LS-HDIB: A Large Scale Handwritten Document Image Binarization Dataset
Link to download initial dataset is here.
git clone https://github.com/kaustubh-sadekar/LS-HDIB.git
cd LS-HDIB/
pip install -r requirements.txt
python run.py cpu input_2.jpg unet_best_weights.pth
The output file will be saved as <INPUT_FILE_NAME>_output.jpg
. For this specific example it will be input_2_output.jpg
For better understanding of the input arguments type python run.py -h
Easy to use google colab notebook
NOTE:
We used segmentation_models_pytorch library for all the segmentation models. It has implementations for several segmentation models.