metadata
dataset_info:
features:
- name: img_id
dtype: int64
- name: img_filename
dtype: string
- name: caption
dtype: string
- name: QA_pairs
sequence:
sequence: string
splits:
- name: 2018_train
num_bytes: 68830965
num_examples: 123287
download_size: 30828814
dataset_size: 68830965