metadata
license: openrail
language:
- en
task_categories:
- document-question-answering
size_categories:
- n<1K
Dataset Card for funsd-vqa
Dataset Description
- Homepage: https://huggingface.co/datasets/munish0838/funsd_vqa
- Repository: https://github.com/munish0838/FUNSD
- Point of Contact: munishkumar19042002@gmail.com
Dataset Summary
This dataset has been processed to be used by Donut model for DocVQA fine tuninf on FUNSD dataset. The final dataset is in .jsonl file format.
Languages
- English
Dataset Structure
Data Fields
id -> Name of Image file/json file
file_name -> Path of image file
questions -> array of all questions in corresponding to the image
words -> list of all words present in image
bounding_boxes -> contains bounding box of all words
answers -> array of all answers in corresponding to the image
grount_truth -> has gt_parses in donut required format for processing
Dataset Creation
Refer this github repo link: https://github.com/munish0838/FUNSD