MNIST Digit Classifier with GUI

A Python application for training, saving, loading, and using a neural network to classify handwritten digits from the MNIST dataset. The application features a Tkinter-based GUI for easy interaction.

Features

• Train neural networks with customizable hyperparameters
• Save and load trained models
• Interactive drawing canvas for digit classification
• Real-time probability distribution display
• User-friendly GUI interface

Project Structure

my_mnist_app/
│
├── main.py                # Application entry point
├── model.py              # PyTorch model definitions
├── train_utils.py        # Training utilities
├── gui/
│   ├── train_gui.py      # Training interface
│   └── draw_gui.py       # Drawing and classification interface
│
├── README.md
└── requirements.txt

Installation

Prerequisites

• Python 3.7+
• pip package manager

Setup

1. Clone the repository:

git clone https://github.com/yourusername/my_mnist_app.git
cd my_mnist_app

2. Create and activate virtual environment:

python -m venv venv

# Windows
venv\Scripts\activate

# macOS/Linux
source venv/bin/activate

3. Install dependencies:

pip install -r requirements.txt

Usage

Running the Application

python main.py

Training a Model

1. Set hyperparameters:
   • Learning rate
   • Batch size
   • Number of epochs
2. Click "Train Model"
3. Monitor training progress
4. Save model when complete

Using the Drawing Interface

1. Click "Draw & Classify"
2. Draw a digit in the canvas
3. Click "Classify" to get predictions
4. Use "Clear" to reset the canvas

Troubleshooting

SSL Certificate Errors

If encountering SSL errors during MNIST download:

pip install --upgrade certifi

Poor Classification Results

• Ensure digits are centered in the canvas
• Try increasing training epochs
• Adjust learning rate and batch size
• Verify image preprocessing

Tips for Best Results

• Train for multiple epochs to improve accuracy
• Use appropriate learning rates (e.g., 0.001)
• Draw digits clearly and centered
• Save models regularly to avoid retraining

License

MIT License

Acknowledgements

• PyTorch
• Tkinter
• MNIST Dataset

---

For more information or to report issues, please visit the project repository.