "QA-SRL was proposed as an open schema for semantic roles, in which the relation between an argument and a predicate is expressed as a natural-language question containing the predicate (“Where was someone educated?”) whose answer is the argument (“Princeton”). The authors collected about 19,000 question-answer pairs from 3,200 sentences." }, { "dkey": "Tasty Videos", "dval": "A collection of 2511 recipes for zero-shot learning, recognition and anticipation." }, { "dkey": "RareAct", "dval": "RareAct is a video dataset of unusual actions, including actions like “blend phone”, “cut keyboard” and “microwave shoes”. It aims at evaluating the zero-shot and few-shot compositionality of action recognition models for unlikely compositions of common action verbs and object nouns. It contains 122 different actions which were obtained by combining verbs and nouns rarely co-occurring together in the large-scale textual corpus from HowTo100M, but that frequently appear separately." }, { "dkey": "LAD", "dval": "LAD (Large-scale Attribute Dataset) has 78,017 images of 5 super-classes and 230 classes. The image number of LAD is larger than the sum of the four most popular attribute datasets (AwA, CUB, aP/aY and SUN). 359 attributes of visual, semantic and subjective properties are defined and annotated in instance-level." }, { "dkey": "decaNLP", "dval": "Natural Language Decathlon Benchmark (decaNLP) is a challenge that spans ten tasks: question answering, machine translation, summarization, natural language inference, sentiment analysis, semantic role labeling, zero-shot relation extraction, goal-oriented dialogue, semantic parsing, and commonsense pronoun resolution. The tasks as cast as question answering over a context." } ]

[ "arXiv Summarization Dataset", "Visual Genome", "PETA", "RAF-DB", "WFLW", "FFHQ", "LAG" ]

[ "MORPH", "CelebA" ]

[ { "dkey": "MORPH", "dval": "MORPH is a facial age estimation dataset, which contains 55,134 facial images of 13,617 subjects ranging from 16 to 77 years old." }, { "dkey": "CelebA", "dval": "CelebFaces Attributes dataset contains 202,599 face images of the size 178×218 from 10,177 celebrities, each annotated with 40 binary labels indicating facial attributes like hair color, gender and age." }, { "dkey": "arXiv Summarization Dataset", "dval": "This is a dataset for evaluating summarisation methods for research papers." }, { "dkey": "Visual Genome", "dval": "Visual Genome contains Visual Question Answering data in a multi-choice setting. It consists of 101,174 images from MSCOCO with 1.7 million QA pairs, 17 questions per image on average. Compared to the Visual Question Answering dataset, Visual Genome represents a more balanced distribution over 6 question types: What, Where, When, Who, Why and How. The Visual Genome dataset also presents 108K images with densely annotated objects, attributes and relationships." }, { "dkey": "PETA", "dval": "The PEdesTrian Attribute dataset (PETA) is a dataset fore recognizing pedestrian attributes, such as gender and clothing style, at a far distance. It is of interest in video surveillance scenarios where face and body close-shots and hardly available. It consists of 19,000 pedestrian images with 65 attributes (61 binary and 4 multi-class). Those images contain 8705 persons." }, { "dkey": "RAF-DB", "dval": "The Real-world Affective Faces Database (RAF-DB) is a dataset for facial expression. It contains 29672 facial images tagged with basic or compound expressions by 40 independent taggers. Images in this database are of great variability in subjects' age, gender and ethnicity, head poses, lighting conditions, occlusions, (e.g. glasses, facial hair or self-occlusion), post-processing operations (e.g. various filters and special effects), etc." }, { "dkey": "WFLW", "dval": "The Wider Facial Landmarks in the Wild or WFLW database contains 10000 faces (7500 for training and 2500 for testing) with 98 annotated landmarks. This database also features rich attribute annotations in terms of occlusion, head pose, make-up, illumination, blur and expressions." }, { "dkey": "FFHQ", "dval": "Flickr-Faces-HQ (FFHQ) consists of 70,000 high-quality PNG images at 1024×1024 resolution and contains considerable variation in terms of age, ethnicity and image background. It also has good coverage of accessories such as eyeglasses, sunglasses, hats, etc. The images were crawled from Flickr, thus inheriting all the biases of that website, and automatically aligned and cropped using dlib. Only images under permissive licenses were collected. Various automatic filters were used to prune the set, and finally Amazon Mechanical Turk was used to remove the occasional statues, paintings, or photos of photos." }, { "dkey": "LAG", "dval": "Includes 5,824 fundus images labeled with either positive glaucoma (2,392) or negative glaucoma (3,432)." } ]

[ "VidSTG", "DCASE 2014", "Places", "VoxPopuli", "Email-EU", "UTD-MHAD" ]

[ "UCF101", "HMDB51" ]

[ { "dkey": "UCF101", "dval": "UCF101 dataset is an extension of UCF50 and consists of 13,320 video clips, which are classified into 101 categories. These 101 categories can be classified into 5 types (Body motion, Human-human interactions, Human-object interactions, Playing musical instruments and Sports). The total length of these video clips is over 27 hours. All the videos are collected from YouTube and have a fixed frame rate of 25 FPS with the resolution of 320 × 240." }, { "dkey": "HMDB51", "dval": "The HMDB51 dataset is a large collection of realistic videos from various sources, including movies and web videos. The dataset is composed of 6,766 video clips from 51 action categories (such as “jump”, “kiss” and “laugh”), with each category containing at least 101 clips. The original evaluation scheme uses three different training/testing splits. In each split, each action class has 70 clips for training and 30 clips for testing. The average accuracy over these three splits is used to measure the final performance." }, { "dkey": "VidSTG", "dval": "The VidSTG dataset is a spatio-temporal video grounding dataset constructed based on the video relation dataset VidOR. VidOR contains 7,000, 835 and 2,165 videos for training, validation and testing, respectively. The goal of the Spatio-Temporal Video Grounding task (STVG) is to localize the spatio-temporal section of an untrimmed video that matches a given sentence depicting an object." }, { "dkey": "DCASE 2014", "dval": "DCASE2014 is an audio classification benchmark." }, { "dkey": "Places", "dval": "The Places dataset is proposed for scene recognition and contains more than 2.5 million images covering more than 205 scene categories with more than 5,000 images per category." }, { "dkey": "VoxPopuli", "dval": "VoxPopuli is a large-scale multilingual corpus providing 100K hours of unlabelled speech data in 23 languages. It is the largest open data to date for unsupervised representation learning as well as semi-supervised learning. VoxPopuli also contains 1.8K hours of transcribed speeches in 16 languages and their aligned oral interpretations into 5 other languages totaling 5.1K hours." }, { "dkey": "Email-EU", "dval": "EmailEU is a directed temporal network constructed from email exchanges in a large European research institution for a 803-day period. It contains 986 email addresses as nodes and 332,334 emails as edges with timestamps. There are 42 ground truth departments in the dataset." }, { "dkey": "UTD-MHAD", "dval": "The UTD-MHAD dataset consists of 27 different actions performed by 8 subjects. Each subject repeated the action for 4 times, resulting in 861 action sequences in total. The RGB, depth, skeleton and the inertial sensor signals were recorded." } ]

[ "GSL", "RoboNet", "AnimalWeb", "Million-AID", "StereoSet", "DailyDialog++" ]

[ "QNLI", "WikiText-103", "MRPC", "CoLA", "GLUE" ]

[ { "dkey": "QNLI", "dval": "The QNLI (Question-answering NLI) dataset is a Natural Language Inference dataset automatically derived from the Stanford Question Answering Dataset v1.1 (SQuAD). SQuAD v1.1 consists of question-paragraph pairs, where one of the sentences in the paragraph (drawn from Wikipedia) contains the answer to the corresponding question (written by an annotator). The dataset was converted into sentence pair classification by forming a pair between each question and each sentence in the corresponding context, and filtering out pairs with low lexical overlap between the question and the context sentence. The task is to determine whether the context sentence contains the answer to the question. This modified version of the original task removes the requirement that the model select the exact answer, but also removes the simplifying assumptions that the answer is always present in the input and that lexical overlap is a reliable cue. The QNLI dataset is part of GLEU benchmark." }, { "dkey": "WikiText-103", "dval": "The WikiText language modeling dataset is a collection of over 100 million tokens extracted from the set of verified Good and Featured articles on Wikipedia. The dataset is available under the Creative Commons Attribution-ShareAlike License.\n\nCompared to the preprocessed version of Penn Treebank (PTB), WikiText-2 is over 2 times larger and WikiText-103 is over 110 times larger. The WikiText dataset also features a far larger vocabulary and retains the original case, punctuation and numbers - all of which are removed in PTB. As it is composed of full articles, the dataset is well suited for models that can take advantage of long term dependencies." }, { "dkey": "MRPC", "dval": "Microsoft Research Paraphrase Corpus (MRPC) is a corpus consists of 5,801 sentence pairs collected from newswire articles. Each pair is labelled if it is a paraphrase or not by human annotators. The whole set is divided into a training subset (4,076 sentence pairs of which 2,753 are paraphrases) and a test subset (1,725 pairs of which 1,147 are paraphrases)." }, { "dkey": "CoLA", "dval": "The Corpus of Linguistic Acceptability (CoLA) consists of 10657 sentences from 23 linguistics publications, expertly annotated for acceptability (grammaticality) by their original authors. The public version contains 9594 sentences belonging to training and development sets, and excludes 1063 sentences belonging to a held out test set." }, { "dkey": "GLUE", "dval": "General Language Understanding Evaluation (GLUE) benchmark is a collection of nine natural language understanding tasks, including single-sentence tasks CoLA and SST-2, similarity and paraphrasing tasks MRPC, STS-B and QQP, and natural language inference tasks MNLI, QNLI, RTE and WNLI." }, { "dkey": "GSL", "dval": "Dataset Description\nThe Greek Sign Language (GSL) is a large-scale RGB+D dataset, suitable for Sign Language Recognition (SLR) and Sign Language Translation (SLT). The video captures are conducted using an Intel RealSense D435 RGB+D camera at a rate of 30 fps. Both the RGB and the depth streams are acquired in the same spatial resolution of 848×480 pixels. To increase variability in the videos, the camera position and orientation is slightly altered within subsequent recordings. Seven different signers are employed to perform 5 individual and commonly met scenarios in different public services. The average length of each scenario is twenty sentences.\n\nThe dataset contains 10,290 sentence instances, 40,785 gloss instances, 310 unique glosses (vocabulary size) and 331 unique sentences, with 4.23 glosses per sentence on average. Each signer is asked to perform the pre-defined dialogues five consecutive times. In all cases, the simulation considers a deaf person communicating with a single public service employee. The involved signer performs the sequence of glosses of both agents in the discussion. For the annotation of each gloss sequence, GSL linguistic experts are involved. The given annotations are at individual gloss and gloss sequence level. A translation of the gloss sentences to spoken Greek is also provided.\n\nEvaluation\nThe GSL dataset includes the 3 evaluation setups:\n\n\n\nSigner-dependent continuous sign language recognition (GSL SD) – roughly 80% of videos are used for training, corresponding to 8,189 instances. The rest 1,063 (10%) were kept for validation and 1,043 (10%) for testing.\n\n\n\nSigner-independent continuous sign language recognition (GSL SI) – the selected test gloss sequences are not used in the training set, while all the individual glosses exist in the training set. In GSL SI, the recordings of one signer are left out for validation and testing (588 and 881 instances, respectively). The rest 8821 instances are utilized for training.\n\n\n\nIsolated gloss sign language recognition (GSL isol.) – The validation set consists of 2,231 gloss instances, the test set 3,500, while the remaining 34,995 are used for training. All 310 unique glosses are seen in the training set.\n\n\n\nFor more info and results, advice our paper\n\nPaper Abstract: A Comprehensive Study on Sign Language Recognition Methods, Adaloglou et al. 2020\nIn this paper, a comparative experimental assessment of computer vision-based methods for sign language recognition is conducted. By implementing the most recent deep neural network methods in this field, a thorough evaluation on multiple publicly available datasets is performed. The aim of the present study is to provide insights on sign language recognition, focusing on mapping non-segmented video streams to glosses. For this task, two new sequence training criteria, known from the fields of speech and scene text recognition, are introduced. Furthermore, a\nplethora of pretraining schemes are thoroughly discussed. Finally, a new RGB+D dataset for the Greek sign language is created. To the best of our knowledge, this is the first sign language dataset where sentence and gloss level annotations are provided for every video capture.\n\nArxiv link" }, { "dkey": "RoboNet", "dval": "An open database for sharing robotic experience, which provides an initial pool of 15 million video frames, from 7 different robot platforms, and study how it can be used to learn generalizable models for vision-based robotic manipulation." }, { "dkey": "AnimalWeb", "dval": "A large-scale, hierarchical annotated dataset of animal faces, featuring 21.9K faces from 334 diverse species and 21 animal orders across biological taxonomy. These faces are captured `in-the-wild' conditions and are consistently annotated with 9 landmarks on key facial features. The proposed dataset is structured and scalable by design; its development underwent four systematic stages involving rigorous, manual annotation effort of over 6K man-hours." }, { "dkey": "Million-AID", "dval": "Million-AID is a large-scale benchmark dataset containing a million instances for RS scene classification. There are 51 semantic scene categories in Million-AID. And the scene categories are customized to match the land-use classification standards, which greatly enhance the practicability of the constructed Million-AID. Different form the existing scene classification datasets of which categories are organized with parallel or uncertain relationships, scene categories in Million-AID are organized with systematic relationship architecture, giving it superiority in management and scalability. Specifically, the scene categories in Million-AID are organized by the hierarchical category network of a three-level tree: 51 leaf nodes fall into 28 parent nodes at the second level which are grouped into 8 nodes at the first level, representing the 8 underlying scene categories of agriculture land, commercial land, industrial land, public service land, residential land, transportation land, unutilized land, and water area. The scene category network provides the dataset with excellent organization of relationship among different scene categories and also the property of scalability. The number of images in each scene category ranges from 2,000 to 45,000, endowing the dataset with the property of long tail distribution. Besides, Million-AID has superiorities over the existing scene classification datasets owing to its high spatial resolution, large scale, and global distribution." }, { "dkey": "StereoSet", "dval": "A large-scale natural dataset in English to measure stereotypical biases in four domains: gender, profession, race, and religion." }, { "dkey": "DailyDialog++", "dval": "Consists of (i) five relevant responses for each context and (ii) five adversarially crafted irrelevant responses for each context." } ]

[ "HotpotQA", "TGIF-QA", "MultiReQA", "MovieQA", "JEC-QA", "LEAF-QA" ]

[ "RACE", "SQuAD" ]

[ { "dkey": "RACE", "dval": "The ReAding Comprehension dataset from Examinations (RACE) dataset is a machine reading comprehension dataset consisting of 27,933 passages and 97,867 questions from English exams, targeting Chinese students aged 12-18. RACE consists of two subsets, RACE-M and RACE-H, from middle school and high school exams, respectively. RACE-M has 28,293 questions and RACE-H has 69,574. Each question is associated with 4 candidate answers, one of which is correct. The data generation process of RACE differs from most machine reading comprehension datasets - instead of generating questions and answers by heuristics or crowd-sourcing, questions in RACE are specifically designed for testing human reading skills, and are created by domain experts." }, { "dkey": "SQuAD", "dval": "The Stanford Question Answering Dataset (SQuAD) is a collection of question-answer pairs derived from Wikipedia articles. In SQuAD, the correct answers of questions can be any sequence of tokens in the given text. Because the questions and answers are produced by humans through crowdsourcing, it is more diverse than some other question-answering datasets. SQuAD 1.1 contains 107,785 question-answer pairs on 536 articles. SQuAD2.0 (open-domain SQuAD, SQuAD-Open), the latest version, combines the 100,000 questions in SQuAD1.1 with over 50,000 un-answerable questions written adversarially by crowdworkers in forms that are similar to the answerable ones." }, { "dkey": "HotpotQA", "dval": "HotpotQA is a question answering dataset collected on the English Wikipedia, containing about 113K crowd-sourced questions that are constructed to require the introduction paragraphs of two Wikipedia articles to answer. Each question in the dataset comes with the two gold paragraphs, as well as a list of sentences in these paragraphs that crowdworkers identify as supporting facts necessary to answer the question. \n\nA diverse range of reasoning strategies are featured in HotpotQA, including questions involving missing entities in the question, intersection questions (What satisfies property A and property B?), and comparison questions, where two entities are compared by a common attribute, among others. In the few-document distractor setting, the QA models are given ten paragraphs in which the gold paragraphs are guaranteed to be found; in the open-domain fullwiki setting, the models are only given the question and the entire Wikipedia. Models are evaluated on their answer accuracy and explainability, where the former is measured as overlap between the predicted and gold answers with exact match (EM) and unigram F1, and the latter concerns how well the predicted supporting fact sentences match human annotation (Supporting Fact EM/F1). A joint metric is also reported on this dataset, which encourages systems to perform well on both tasks simultaneously." }, { "dkey": "TGIF-QA", "dval": "The TGIF-QA dataset contains 165K QA pairs for the animated GIFs from the TGIF dataset [Li et al. CVPR 2016]. The question & answer pairs are collected via crowdsourcing with a carefully designed user interface to ensure quality. The dataset can be used to evaluate video-based Visual Question Answering techniques." }, { "dkey": "MultiReQA", "dval": "MultiReQA is a cross-domain evaluation for retrieval question answering models. Retrieval question answering (ReQA) is the task of retrieving a sentence-level answer to a question from an open corpus. MultiReQA is a new multi-domain ReQA evaluation suite composed of eight retrieval QA tasks drawn from publicly available QA datasets from the MRQA shared task.\nMultiReQA contains the sentence boundary annotation from eight publicly available QA datasets including SearchQA, TriviaQA, HotpotQA, NaturalQuestions, SQuAD, BioASQ, RelationExtraction, and TextbookQA. Five of these datasets, including SearchQA, TriviaQA, HotpotQA, NaturalQuestions, SQuAD, contain both training and test data, and three, in cluding BioASQ, RelationExtraction, TextbookQA, contain only the test data." }, { "dkey": "MovieQA", "dval": "The MovieQA dataset is a dataset for movie question answering. to evaluate automatic story comprehension from both video and text. The data set consists of almost 15,000 multiple choice question answers obtained from over 400 movies and features high semantic diversity. Each question comes with a set of five highly plausible answers; only one of which is correct. The questions can be answered using multiple sources of information: movie clips, plots, subtitles, and for a subset scripts and DVS." }, { "dkey": "JEC-QA", "dval": "JEC-QA is a LQA (Legal Question Answering) dataset collected from the National Judicial Examination of China. It contains 26,365 multiple-choice and multiple-answer questions in total. The task of the dataset is to predict the answer using the questions and relevant articles. To do well on JEC-QA, both retrieving and answering are important." }, { "dkey": "LEAF-QA", "dval": "LEAF-QA, a comprehensive dataset of 250,000 densely annotated figures/charts, constructed from real-world open data sources, along with ~2 million question-answer (QA) pairs querying the structure and semantics of these charts. LEAF-QA highlights the problem of multimodal QA, which is notably different from conventional visual QA (VQA), and has recently gained interest in the community. Furthermore, LEAF-QA is significantly more complex than previous attempts at chart QA, viz. FigureQA and DVQA, which present only limited variations in chart data. LEAF-QA being constructed from real-world sources, requires a novel architecture to enable question answering." } ]

[ "AFLW2000-3D", "FaceForensics", "300-VW", "LS3D-W", "WFLW", "SoF" ]

[ "AFW", "AFLW" ]

[ { "dkey": "AFW", "dval": "AFW (Annotated Faces in the Wild) is a face detection dataset that contains 205 images with 468 faces. Each face image is labeled with at most 6 landmarks with visibility labels, as well as a bounding box." }, { "dkey": "AFLW", "dval": "The Annotated Facial Landmarks in the Wild (AFLW) is a large-scale collection of annotated face images gathered from Flickr, exhibiting a large variety in appearance (e.g., pose, expression, ethnicity, age, gender) as well as general imaging and environmental conditions. In total about 25K faces are annotated with up to 21 landmarks per image." }, { "dkey": "AFLW2000-3D", "dval": "AFLW2000-3D is a dataset of 2000 images that have been annotated with image-level 68-point 3D facial landmarks. This dataset is used for evaluation of 3D facial landmark detection models. The head poses are very diverse and often hard to be detected by a CNN-based face detector." }, { "dkey": "FaceForensics", "dval": "FaceForensics is a video dataset consisting of more than 500,000 frames containing faces from 1004 videos that can be used to study image or video forgeries. All videos are downloaded from Youtube and are cut down to short continuous clips that contain mostly frontal faces. This dataset has two versions:\n\n\n\nSource-to-Target: where the authors reenact over 1000 videos with new facial expressions extracted from other videos, which e.g. can be used to train a classifier to detect fake images or videos.\n\n\n\nSelfreenactment: where the authors use Face2Face to reenact the facial expressions of videos with their own facial expressions as input to get pairs of videos, which e.g. can be used to train supervised generative refinement models." }, { "dkey": "300-VW", "dval": "300 Videos in the Wild (300-VW) is a dataset for evaluating facial landmark tracking algorithms in the wild. The dataset authors collected a large number of long facial videos recorded in the wild. Each video has duration of ~1 minute (at 25-30 fps). All frames have been annotated with regards to the same mark-up (i.e. set of facial landmarks) used in the 300 W competition as well (a total of 68 landmarks). The dataset includes 114 videos (circa 1 min each)." }, { "dkey": "LS3D-W", "dval": "A 3D facial landmark dataset of around 230,000 images." }, { "dkey": "WFLW", "dval": "The Wider Facial Landmarks in the Wild or WFLW database contains 10000 faces (7500 for training and 2500 for testing) with 98 annotated landmarks. This database also features rich attribute annotations in terms of occlusion, head pose, make-up, illumination, blur and expressions." }, { "dkey": "SoF", "dval": "The Specs on Faces (SoF) dataset, a collection of 42,592 (2,662×16) images for 112 persons (66 males and 46 females) who wear glasses under different illumination conditions. The dataset is FREE for reasonable academic fair use. The dataset presents a new challenge regarding face detection and recognition. It is focused on two challenges: harsh illumination environments and face occlusions, which highly affect face detection, recognition, and classification. The glasses are the common natural occlusion in all images of the dataset. However, there are two more synthetic occlusions (nose and mouth) added to each image. Moreover, three image filters, that may evade face detectors and facial recognition systems, were applied to each image. All generated images are categorized into three levels of difficulty (easy, medium, and hard). That enlarges the number of images to be 42,592 images (26,112 male images and 16,480 female images). There is metadata for each image that contains many information such as: the subject ID, facial landmarks, face and glasses rectangles, gender and age labels, year that the photo was taken, facial emotion, glasses type, and more." } ]

[ "CDTB", "Dialogue State Tracking Challenge", "Glint360K", "MARS", "VideoMem", "WikiText-103", "WikiText-2" ]

[ "VOT2018", "LaSOT" ]

[ { "dkey": "VOT2018", "dval": "VOT2018 is a dataset for visual object tracking. It consists of 60 challenging videos collected from real-life datasets." }, { "dkey": "LaSOT", "dval": "LaSOT is a high-quality benchmark for Large-scale Single Object Tracking. LaSOT consists of 1,400 sequences with more than 3.5M frames in total. Each frame in these sequences is carefully and manually annotated with a bounding box, making LaSOT one of the largest densely annotated\ntracking benchmark. The average video length of LaSOT\nis more than 2,500 frames, and each sequence comprises\nvarious challenges deriving from the wild where target objects may disappear and re-appear again in the view." }, { "dkey": "CDTB", "dval": "dataset is recorded by several passive and active RGB-D setups and contains indoor as well as outdoor sequences acquired in direct sunlight. The sequences were recorded to contain significant object pose change, clutter, occlusion, and periods of long-term target absence to enable tracker evaluation under realistic conditions. Sequences are per-frame annotated with 13 visual attributes for detailed analysis. It contains around 100,000 samples.)" }, { "dkey": "Dialogue State Tracking Challenge", "dval": "The Dialog State Tracking Challenges 2 & 3 (DSTC2&3) were research challenge focused on improving the state of the art in tracking the state of spoken dialog systems. State tracking, sometimes called belief tracking, refers to accurately estimating the user's goal as a dialog progresses. Accurate state tracking is desirable because it provides robustness to errors in speech recognition, and helps reduce ambiguity inherent in language within a temporal process like dialog.\nIn these challenges, participants were given labelled corpora of dialogs to develop state tracking algorithms. The trackers were then evaluated on a common set of held-out dialogs, which were released, un-labelled, during a one week period.\n\nThe corpus was collected using Amazon Mechanical Turk, and consists of dialogs in two domains: restaurant information, and tourist information. Tourist information subsumes restaurant information, and includes bars, cafés etc. as well as multiple new slots. There were two rounds of evaluation using this data:\n\nDSTC 2 released a large number of training dialogs related to restaurant search. Compared to DSTC (which was in the bus timetables domain), DSTC 2 introduces changing user goals, tracking 'requested slots' as well as the new restaurants domain. Results from DSTC 2 were presented at SIGDIAL 2014.\nDSTC 3 addressed the problem of adaption to a new domain - tourist information. DSTC 3 releases a small amount of labelled data in the tourist information domain; participants will use this data plus the restaurant data from DSTC 2 for training.\nDialogs used for training are fully labelled; user transcriptions, user dialog-act semantics and dialog state are all annotated. (This corpus therefore is also suitable for studies in Spoken Language Understanding.)" }, { "dkey": "Glint360K", "dval": "The largest and cleanest face recognition dataset Glint360K, \nwhich contains 17,091,657 images of 360,232 individuals, baseline models trained on Glint360K can easily achieve state-of-the-art performance." }, { "dkey": "MARS", "dval": "MARS (Motion Analysis and Re-identification Set) is a large scale video based person reidentification dataset, an extension of the Market-1501 dataset. It has been collected from six near-synchronized cameras. It consists of 1,261 different pedestrians, who are captured by at least 2 cameras. The variations in poses, colors and illuminations of pedestrians, as well as the poor image quality, make it very difficult to yield high matching accuracy. Moreover, the dataset contains 3,248 distractors in order to make it more realistic. Deformable Part Model and GMMCP tracker were used to automatically generate the tracklets (mostly 25-50 frames long)." }, { "dkey": "VideoMem", "dval": "Composed of 10,000 videos annotated with memorability scores. In contrast to previous work on image memorability -- where memorability was measured a few minutes after memorization -- memory performance is measured twice: a few minutes after memorization and again 24-72 hours later." }, { "dkey": "WikiText-103", "dval": "The WikiText language modeling dataset is a collection of over 100 million tokens extracted from the set of verified Good and Featured articles on Wikipedia. The dataset is available under the Creative Commons Attribution-ShareAlike License.\n\nCompared to the preprocessed version of Penn Treebank (PTB), WikiText-2 is over 2 times larger and WikiText-103 is over 110 times larger. The WikiText dataset also features a far larger vocabulary and retains the original case, punctuation and numbers - all of which are removed in PTB. As it is composed of full articles, the dataset is well suited for models that can take advantage of long term dependencies." }, { "dkey": "WikiText-2", "dval": "The WikiText language modeling dataset is a collection of over 100 million tokens extracted from the set of verified Good and Featured articles on Wikipedia. The dataset is available under the Creative Commons Attribution-ShareAlike License.\n\nCompared to the preprocessed version of Penn Treebank (PTB), WikiText-2 is over 2 times larger and WikiText-103 is over 110 times larger. The WikiText dataset also features a far larger vocabulary and retains the original case, punctuation and numbers - all of which are removed in PTB. As it is composed of full articles, the dataset is well suited for models that can take advantage of long term dependencies." } ]

[ "UAVA", "Talk2Car", "MARS", "DIV2K", "BLURB", "House3D Environment" ]

[ "CIFAR-10", "CelebA" ]

[ { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "CelebA", "dval": "CelebFaces Attributes dataset contains 202,599 face images of the size 178×218 from 10,177 celebrities, each annotated with 40 binary labels indicating facial attributes like hair color, gender and age." }, { "dkey": "UAVA", "dval": "The UAVA,<i>UAV-Assistant</i>, dataset is specifically designed for fostering applications which consider UAVs and humans as cooperative agents.\nWe employ a real-world 3D scanned dataset (<a href=\"https://niessner.github.io/Matterport/\">Matterport3D</a>), physically-based rendering, a gamified simulator for realistic drone navigation trajectory collection, to generate realistic multimodal data both from the user’s exocentric view of the drone, as well as the drone’s egocentric view." }, { "dkey": "Talk2Car", "dval": "The Talk2Car dataset finds itself at the intersection of various research domains, promoting the development of cross-disciplinary solutions for improving the state-of-the-art in grounding natural language into visual space. The annotations were gathered with the following aspects in mind:\nFree-form high quality natural language commands, that stimulate the development of solutions that can operate in the wild.\nA realistic task setting. Specifically, the authors consider an autonomous driving setting, where a passenger can control the actions of an Autonomous Vehicle by giving commands in natural language.\nThe Talk2Car dataset was build on top of the nuScenes dataset to include an extensive suite of sensor modalities, i.e. semantic maps, GPS, LIDAR, RADAR and 360-degree RGB images annotated with 3D bounding boxes. Such variety of input modalities sets the object referral task on the Talk2Car dataset apart from related challenges, where additional sensor modalities are generally missing." }, { "dkey": "MARS", "dval": "MARS (Motion Analysis and Re-identification Set) is a large scale video based person reidentification dataset, an extension of the Market-1501 dataset. It has been collected from six near-synchronized cameras. It consists of 1,261 different pedestrians, who are captured by at least 2 cameras. The variations in poses, colors and illuminations of pedestrians, as well as the poor image quality, make it very difficult to yield high matching accuracy. Moreover, the dataset contains 3,248 distractors in order to make it more realistic. Deformable Part Model and GMMCP tracker were used to automatically generate the tracklets (mostly 25-50 frames long)." }, { "dkey": "DIV2K", "dval": "DIV2K is a popular single-image super-resolution dataset which contains 1,000 images with different scenes and is splitted to 800 for training, 100 for validation and 100 for testing. It was collected for NTIRE2017 and NTIRE2018 Super-Resolution Challenges in order to encourage research on image super-resolution with more realistic degradation. This dataset contains low resolution images with different types of degradations. Apart from the standard bicubic downsampling, several types of degradations are considered in synthesizing low resolution images for different tracks of the challenges. Track 2 of NTIRE 2017 contains low resolution images with unknown x4 downscaling. Track 2 and track 4 of NTIRE 2018 correspond to realistic mild ×4 and realistic wild ×4 adverse conditions, respectively. Low-resolution images under realistic mild x4 setting suffer from motion blur, Poisson noise and pixel shifting. Degradations under realistic wild x4 setting are further extended to be of different levels from image to image." }, { "dkey": "BLURB", "dval": "BLURB is a collection of resources for biomedical natural language processing. In general domains such as newswire and the Web, comprehensive benchmarks and leaderboards such as GLUE have greatly accelerated progress in open-domain NLP. In biomedicine, however, such resources are ostensibly scarce. In the past, there have been a plethora of shared tasks in biomedical NLP, such as BioCreative, BioNLP Shared Tasks, SemEval, and BioASQ, to name just a few. These efforts have played a significant role in fueling interest and progress by the research community, but they typically focus on individual tasks. The advent of neural language models such as BERTs provides a unifying foundation to leverage transfer learning from unlabeled text to support a wide range of NLP applications. To accelerate progress in biomedical pretraining strategies and task-specific methods, it is thus imperative to create a broad-coverage benchmark encompassing diverse biomedical tasks.\n\nInspired by prior efforts toward this direction (e.g., BLUE), BLURB (short for Biomedical Language Understanding and Reasoning Benchmark) was created. BLURB comprises of a comprehensive benchmark for PubMed-based biomedical NLP applications, as well as a leaderboard for tracking progress by the community. BLURB includes thirteen publicly available datasets in six diverse tasks. To avoid placing undue emphasis on tasks with many available datasets, such as named entity recognition (NER), BLURB reports the macro average across all tasks as the main score. The BLURB leaderboard is model-agnostic. Any system capable of producing the test predictions using the same training and development data can participate. The main goal of BLURB is to lower the entry barrier in biomedical NLP and help accelerate progress in this vitally important field for positive societal and human impact." }, { "dkey": "House3D Environment", "dval": "A rich, extensible and efficient environment that contains 45,622 human-designed 3D scenes of visually realistic houses, ranging from single-room studios to multi-storied houses, equipped with a diverse set of fully labeled 3D objects, textures and scene layouts, based on the SUNCG dataset (Song et.al.)" } ]

[ "SNIPS", "ConvAI2", "FaceForensics", "FaceWarehouse", "RAF-DB", "UTKFace" ]

[ "FG-NET", "CACD", "MegaFace" ]

[ { "dkey": "FG-NET", "dval": "FGNet is a dataset for age estimation and face recognition across ages. It is composed of a total of 1,002 images of 82 people with age range from 0 to 69 and an age gap up to 45 years" }, { "dkey": "CACD", "dval": "The Cross-Age Celebrity Dataset (CACD) contains 163,446 images from 2,000 celebrities collected from the Internet. The images are collected from search engines using celebrity name and year (2004-2013) as keywords. Therefore, it is possible to estimate the ages of the celebrities on the images by simply subtract the birth year from the year of which the photo was taken." }, { "dkey": "MegaFace", "dval": "MegaFace was a publicly available dataset which is used for evaluating the performance of face recognition algorithms with up to a million distractors (i.e., up to a million people who are not in the test set). MegaFace contains 1M images from 690K individuals with unconstrained pose, expression, lighting, and exposure. MegaFace captures many different subjects rather than many images of a small number of subjects. The gallery set of MegaFace is collected from a subset of Flickr. The probe set of MegaFace used in the challenge consists of two databases; Facescrub and FGNet. FGNet contains 975 images of 82 individuals, each with several images spanning ages from 0 to 69. Facescrub dataset contains more than 100K face images of 530 people. The MegaFace challenge evaluates performance of face recognition algorithms by increasing the numbers of “distractors” (going from 10 to 1M) in the gallery set. In order to evaluate the face recognition algorithms fairly, MegaFace challenge has two protocols including large or small training sets. If a training set has more than 0.5M images and 20K subjects, it is considered as large. Otherwise, it is considered as small.\n\nNOTE: This dataset has been retired." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "FaceForensics", "dval": "FaceForensics is a video dataset consisting of more than 500,000 frames containing faces from 1004 videos that can be used to study image or video forgeries. All videos are downloaded from Youtube and are cut down to short continuous clips that contain mostly frontal faces. This dataset has two versions:\n\n\n\nSource-to-Target: where the authors reenact over 1000 videos with new facial expressions extracted from other videos, which e.g. can be used to train a classifier to detect fake images or videos.\n\n\n\nSelfreenactment: where the authors use Face2Face to reenact the facial expressions of videos with their own facial expressions as input to get pairs of videos, which e.g. can be used to train supervised generative refinement models." }, { "dkey": "FaceWarehouse", "dval": "FaceWarehouse is a 3D facial expression database that provides the facial geometry of 150 subjects, covering a wide range of ages and ethnic backgrounds." }, { "dkey": "RAF-DB", "dval": "The Real-world Affective Faces Database (RAF-DB) is a dataset for facial expression. It contains 29672 facial images tagged with basic or compound expressions by 40 independent taggers. Images in this database are of great variability in subjects' age, gender and ethnicity, head poses, lighting conditions, occlusions, (e.g. glasses, facial hair or self-occlusion), post-processing operations (e.g. various filters and special effects), etc." }, { "dkey": "UTKFace", "dval": "The UTKFace dataset is a large-scale face dataset with long age span (range from 0 to 116 years old). The dataset consists of over 20,000 face images with annotations of age, gender, and ethnicity. The images cover large variation in pose, facial expression, illumination, occlusion, resolution, etc. This dataset could be used on a variety of tasks, e.g., face detection, age estimation, age progression/regression, landmark localization, etc." } ]

[ "2D-3D Match Dataset", "People Snapshot Dataset", "Semantic3D", "IntrA", "ScanRefer Dataset", "Florence" ]

[ "SUNCG", "ScanNet", "ShapeNet", "Scan2CAD" ]

[ { "dkey": "SUNCG", "dval": "SUNCG is a large-scale dataset of synthetic 3D scenes with dense volumetric annotations.\n\nThe dataset is currently not available." }, { "dkey": "ScanNet", "dval": "ScanNet is an instance-level indoor RGB-D dataset that includes both 2D and 3D data. It is a collection of labeled voxels rather than points or objects. Up to now, ScanNet v2, the newest version of ScanNet, has collected 1513 annotated scans with an approximate 90% surface coverage. In the semantic segmentation task, this dataset is marked in 20 classes of annotated 3D voxelized objects." }, { "dkey": "ShapeNet", "dval": "ShapeNet is a large scale repository for 3D CAD models developed by researchers from Stanford University, Princeton University and the Toyota Technological Institute at Chicago, USA. The repository contains over 300M models with 220,000 classified into 3,135 classes arranged using WordNet hypernym-hyponym relationships. ShapeNet Parts subset contains 31,693 meshes categorised into 16 common object classes (i.e. table, chair, plane etc.). Each shapes ground truth contains 2-5 parts (with a total of 50 part classes)." }, { "dkey": "Scan2CAD", "dval": "Scan2CAD is an alignment dataset based on 1506 ScanNet scans with 97607 annotated keypoints pairs between 14225 (3049 unique) CAD models from ShapeNet and their counterpart objects in the scans. The top 3 annotated model classes are chairs, tables and cabinets which arises due to the nature of indoor scenes in ScanNet. The number of objects aligned per scene ranges from 1 to 40 with an average of 9.3.\n\nAdditionally, all ShapeNet CAD models used in the Scan2CAD dataset are annotated with their rotational symmetries: either none, 2-fold, 4-fold or infinite rotational symmetries around a canonical axis of the object." }, { "dkey": "2D-3D Match Dataset", "dval": "2D-3D Match Dataset is a new dataset of 2D-3D correspondences by leveraging the availability of several 3D datasets from RGB-D scans. Specifically, the data from SceneNN and 3DMatch are used. The training dataset consists of 110 RGB-D scans, of which 56 scenes are from SceneNN and 54 scenes are from 3DMatch. The 2D-3D correspondence data is generated as follows. Given a 3D point which is randomly sampled from a 3D point cloud, a set of 3D patches from different scanning views are extracted. To find a 2D-3D correspondence, for each 3D patch, its 3D position is re-projected into all RGB-D frames for which the point lies in the camera frustum, taking occlusion into account. The corresponding local 2D patches around the re-projected point are extracted. In total, around 1.4 millions 2D-3D correspondences are collected." }, { "dkey": "People Snapshot Dataset", "dval": "Enables detailed human body model reconstruction in clothing from a single monocular RGB video without requiring a pre scanned template or manually clicked points." }, { "dkey": "Semantic3D", "dval": "Semantic3D is a point cloud dataset of scanned outdoor scenes with over 3 billion points. It contains 15 training and 15 test scenes annotated with 8 class labels. This large labelled 3D point cloud data set of natural covers a range of diverse urban scenes: churches, streets, railroad tracks, squares, villages, soccer fields, castles to name just a few. The point clouds provided are scanned statically with state-of-the-art equipment and contain very fine details." }, { "dkey": "IntrA", "dval": "IntrA is an open-access 3D intracranial aneurysm dataset that makes the application of points-based and mesh-based classification and segmentation models available. This dataset can be used to diagnose intracranial aneurysms and to extract the neck for a clipping operation in medicine and other areas of deep learning, such as normal estimation and surface reconstruction.\n\n103 3D models of entire brain vessels are collected by reconstructing scanned 2D MRA images of patients (the raw 2D MRA images are not published due to medical ethics).\n1909 blood vessel segments are generated automatically from the complete models, including 1694 healthy vessel segments and 215 aneurysm segments for diagnosis.\n116 aneurysm segments are divided and annotated manually by medical experts; the scale of each aneurysm segment is based on the need for a preoperative examination.\nGeodesic distance matrices are computed and included for each annotated 3D segment, because the expression of the geodesic distance is more accurate than Euclidean distance according to the shape of vessels." }, { "dkey": "ScanRefer Dataset", "dval": "Contains 51,583 descriptions of 11,046 objects from 800 ScanNet scenes. ScanRefer is the first large-scale effort to perform object localization via natural language expression directly in 3D." }, { "dkey": "Florence", "dval": "The Florence 3D faces dataset consists of:\n\n\nHigh-resolution 3D scans of human faces from many subjects.\nSeveral video sequences of varying resolution, conditions and zoom level for each subject.\nEach subject is recorded in the following situations:\nIn a controlled setting in HD video.\nIn a less-constrained (but still indoor) setting using a standard, PTZ surveillance camera.\nIn an unconstrained, outdoor environment under challenging recording conditions." } ]

[ "VATEX", "Conceptual Captions", "SNIPS", "SWAG", "Fashion IQ", "BanglaLekhaImageCaptions", "Django" ]

[ "Flickr30k", "COCO" ]

[ { "dkey": "Flickr30k", "dval": "The Flickr30k dataset contains 31,000 images collected from Flickr, together with 5 reference sentences provided by human annotators." }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "VATEX", "dval": "VATEX is multilingual, large, linguistically complex, and diverse dataset in terms of both video and natural language descriptions. It has two tasks for video-and-language research: (1) Multilingual Video Captioning, aimed at describing a video in various languages with a compact unified captioning model, and (2) Video-guided Machine Translation, to translate a source language description into the target language using the video information as additional spatiotemporal context." }, { "dkey": "Conceptual Captions", "dval": "Automatic image captioning is the task of producing a natural-language utterance (usually a sentence) that correctly reflects the visual content of an image. Up to this point, the resource most used for this task was the MS-COCO dataset, containing around 120,000 images and 5-way image-caption annotations (produced by paid annotators).\n\nGoogle's Conceptual Captions dataset has more than 3 million images, paired with natural-language captions. In contrast with the curated style of the MS-COCO images, Conceptual Captions images and their raw descriptions are harvested from the web, and therefore represent a wider variety of styles. The raw descriptions are harvested from the Alt-text HTML attribute associated with web images. The authors developed an automatic pipeline that extracts, filters, and transforms candidate image/caption pairs, with the goal of achieving a balance of cleanliness, informativeness, fluency, and learnability of the resulting captions." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "SWAG", "dval": "Given a partial description like \"she opened the hood of the car,\" humans can reason about the situation and anticipate what might come next (\"then, she examined the engine\"). SWAG (Situations With Adversarial Generations) is a large-scale dataset for this task of grounded commonsense inference, unifying natural language inference and physically grounded reasoning.\n\nThe dataset consists of 113k multiple choice questions about grounded situations. Each question is a video caption from LSMDC or ActivityNet Captions, with four answer choices about what might happen next in the scene. The correct answer is the (real) video caption for the next event in the video; the three incorrect answers are adversarially generated and human verified, so as to fool machines but not humans. The authors aim for SWAG to be a benchmark for evaluating grounded commonsense NLI and for learning representations." }, { "dkey": "Fashion IQ", "dval": "Fashion IQ support and advance research on interactive fashion image retrieval. Fashion IQ is the first fashion dataset to provide human-generated captions that distinguish similar pairs of garment images together with side-information consisting of real-world product descriptions and derived visual attribute labels for these images." }, { "dkey": "BanglaLekhaImageCaptions", "dval": "This dataset consists of images and annotations in Bengali. The images are human annotated in Bengali by two adult native Bengali speakers. All popular image captioning datasets have a predominant western cultural bias with the annotations done in English. Using such datasets to train an image captioning system assumes that a good English to target language translation system exists and that the original dataset had elements of the target culture. Both these assumptions are false, leading to the need of a culturally relevant dataset in Bengali, to generate appropriate image captions of images relevant to the Bangladeshi and wider subcontinental context. The dataset presented consists of 9,154 images." }, { "dkey": "Django", "dval": "The Django dataset is a dataset for code generation comprising of 16000 training, 1000 development and 1805 test annotations. Each data point consists of a line of Python code together with a manually created natural language description." } ]

[ "BiasBios", "GYAFC", "ShapeNet", "SVT", "2D Hela" ]

[ "ImageNet", "CIFAR-10" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "BiasBios", "dval": "The purpose of this dataset was to study gender bias in occupations. Online biographies, written in English, were collected to find the names, pronouns, and occupations. Twenty-eight most frequent occupations were identified based on their appearances. The resulting dataset consists of 397,340 biographies spanning twenty-eight different occupations. Of these occupations, the professor is the most frequent, with 118,400 biographies, while the rapper is the least frequent, with 1,406 biographies. \nImportant information about the biographies:\n1. The longest biography is 194 tokens, while the shortest is eighteen; the median biography length is seventy-two tokens.\n2. It should be noted that the demographics of online biographies’ subjects differ from those of the overall workforce and that this dataset does not contain all biographies on the Internet." }, { "dkey": "GYAFC", "dval": "Grammarly’s Yahoo Answers Formality Corpus (GYAFC) is the largest dataset for any style containing a total of 110K informal / formal sentence pairs.\n\nYahoo Answers is a question answering forum, contains a large number of informal sentences and allows redistribution of data. The authors used the Yahoo Answers L6 corpus to create the GYAFC dataset of informal and formal sentence pairs. In order to ensure a uniform distribution of data, they removed sentences that are questions, contain URLs, and are shorter than 5 words or longer than 25. After these preprocessing steps, 40 million sentences remain. \n\nThe Yahoo Answers corpus consists of several different domains like Business, Entertainment & Music, Travel, Food, etc. Pavlick and Tetreault formality classifier (PT16) shows that the formality level varies significantly\nacross different genres. In order to control for this variation, the authors work with two specific domains that contain the most informal sentences and show results on training and testing within those categories. The authors use the formality classifier from PT16 to identify informal sentences and train this classifier on the Answers genre of the PT16 corpus\nwhich consists of nearly 5,000 randomly selected sentences from Yahoo Answers manually annotated on a scale of -3 (very informal) to 3 (very formal). They find that the domains of Entertainment & Music and Family & Relationships contain the most informal sentences and create the GYAFC dataset using these domains." }, { "dkey": "ShapeNet", "dval": "ShapeNet is a large scale repository for 3D CAD models developed by researchers from Stanford University, Princeton University and the Toyota Technological Institute at Chicago, USA. The repository contains over 300M models with 220,000 classified into 3,135 classes arranged using WordNet hypernym-hyponym relationships. ShapeNet Parts subset contains 31,693 meshes categorised into 16 common object classes (i.e. table, chair, plane etc.). Each shapes ground truth contains 2-5 parts (with a total of 50 part classes)." }, { "dkey": "SVT", "dval": "The Street View Text (SVT) dataset was harvested from Google Street View. Image text in this data exhibits high variability and often has low resolution. In dealing with outdoor street level imagery, we note two characteristics. (1) Image text often comes from business signage and (2) business names are easily available through geographic business searches. These factors make the SVT set uniquely suited for word spotting in the wild: given a street view image, the goal is to identify words from nearby businesses.\n\nNote: the dataset has undergone revision since the time it was evaluated in this publication. Please consult the ICCV2011 paper for most up-to-date results." }, { "dkey": "2D Hela", "dval": "2D HeLa is a dataset of fluorescence microscopy images of HeLa cells stained with various organelle-specific fluorescent dyes. The images include 10 organelles, which are DNA (Nuclei), ER (Endoplasmic reticulum), Giantin, (cis/medial Golgi), GPP130 (cis Golgi), Lamp2 (Lysosomes), Mitochondria, Nucleolin (Nucleoli), Actin, TfR (Endosomes), Tubulin.\nThe purpose of the dataset is to train a computer program to automatically identify sub-cellular organelles.\n\nPaper: M. V. Boland and R. F. Murphy (2001). A Neural Network Classifier Capable of Recognizing the Patterns of all Major Subcellular Structures in Fluorescence Microscope Images of HeLa Cells. Bioinformatics 17:1213-1223" } ]

[ "e-SNLI", "GLUE", "IMPPRES", "SNLI-VE", "NLI-TR", "SNIPS", "Violin" ]

[ "Flickr30k", "SNLI", "MultiNLI", "SentEval" ]

[ { "dkey": "Flickr30k", "dval": "The Flickr30k dataset contains 31,000 images collected from Flickr, together with 5 reference sentences provided by human annotators." }, { "dkey": "SNLI", "dval": "The SNLI dataset (Stanford Natural Language Inference) consists of 570k sentence-pairs manually labeled as entailment, contradiction, and neutral. Premises are image captions from Flickr30k, while hypotheses were generated by crowd-sourced annotators who were shown a premise and asked to generate entailing, contradicting, and neutral sentences. Annotators were instructed to judge the relation between sentences given that they describe the same event. Each pair is labeled as “entailment”, “neutral”, “contradiction” or “-”, where “-” indicates that an agreement could not be reached." }, { "dkey": "MultiNLI", "dval": "The Multi-Genre Natural Language Inference (MultiNLI) dataset has 433K sentence pairs. Its size and mode of collection are modeled closely like SNLI. MultiNLI offers ten distinct genres (Face-to-face, Telephone, 9/11, Travel, Letters, Oxford University Press, Slate, Verbatim, Goverment and Fiction) of written and spoken English data. There are matched dev/test sets which are derived from the same sources as those in the training set, and mismatched sets which do not closely resemble any seen at training time." }, { "dkey": "SentEval", "dval": "SentEval is a toolkit for evaluating the quality of universal sentence representations. SentEval encompasses a variety of tasks, including binary and multi-class classification, natural language inference and sentence similarity. The set of tasks was selected based on what appears to be the community consensus regarding the appropriate evaluations for universal sentence representations. The toolkit comes with scripts to download and preprocess datasets, and an easy interface to evaluate sentence encoders." }, { "dkey": "e-SNLI", "dval": "e-SNLI is used for various goals, such as obtaining full sentence justifications of a model's decisions, improving universal sentence representations and transferring to out-of-domain NLI datasets." }, { "dkey": "GLUE", "dval": "General Language Understanding Evaluation (GLUE) benchmark is a collection of nine natural language understanding tasks, including single-sentence tasks CoLA and SST-2, similarity and paraphrasing tasks MRPC, STS-B and QQP, and natural language inference tasks MNLI, QNLI, RTE and WNLI." }, { "dkey": "IMPPRES", "dval": "An IMPlicature and PRESupposition diagnostic dataset (IMPPRES), consisting of >25k semiautomatically generated sentence pairs illustrating well-studied pragmatic inference types." }, { "dkey": "SNLI-VE", "dval": "Visual Entailment (VE) consists of image-sentence pairs whereby a premise is defined by an image, rather than a natural language sentence as in traditional Textual Entailment tasks. The goal of a trained VE model is to predict whether the image semantically entails the text. SNLI-VE is a dataset for VE which is based on the Stanford Natural Language Inference corpus and Flickr30k dataset." }, { "dkey": "NLI-TR", "dval": "Natural Language Inference in Turkish (NLI-TR) provides translations of two large English NLI datasets into Turkish and had a team of experts validate their translation quality and fidelity to the original labels." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "Violin", "dval": "Video-and-Language Inference is the task of joint multimodal understanding of video and text. Given a video clip with aligned subtitles as premise, paired with a natural language hypothesis based on the video content, a model needs to infer whether the hypothesis is entailed or contradicted by the given video clip. The Violin dataset is a dataset for this task which consists of 95,322 video-hypothesis pairs from 15,887 video clips, spanning over 582 hours of video. These video clips contain rich content with diverse temporal dynamics, event shifts, and people interactions, collected from two sources: (i) popular TV shows, and (ii) movie clips from YouTube channels." } ]

[ "RITE", "HRF", "ADAM", "G1020", "ORVS", "ROSE" ]

[ "STARE", "DRIVE", "CHASE_DB1" ]

[ { "dkey": "STARE", "dval": "The STARE (Structured Analysis of the Retina) dataset is a dataset for retinal vessel segmentation. It contains 20 equal-sized (700×605) color fundus images. For each image, two groups of annotations are provided.." }, { "dkey": "DRIVE", "dval": "The Digital Retinal Images for Vessel Extraction (DRIVE) dataset is a dataset for retinal vessel segmentation. It consists of a total of JPEG 40 color fundus images; including 7 abnormal pathology cases. The images were obtained from a diabetic retinopathy screening program in the Netherlands. The images were acquired using Canon CR5 non-mydriatic 3CCD camera with FOV equals to 45 degrees. Each image resolution is 584*565 pixels with eight bits per color channel (3 channels). \n\nThe set of 40 images was equally divided into 20 images for the training set and 20 images for the testing set. Inside both sets, for each image, there is circular field of view (FOV) mask of diameter that is approximately 540 pixels. Inside training set, for each image, one manual segmentation by an ophthalmological expert has been applied. Inside testing set, for each image, two manual segmentations have been applied by two different observers, where the first observer segmentation is accepted as the ground-truth for performance evaluation." }, { "dkey": "CHASE_DB1", "dval": "CHASE_DB1 is a dataset for retinal vessel segmentation which contains 28 color retina images with the size of 999×960 pixels which are collected from both left and right eyes of 14 school children. Each image is annotated by two independent human experts." }, { "dkey": "RITE", "dval": "The RITE (Retinal Images vessel Tree Extraction) is a database that enables comparative studies on segmentation or classification of arteries and veins on retinal fundus images, which is established based on the public available DRIVE database (Digital Retinal Images for Vessel Extraction).\n\nRITE contains 40 sets of images, equally separated into a training subset and a test subset, the same as DRIVE. The two subsets are built from the corresponding two subsets in DRIVE. For each set, there is a fundus photograph, a vessel reference standard, and a Arteries/Veins (A/V) reference standard. \n\n\nThe fundus photograph is inherited from DRIVE. \nFor the training set, the vessel reference standard is a modified version of 1st_manual from DRIVE. \nFor the test set, the vessel reference standard is 2nd_manual from DRIVE. \nFor the A/V reference standard, four types of vessels are labelled using four colors based on the vessel reference standard. \nArteries are labelled in red; veins are labelled in blue; the overlapping of arteries and veins are labelled in green; the vessels which are uncertain are labelled in white. \nThe fundus photograph is in tif format. And the vessel reference standard and the A/V reference standard are in png format. \n\nThe dataset is described in more detail in our paper, which you will cite if you use the dataset in any way: \n\nHu Q, Abràmoff MD, Garvin MK. Automated separation of binary overlapping trees in low-contrast color retinal images. Med Image Comput Comput Assist Interv. 2013;16(Pt 2):436-43. PubMed PMID: 24579170 https://doi.org/10.1007/978-3-642-40763-5_54" }, { "dkey": "HRF", "dval": "The HRF dataset is a dataset for retinal vessel segmentation which comprises 45 images and is organized as 15 subsets. Each subset contains one healthy fundus image, one image of patient with diabetic retinopathy and one glaucoma image. The image sizes are 3,304 x 2,336, with a training/testing image split of 22/23." }, { "dkey": "ADAM", "dval": "ADAM is organized as a half day Challenge, a Satellite Event of the ISBI 2020 conference in Iowa City, Iowa, USA.\n\nThe ADAM challenge focuses on the investigation and development of algorithms associated with the diagnosis of Age-related Macular degeneration (AMD) and segmentation of lesions in fundus photos from AMD patients. The goal of the challenge is to evaluate and compare automated algorithms for the detection of AMD on a common dataset of retinal fundus images. We invite the medical image analysis community to participate by developing and testing existing and novel automated fundus classification and segmentation methods.\n\nInstructions: \nADAM: Automatic Detection challenge on Age-related Macular degeneration\n\nLink: https://amd.grand-challenge.org\n\nAge-related macular degeneration, abbreviated as AMD, is a degenerative disorder in the macular region. It mainly occurs in people older than 45 years old and its incidence rate is even higher than diabetic retinopathy in the elderly. \n\nThe etiology of AMD is not fully understood, which could be related to multiple factors, including genetics, chronic photodestruction effect, and nutritional disorder. AMD is classified into Dry AMD and Wet AMD. Dry AMD (also called nonexudative AMD) is not neovascular. It is characterized by progressive atrophy of retinal pigment epithelium (RPE). In the late stage, drusen and the large area of atrophy could be observed under ophthalmoscopy. Wet AMD (also called neovascular or exudative AMD), is characterized by active neovascularization under RPE, subsequently causing exudation, hemorrhage, and scarring, and will eventually cause irreversible damage to the photoreceptors and rapid vision loss if left untreated.\n\nAn early diagnosis of AMD is crucial to treatment and prognosis. Fundus photo is one of the basic examinations. The current dataset is composed of AMD and non-AMD (myopia, normal control, etc.) photos. Typical signs of AMD that can be found in these photos include drusen, exudation, hemorrhage, etc. \n\nThe ADAM challenge has 4 tasks:\n\nTask 1: Classification of AMD and non-AMD fundus images.\n\nTask 2: Detection and segmentation of optic disc.\n\nTask 3: Localization of fovea.\n\nTask 4: Detection and Segmentation of lesions from fundus images." }, { "dkey": "G1020", "dval": "A large publicly available retinal fundus image dataset for glaucoma classification called G1020. The dataset is curated by conforming to standard practices in routine ophthalmology and it is expected to serve as standard benchmark dataset for glaucoma detection. This database consists of 1020 high resolution colour fundus images and provides ground truth annotations for glaucoma diagnosis, optic disc and optic cup segmentation, vertical cup-to-disc ratio, size of neuroretinal rim in inferior, superior, nasal and temporal quadrants, and bounding box location for optic disc." }, { "dkey": "ORVS", "dval": "The ORVS dataset has been newly established as a collaboration between the computer science and visual-science departments at the University of Calgary.\n\nThis dataset contains 49 images (42 training and seven testing images) collected from a clinic in Calgary-Canada. All images were acquired with a Zeiss Visucam 200 with 30 degrees field of view (FOV). The image size is 1444×1444 with 24 bits per pixel. Images and are stored in JPEG format with low compression, which is common in ophthalmology practice. All images were manually traced by an expert who a has been working in the field of retinal-image analysis and went through training. The expert was asked to label all pixels belonging to retinal vessels. The Windows Paint 3D tool was used to manually label the images." }, { "dkey": "ROSE", "dval": "Retinal OCTA SEgmentation dataset (ROSE) consists of 229 OCTA images with vessel annotations at either centerline-level or pixel level." } ]