MNIST Handwritten Digit Classification using CNN

This project trains a Convolutional Neural Network (CNN) using the MNIST dataset to classify handwritten digits (0–9). The model is built and trained using TensorFlow & Keras.

Features

• Loads and preprocesses MNIST dataset
• Deep CNN model for digit recognition
• Dropout for regularization
• Achieves high accuracy with just a few epochs
• Simple and beginner friendly code structure
• Includes sample predictions on test images

Installation & Setup

Make sure you have Python 3.7+ installed.

# Clone the repository
git clone https://github.com/<your-username>/<repo>.git
cd <repo>

# Install dependencies
pip install tensorflow

Repository structure

Repo
├── digit_recog.ipynb      # Main Jupyter Notebook
├── README.md            # Project Documentation
└── CONTRIBUTING.md      # Contributing Guidelines 

Scope for contributions

📌 For contributing guidelines, refer to Contributing.md.

⭐️ Model Enhancements

• Add more CNN layers / different architecture (e.g., BatchNorm, Dropout tuning)
• Use different optimizers / learning rate schedulers
• Improve accuracy through data augmentation
• Add early stopping + checkpoint saving

⭐️ Code Improvements

• Modularize the script (split into train.py, model.py, utils.py)
• Add configuration support (YAML/JSON)
• Add argument parser (argparse) for flexibility
• Add documentation and inline comments

⭐️ Testing & Validation

• Write unit tests for data loading/model building
• Add validation metrics like confusion matrix, F1 score