--- dataset_info: - config_name: Baseball_bat features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string - name: st3_masks sequence: image - name: st3_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st3_scores sequence: float64 - name: st3_sorts sequence: int64 - name: st3_eachround sequence: string splits: - name: train num_bytes: 252424296.0 num_examples: 400 download_size: 246777468 dataset_size: 252424296.0 - config_name: Bird features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string - name: st3_masks sequence: image - name: st3_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st3_scores sequence: float64 - name: st3_sorts sequence: int64 - name: st3_eachround sequence: string splits: - name: train num_bytes: 230220917.0 num_examples: 400 download_size: 228585015 dataset_size: 230220917.0 - config_name: Breast features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string - name: st3_masks sequence: image - name: st3_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st3_scores sequence: float64 - name: st3_sorts sequence: int64 - name: st3_eachround sequence: string splits: - name: train num_bytes: 98226368.0 num_examples: 400 download_size: 94721489 dataset_size: 98226368.0 - config_name: Bus features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string - name: st3_masks sequence: image - name: st3_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st3_scores sequence: float64 - name: st3_sorts sequence: int64 - name: st3_eachround sequence: string splits: - name: train num_bytes: 272616938.0 num_examples: 400 download_size: 263304236 dataset_size: 272616938.0 - config_name: Cat features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string - name: st3_masks sequence: image - name: st3_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st3_scores sequence: float64 - name: st3_sorts sequence: int64 - name: st3_eachround sequence: string - name: st4_masks sequence: image - name: st4_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st4_scores sequence: float64 - name: st4_sorts sequence: int64 - name: st4_eachround sequence: string splits: - name: train num_bytes: 243745854.0 num_examples: 400 download_size: 237513847 dataset_size: 243745854.0 - config_name: Chalk_group features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string splits: - name: train num_bytes: 56274325.0 num_examples: 200 download_size: 54738809 dataset_size: 56274325.0 - config_name: Clock features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string - name: st3_masks sequence: image - name: st3_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st3_scores sequence: float64 - name: st3_sorts sequence: int64 - name: st3_eachround sequence: string splits: - name: train num_bytes: 252120550.0 num_examples: 400 download_size: 235007162 dataset_size: 252120550.0 - config_name: Cow features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string splits: - name: train num_bytes: 320848441.0 num_examples: 400 download_size: 306582420 dataset_size: 320848441.0 - config_name: Dog features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string - name: st3_masks sequence: image - name: st3_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st3_scores sequence: float64 - name: st3_sorts sequence: int64 - name: st3_eachround sequence: string splits: - name: train num_bytes: 267627619.0 num_examples: 400 download_size: 256377237 dataset_size: 267627619.0 - config_name: Dolphin_above features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string - name: st3_masks sequence: image - name: st3_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st3_scores sequence: float64 - name: st3_sorts sequence: int64 - name: st3_eachround sequence: string splits: - name: train num_bytes: 199319772.0 num_examples: 400 download_size: 187854559 dataset_size: 199319772.0 - config_name: Dolphin_below features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string - name: st3_masks sequence: image - name: st3_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st3_scores sequence: float64 - name: st3_sorts sequence: int64 - name: st3_eachround sequence: string splits: - name: train num_bytes: 146521148.0 num_examples: 400 download_size: 130324944 dataset_size: 146521148.0 - config_name: Pizza features: - name: images dtype: image - name: labels dtype: image splits: - name: train num_bytes: 241737760.0 num_examples: 400 download_size: 241312980 dataset_size: 241737760.0 - config_name: Polyp features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string - name: st3_masks sequence: image - name: st3_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st3_scores sequence: float64 - name: st3_sorts sequence: int64 - name: st3_eachround sequence: string splits: - name: train num_bytes: 207519159.0 num_examples: 400 download_size: 194635329 dataset_size: 207519159.0 - config_name: Salt_dome features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string - name: st3_masks sequence: image - name: st3_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st3_scores sequence: float64 - name: st3_sorts sequence: int64 - name: st3_eachround sequence: string splits: - name: train num_bytes: 54670772.0 num_examples: 200 download_size: 53194891 dataset_size: 54670772.0 - config_name: Skin features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string - name: st3_masks sequence: image - name: st3_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st3_scores sequence: float64 - name: st3_sorts sequence: int64 - name: st3_eachround sequence: string - name: st4_masks sequence: image - name: st4_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st4_scores sequence: float64 - name: st4_sorts sequence: int64 - name: st4_eachround sequence: string splits: - name: train num_bytes: 1262739004.0 num_examples: 400 download_size: 1237901679 dataset_size: 1262739004.0 - config_name: Stop_sign features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string - name: st3_masks sequence: image - name: st3_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st3_scores sequence: float64 - name: st3_sorts sequence: int64 - name: st3_eachround sequence: string splits: - name: train num_bytes: 234295025.0 num_examples: 400 download_size: 231574849 dataset_size: 234295025.0 - config_name: Tie features: - name: images dtype: image - name: labels dtype: image - name: st1_masks sequence: image - name: st1_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st1_scores sequence: float64 - name: st1_sorts sequence: int64 - name: st1_eachround sequence: string - name: st2_masks sequence: image - name: st2_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st2_scores sequence: float64 - name: st2_sorts sequence: int64 - name: st2_eachround sequence: string - name: st3_masks sequence: image - name: st3_points struct: - name: green sequence: sequence: sequence: int64 - name: red sequence: sequence: sequence: int64 - name: st3_scores sequence: float64 - name: st3_sorts sequence: int64 - name: st3_eachround sequence: string splits: - name: train num_bytes: 224976900.0 num_examples: 400 download_size: 220259941 dataset_size: 224976900.0 configs: - config_name: Baseball_bat data_files: - split: train path: Baseball_bat/train-* - config_name: Bird data_files: - split: train path: Bird/train-* - config_name: Breast data_files: - split: train path: Breast/train-* - config_name: Bus data_files: - split: train path: Bus/train-* - config_name: Cat data_files: - split: train path: Cat/train-* - config_name: Chalk_group data_files: - split: train path: Chalk_group/train-* - config_name: Clock data_files: - split: train path: Clock/train-* - config_name: Cow data_files: - split: train path: Cow/train-* - config_name: Dog data_files: - split: train path: Dog/train-* - config_name: Dolphin_above data_files: - split: train path: Dolphin_above/train-* - config_name: Dolphin_below data_files: - split: train path: Dolphin_below/train-* - config_name: Pizza data_files: - split: train path: Pizza/train-* - config_name: Polyp data_files: - split: train path: Polyp/train-* - config_name: Salt_dome data_files: - split: train path: Salt_dome/train-* - config_name: Skin data_files: - split: train path: Skin/train-* - config_name: Stop_sign data_files: - split: train path: Stop_sign/train-* - config_name: Tie data_files: - split: train path: Tie/train-* --- # Abstract The remarkable capabilities of the Segment Anything Model (SAM) for tackling image segmentation tasks in an intuitive and interactive manner has sparked interest in the design of effective visual prompts. Such interest has led to the creation of automated point prompt selection strategies, typically motivated from a feature extraction perspective. However, there is still very little understanding of how appropriate these automated visual prompting strategies are, particularly when compared to humans, across diverse image domains. Additionally, the performance benefits of including such automated visual prompting strategies within the finetuning process of SAM also remains unexplored, as does the effect of interpretable factors like distance between the prompt points on segmentation performance. To bridge these gaps, we leverage a recently released visual prompting dataset, PointPrompt, and introduce a number of benchmarking tasks that provide an array of opportunities to improve the understanding of the way human prompts differ from automated ones and what underlying factors make for effective visual prompts. We demonstrate that the resulting segmentation scores obtained by humans are approximately 29% higher than those given by automated strategies and identify potential features that are indicative of prompting performance with R2 scores over 0.5. Additionally, we demonstrate that performance when using automated methods can be improved by up to 68% via a finetuning approach. Overall, our experiments not only showcase the existing gap between human prompts and automated methods, but also highlight potential avenues through which this gap can be leveraged to improve effective visual prompt design. Further details along with the dataset links and codes are available at [this link](https://alregib.ece.gatech.edu/pointprompt-a-visual-prompting-dataset-based-on-the-segment-anything-model/). # Prompting Data - **Masks**: Contains a list of the binary masks produced for each image, where `masks[i]` contains the mask for during timestamp `i` - **Points**: Contains the inclusion and exclusion points. Each image has an outer list of size `(t,)` where `t` is the number of timesteps for that image and an inner list of size `(n, 2)` where `n` is the number of points at a given timestep - **Scores**: Contains the scores at each timestep for every image (mIoU) - **Sorts**: Contains sorted timestamp indexes, going from max to min based on the score - **Eachround**: Indicates which timesteps belong to each of the two rounds (if they exist). Each entry contains a list of lenght `t` (number of timestamps) where values of `0` corresponds to timestamps that belong to the first round and values of `1` correspond to timestamps that belong to the second round # Quick usage: - To get the best (highes score) mask for a given image : `masks[sorts[0]]` - To get the best set of prompts for that image : `green[sorts[0]]` and `red[sorts[0]]` - To get which round produced the highest score in that image : `eachround[sorts[0]]` # Data Download Sample code to download the dataset: ```python from datasets import load_dataset # Download the 'Bird' subset from HuggingFace pointprompt_bird = load_dataset('gOLIVES/PointPrompt', 'Bird', split='train') # Print the scores the scores from st1 for the first image print(pointprompt_bird[0]['st1_scores']) ``` # Links **Associated Website**: https://alregib.ece.gatech.edu/ **Paper**: https://arxiv.org/pdf/2410.22048v1 # Citation If you find the work useful, please include the following citation in your work: ``` J. Quesada∗, Z. Fowler∗, M. Alotaibi, M. Prabhushankar, and G. AlRegib, ”Benchmarking Human and Automated Prompting in the Segment Anything Model”, In IEEE International Conference on Big Data 2024, Washington DC, USA ```