Skip to content

This is an implementation for the paper Enhance to Read Better: A Multi-Task Adversarial Network for Handwritten Document Image Enhancement designed to enhance the document quality before the recognition process. It could be used for document cleaning and binarization. The weights are available to test the enhancement.

License

Notifications You must be signed in to change notification settings

sanakhamekhem/GAN-HTR

Repository files navigation

Description

This is an implementation for the paper "Enhance to Read Better: A Multi-Task Adversarial Network for Handwritten Document Image Enhancement" designed to enhance the document quality before the recognition process. It could be used for document cleaning and binarization.

• A Generative Adversarial Network for handwritten document image binarization.

• We perform document binarization while ensuring text readability, simultaneously, by integrating a handwritten text recognition component within the proposed architecture.

• The proposed model enhances different forms of documents, independently of the text language.

• We achieve state-of-the-art performance on the public H-DIBCO datasets.

License

This work is only allowed for academic research use. For commercial use, please contact the author.

Requirements

install the requirements.txt pip install requirements.txt

Insert distortions on text line images

python distort_image_khatt.py

Train the GAN-HTR system using handwritten texts images datasets

python GAN_AHTR.py

Document binarization

python eval_Dibco_2010.py

Ckeckpoints to test using the HDIBCO 2010 are available :

https://usaupload.com/5Guu/discriminator_weights.h5

https://usaupload.com/5GuG/generator_weights.h5

Train a handwriting text recognition system

python train_khatt_basic_distorted.py

Image

H03

Binarzed Image

b_predicted_3

Citation:

If this work was useful for you, please cite it as:

@article{KHAMEKHEMJEMNI2022108370, title = {Enhance to read better: A Multi-Task Adversarial Network for Handwritten Document Image Enhancement}, journal = {Pattern Recognition}, volume = {123}, pages = {108370}, year = {2022}, doi = {https://doi.org/10.1016/j.patcog.2021.108370}, author = {Sana {Khamekhem Jemni} and Mohamed Ali Souibgui and Yousri Kessentini and Alicia Fornés}, }

About

This is an implementation for the paper Enhance to Read Better: A Multi-Task Adversarial Network for Handwritten Document Image Enhancement designed to enhance the document quality before the recognition process. It could be used for document cleaning and binarization. The weights are available to test the enhancement.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages