method: Donut (fine-tuned)2023-03-17

Authors: Ryota Nakao, Youngmin Baek, Yamato Okamoto, Shuta Ichimura, Yu Nakagome

Affiliation: LINE Corporation

Description: We finetuned Donut (model: https://huggingface.co/naver-clova-ix/donut-base-finetuned-zhtrainticket/tree/official) with augmentations and label smoothing.
We filtered out 105 images from the training set because they seems to have incorrect labels.

@misc{kim2022ocrfree, title={OCR-free Document Understanding Transformer}, author={Geewook Kim and Teakgyu Hong and Moonbin Yim and Jeongyeon Nam and Jinyoung Park and Jinyeong Yim and Wonseok Hwang and Sangdoo Yun and Dongyoon Han and Seunghyun Park}, year={2022}, eprint={2111.15664}, archivePrefix={arXiv}, primaryClass={cs.LG} }

Source code