This is a tool for pixel wise segmentation. This is developed in order to do use cases like page extraction, textline , word and structure recognition of library documents.
Download the repository and after you prepared your data, just run:
python train.py with config_params.json