This repository is part of the implementation of the handwritten text recognition experiments in the paper Fast writer adaptation with style extractor network for handwritten text recognition. The code is developed based on the Pytorch framework and some code in https://github.com/vloison/Handwritten_Text_Recognition is reused.
For the HETR task (experiments on IAM):
(1)If you only need the backbone recognition network:
,just run the train_CTC_HAM_Vis_Contex.py;
(2)Train the writer style extractor network:
, run the train_WID.py;
(3)Now, you can train the adaptation network, run the train_FWA.py.
The whole pipeline is shown in Alg. 1 and Alg. 2:
If you use our code in your research or wish to refer to the baseline results, please use the following BibTeX entry.
@article{wang2022fast,
title={Fast writer adaptation with style extractor network for handwritten text recognition},
author={Wang, Zi-Rui and Du, Jun},
journal={Neural Networks},
volume={147},
pages={42--52},
year={2022},
publisher={Elsevier}
}