README.md

metadata

license: mit
dataset_info:
  features:
    - name: input
      dtype: image
    - name: output
      dtype: image
    - name: instruction
      dtype: string
    - name: source
      dtype: string
  splits:
    - name: train
      num_bytes: 46176211620.62
      num_examples: 169180
  download_size: 28282347308
  dataset_size: 46176211620.62
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
task_categories:
  - image-to-image
language:
  - en
pretty_name: AURORA
size_categories:
  - 100K<n<1M

Read the paper here: https://arxiv.org/abs/2407.03471. IMPORTANT: Please check out our GitHub repository for more instructions on how to also access the Something-Something-Edit subdataset, which we can't publish directly: https://github.com/McGill-NLP/AURORA