metadata
dataset_info:
features:
- name: image
dtype: image
- name: question
dtype: string
- name: ground_truth
dtype: string
splits:
- name: train
num_bytes: 8164044640.98
num_examples: 68980
download_size: 2667555722
dataset_size: 8164044640.98
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
screen-qa-full
The ScreenQA dataset obtained from github.com/google-research-datasets/screen_qa.
Added images from the Rico dataset.