MNIST Neural Network Web Application

An end-to-end machine learning project that trains a neural network on the MNIST handwritten digit dataset and serves predictions through a full-stack web interface.

Overview

This project demonstrates the complete workflow of a simple deep learning system:

• data loading and preprocessing
• neural network training
• model saving and inference
• backend API integration
• frontend user interface for digit drawing and prediction

Features

• Train a neural network on the MNIST dataset
• Save and load the trained model
• Predict handwritten digits from a web interface
• Backend built with Python
• Frontend built with HTML, CSS, and JavaScript
• Clean project structure for experimentation and extension

Project Structure

mnist-neural-network/
├── backend/
│   ├── main.py
│   ├── model.py
│   └── requirements.txt
├── data/
├── static/
│   ├── app.js
│   ├── index.html
│   └── style.css
├── models/
│   └── mnist_model.pth
├── run.sh
├── .gitignore
└── README.md

## License

This project is licensed under the MIT License.