DOCVQA data set format ?

clovaai / donut

Official Implementation of OCR-free Document Understanding Transformer (Donut) and Synthetic Document Generator (SynthDoG), ECCV 2022

https://arxiv.org/abs/2111.15664

MIT License

5.74k stars 466 forks source link

DOCVQA data set format ? #281

Open tzktz opened 8 months ago

tzktz commented 8 months ago

Any one knows about docvqa dataset format to train my custom dataset...thx in advance :)

svocdfrockz commented 8 months ago

Facing the same issue.. @gwkrsrch , request your help here. What is the gt parse , groundtruth, metadata.jsonl format when we have single answer for one question. And we have lets say 5 qns. Please tell the format for creating dataset for finetuning donut docvqa.

praneetreddy017 commented 8 months ago

Are you guys still facing this issue? Drop a comment and I'll help out with some examples and images

svocdfrockz commented 8 months ago

Yes still facing the same issue

I want to know the docvqa format one question has one answer