donut
Official Implementation of OCR-free Document Understanding Transformer (Donut) and Synthetic Document Generator (SynthDoG), ECCV 2022
MIT License
Ecosystems:
Python
Bot releases are hidden (Show)
- enhance CPU performance by removing torch.bfloat16 related lines 1f4b27c
- update comments and README
donut
- 1.0.8
Published by gwkrsrch about 2 years ago
- support newer versions of
transformerhttps://github.com/clovaai/donut/pull/56 - update
JSONParseEvaluatore33da0e- enhance the logic in normalize_dict and update_cost
- update comments and README
donut
- 1.0.7
Published by gwkrsrch about 2 years ago
- update
JSONParseEvaluator86bcafe- add functions to calculate (strict) F1 accuracy score
- enhance the logic in normalize_dict
- update docstrings and comments
donut
- 1.0.5
Published by gwkrsrch about 2 years ago
-
donut-pythonloads the pre-trained weights from theofficialbranch of the model repositories f174bb0
donut
- 1.0.4
Published by gwkrsrch about 2 years ago
- fixed some minor issues
- convert PIL image to RGB in
prepare_input74507ce - set
max_iterwithconfig.max_steps4130973
- convert PIL image to RGB in
donut
- 1.0.3
Published by gwkrsrch about 2 years ago
- Initial release
Package Rankings
Top 6.29% on Pypi.org
Badges
Extracted from project README
