metadata
dataset_info:
features:
- name: image_path
dtype: string
- name: caption
dtype: string
splits:
- name: train
num_bytes: 55751185
num_examples: 413803
- name: valid
num_bytes: 26444641
num_examples: 202512
download_size: 0
dataset_size: 82195826
Dataset Card for "ar_mscoco2014_text_augmented"
This dataset repo contains the jsonl files for the training and validation dataset of MSCOCO dataset. The 5 captions are unfolded and so we have trainsize * 5 and validsize *5.