Chargrid Pytorch

This is the pytorch implementation of Chargrid paper Arxive Link.

Pre-processing has been taken from Antoine Delplace's repository.
Also, for tensorflow2.0 implementation check out his repository Link Here

See this for model description

Installation

After cloning this repository

Use the package manager pip to install the requirements.

pip install -r requirements.txt

Copy env.sample to .env Then add folder paths for the dataset in .env file

Usage

Run following to preprocess the dataset

python -m data_pipeline.preprocessing
python -m data_pipeline.preprocessing_bis
python -m data_pipeline.preprocessing_ter

Once you have data preprocessed run train.py.

python train.py

Or for resumable training run

python resumable.py

Dataset

This implementation uses The ICDAR 2019 dataset.
For more information and download sources check Niansong Zhang's work

Code Style

We use black code formatter. Before committing please run following in your terminal from the project root directory.

black ./*/*.py

License

GPLV3