metadata

license: apache-2.0
size_categories:
  - 1K<n<10K
task_categories:
  - zero-shot-object-detection
  - zero-shot-image-classification
dataset_info:
  features:
    - name: image
      dtype: image
  splits:
    - name: train
      num_bytes: 1486623411.242
      num_examples: 2338
  download_size: 1482027997
  dataset_size: 1486623411.242
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
tags:
  - flying
  - objects
  - drones
  - planes

Described Flying Objects

This dataset was created for the DSTA BrainHack TIL 2024 competition's VLM task. It includes various flying objects in backgrounds, and augmented with noise. Bounding box and short text description are provided. This is meant for open vocabulary object detection/classification tasks.

This dataset was created by Ooi Xuan Shan from team 12000SGDPLUSHIE. Creation code can be found here.

This dataset includes 2338 images, 9214 bboxes, 69 unique flying objects with their captions.