/bHCR

Bangla Handwritten Character Recognition Using CNN

Primary LanguageJupyter NotebookApache License 2.0Apache-2.0

Bangla Handwritten Character Recognition Using CNN

This project is the implementation of the paper- S. M. Azizul Hakim and Asaduzzaman, "Handwritten Bangla Numeral and Basic Character Recognition Using Deep Convolutional Neural Network," 2019 International Conference on Electrical, Computer and Communication Engineering (ECCE), Cox’sBazar, Bangladesh, 2019, doi:10.1109/ECACE.2019.8679243

Built With

  • openCV for data preprocessing
  • keras on tensorflow backend
  • Implemented on Google Colab

Datasets

BanglaLekha-Isolated is used for train and validation. Some of the images were removed for their ambiguous nature. Prepared test dataset is used for cross-validation.

Architecture

arch

License

This project is licensed under the Apache License 2.0 - see the LICENSE.md file for details