search

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
5.83k stars 476 forks source link

synthdog for document generation using template #144

Open Wyzix33 opened 1 year ago

Wyzix33 commented 1 year ago

Hi, is it possible to use synthdoc to generate images based on a template (layout)? Let's say that i want to fine-tune donut on a specific document, but i only have 10 of them, it would be great to be able to generate similar documents with text from a dataset added to specific coordinates, length and type to be used for fine tune. Can this be done or do you know such a script that can take a image or pdf as template and a text file for configuration where the random text should be inserted? Thanks

bswethav commented 9 months ago

did your find answer to your question. i have similar requirement

hengrui0516 commented 6 months ago

I wonder answer to this question, thx!