coco_qa_dataset / README.md
Slicky325's picture
Merged COCO caption dataset with QA pairs. 5000 exact matches found from 5000 base captions and 5000 QA pairs.
80cf23f verified
metadata
dataset_info:
  features:
    - name: caption
      dtype: string
    - name: question
      dtype: string
    - name: answer
      dtype: string
    - name: image
      dtype: image
  splits:
    - name: train
      num_bytes: 257085349
      num_examples: 5000
  download_size: 256383268
  dataset_size: 257085349
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*