/Sinhala-HWR

Deep learning for Sinhala HWR with web app integration

Primary LanguageJupyter Notebook

Practice Sinhala Alphabet (Using Convolutional neural networks)

by Kavinthe Perera

Note : This repository does not include the model file, the model needs to be trained before locally running application

Required python libraries

  • Flask
  • Tensorflow
  • Numpy
  • Bidict
  • CV2
  • Matplotlib
  • Sklearn

How to run

  1. Open terminal
  2. Change directory to model
  3. Run either train.py or train.ipynb (if notebook installed)
  4. Change back to root directory of the application
  5. Run command python server.py to start flask server

Updating model

  1. Navigate to data directory
  2. Run update script from command python update.py

Base Model Specs

Below values are generated at 30 epochs

  • Accuracy : 0.9447674
  • Loss : 0.2780703274359883