metadata
language:
- en
config_names:
- query
- corpus
- qrels
tags:
- information-retrieval
- multimodal-retrieval
dataset_info:
- config_name: qrels
features:
- name: query-id
dtype: string
- name: Q0
dtype: string
- name: corpus-id
dtype: string
- name: score
dtype: int8
splits:
- name: test
num_examples: 37604
- config_name: corpus
features:
- name: id
dtype: string
- name: modality
dtype: string
- name: text
dtype: string
- name: image
dtype: image
splits:
- name: corpus
num_examples: 114516
- config_name: query
features:
- name: id
dtype: string
- name: modality
dtype: string
- name: text
dtype: string
- name: image
dtype: image
splits:
- name: test
num_examples: 5046
configs:
- config_name: qrels
data_files:
- split: test
path: qrels.parquet
- config_name: corpus
data_files:
- split: corpus
path: corpus.parquet
- config_name: query
data_files:
- split: test
path: query.parquet
license: cc-by-4.0
OK-VQA Image-Text to Text Retreival
@inproceedings{marino2019ok,
title={Ok-vqa: A visual question answering benchmark requiring external knowledge},
author={Marino, Kenneth and Rastegari, Mohammad and Farhadi, Ali and Mottaghi, Roozbeh},
booktitle={Proceedings of the IEEE/cvf conference on computer vision and pattern recognition},
pages={3195--3204},
year={2019}
}