Datasets:
Tasks:
Visual Question Answering
Modalities:
Text
Formats:
text
Sub-tasks:
visual-question-answering
Size:
10K - 100K
License:
metadata
language:
- vi
- en
pretty_name: TextVQA in Vietnamese
source-datasets:
- TextVQA
- OpenImages
tags:
- TextVQA
- TextVQA-vi
license: unknown
task_categories:
- visual-question-answering
task_ids:
- visual-question-answering
TextVQA in Vietnamese
This is Google-translated version of TextVQA in Vietnamese. The process of building Vietnamese version as follows:
- In en/ folder,
- Download
TextVQA_0.5.1_train.json
,TextVQA_0.5.1_val.json
. - By using set data structure, generate txt files of unique text: train_answer_list.txt, train_question_list.txt, val_answer_list.txt, val_question_list.txt.
- Download
- In vi/ folder
- By translating 4 en/.txt files, generate train_answer_list.jsonl, train_question_list.jsonl, val_answer_list.jsonl, val_question_list.jsonl. In each of entry of each file, the key is the original english text, the value is the translated text in vietnamese.
To load Vietnamese version in your code, you need original English version. Then just use English text as key to retrieve Vietnamese value from jsonl files. I provide both English and Vietnamese version.