metadata
dataset_info:
features:
- name: image
dtype: image
- name: question
dtype: string
- name: answer
dtype: string
- name: dataset
dtype: string
splits:
- name: test
num_bytes: 338910200
num_examples: 886
- name: train
num_bytes: 1020339788.41
num_examples: 2743
download_size: 379500980
dataset_size: 1359249988.4099998
configs:
- config_name: default
data_files:
- split: test
path: data/test-*
- split: train
path: data/train-*
task_categories:
- visual-question-answering
- image-feature-extraction
language:
- fr
license: cc-by-nc-4.0
Description
French split of the ByteDance/MTVQA dataset that we processed.
Citation
@misc{tang2024mtvqa,
title={MTVQA: Benchmarking Multilingual Text-Centric Visual Question Answering},
author={Jingqun Tang and Qi Liu and Yongjie Ye and Jinghui Lu and Shu Wei and Chunhui Lin and Wanqing Li and Mohamad Fitri Faiz Bin Mahmood and Hao Feng and Zhen Zhao and Yanjie Wang and Yuliang Liu and Hao Liu and Xiang Bai and Can Huang},
year={2024},
eprint={2405.11985},
archivePrefix={arXiv},
primaryClass={cs.CV}
}