MNIST Digit Recognition

This project implements a Convolutional Neural Network (CNN) for recognizing handwritten digits using the MNIST dataset. The model can classify digits from 0 to 9 with high accuracy.

Features

• MNIST dataset loading and preprocessing
• CNN model implementation using TensorFlow
• Model training and evaluation
• Visualization of results and predictions
• Interactive digit recognition demo

Setup

1. Create a virtual environment (recommended):

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

2. Install dependencies:

pip install -r requirements.txt

3. Run the model:

python digit_recognition.py

Project Structure

• digit_recognition.py: Main script containing the CNN model implementation
• requirements.txt: Project dependencies
• README.md: Project documentation

Model Architecture

• Input layer: 28x28 pixels (grayscale images)
• Convolutional layers with ReLU activation
• MaxPooling layers
• Dense layers
• Output layer: 10 neurons (digits 0-9)

Performance

The model typically achieves:

• Training accuracy: >99%
• Test accuracy: >98%

Output Files

The script generates three visualization files:

1. confusion_matrix.png: Shows the model's prediction accuracy for each digit
2. training_history.png: Displays the training and validation accuracy/loss over time
3. sample_predictions.png: Shows random test images with their true and predicted labels

Dependencies

• numpy==1.24.3
• tensorflow==2.15.0
• matplotlib==3.7.2
• seaborn==0.12.2
• scikit-learn==1.3.0