DaKanjiRecognizer-ML

The machine learning parts of the DaKanjiRecognizer applications.

Setup development environment

install all dependencies:

python -m pip install wheel

python -m pip install -r requirements.txt

Because the database used for this application is needed please follow the setup guide for the ETL_data_reader package.

Kanji recognition CNN

This CNN tries to recognize Kanji characters. A list of all supported characters can be found here and an desktop application using it here..
In the CNN_kanji folder you can find the python files necessary to train this model on your own. When everything is setup correctly training.py can be run to train the network.
For a more interactive approach look at the jupyter notebook. In the releases section pretrained weights can be found.

Credits

CNN for kanji detection

The data on which the neural network was trained on was kindly provided by ETL Character Database

Papers:

• Recognizing Handwritten Japanese Characters Using Deep Convolutional Neural Networks
  
• A neural framework for online recognition of handwritten Kanji characters
  
• Online Handwritten Kanji Recognition Based on Inter-stroke Grammar
  
  

other

pages: button design