🔢 Digit Recognition Neural Network

A C++ application for handwritten digit recognition that leverages the NeuralNetwork library. This project demonstrates how to use a modern neural network implementation to recognize digits (0-9) from images with high accuracy.

✨ Key Features

• 📊 MNIST Dataset Integration: Ready-to-use solution for the standard benchmark in digit recognition
• 🎮 User-friendly Interface: Simple commands to train and test the neural network
• 📈 Performance Visualization: Monitoring of accuracy and loss during training
• 🖼️ Prediction Visualization: See the network's predictions on digit images
• ⚙️ Configuration Options: Customize network architecture and training parameters
• 💾 Model Saving/Loading: Preserve trained models for later use

🔗 Dependencies

• NeuralNetwork Library: The core neural network implementation providing:
  • Modern C++17 implementation
  • Flexible network architecture
  • Various activation functions
  • Efficient training algorithms
  • Model persistence capabilities

🏗️ Quick Start

Prerequisites

• 🛠️ C++17-compatible compiler (GCC 7+, Clang 5+, MSVC 2017+)
• 📋 CMake 3.28 or later
• 📚 NeuralNetwork library (included as a submodule)

🔨 Build Instructions

# Clone the repository with submodules
git clone --recurse-submodules https://github.com/dredstone1/digit-recognition-nn.git

# Navigate to project directory
cd digit-recognition-nn

# Create and enter build directory
mkdir build && cd build

# Configure and build
cmake ..
make

🏃‍♂️ Running the Program

# Run the digit recognition program with loading a pre-trained model
./digit_recognition -l

# Train a new model
./digit_recognition -t

# Continue training the model
./digit_recognition -l -t

📊 Performance

The application achieves approximately 90-93% accuracy on the MNIST test dataset after training, making it suitable for many practical applications requiring digit recognition.

🧩 How It Works

This application:

1. Loads the MNIST dataset into memory
2. Creates a neural network using the NeuralNetwork library with an architecture optimized for digit recognition
3. Trains the network on the training set (or loads a pre-trained model)
4. Evaluates the network on the test set to measure accuracy
5. Allows interactive testing with user-provided digit images

🛠️ Command Line Options

Flag	Description
-t	Train a new model
-l	Load a pre-trained model

🤝 Contributing

Contributions are welcome! Whether you want to improve accuracy, optimize performance, or add new features:

• 🐛 Open issues for bugs or feature requests
• 🔀 Submit pull requests with your improvements
• 💡 Share ideas and suggestions
• 📖 Help improve documentation

📄 License

This project is licensed under the MIT License. See the LICENSE file for complete details.

👨‍💻 Author

dredstone1 - GitHub Profile

---

⭐ If you find this project helpful for your digit recognition tasks, consider giving it a star!