Nepali Handwritten Character Recognization

Dataset: https://archive.ics.uci.edu/ml/machine-learning-databases/00389/

A Convolution Neural Network based approach for the classification of nepali characters. Alphabets from क-ज्ञ and digits from ०-९ can be recognized by the network with about 98.6% accuracy.

We used the following network architecture:

Summary of the model

Libraries used

• tensorflow
• numpy
• keras
• matplotlib
• opencv-python
• Flask
• Pillow

Steps to run the project

1. Clone the project
2. Install the required libraries using pip install -r requirements.txt
   (Mannually install above libraries in case of any error)
3. Go to webiste folder and run python app.py from the command line or terminal.
4. Go to address localhost:3000 from your favourite browser

You can also modify the model by editing Nepali Handwritten.ipnyb file using jupyter notebook.