Hand Digit Recognition using CNN

This repository contains a Convolutional Neural Network (CNN) implementation for handwritten digit recognition using TensorFlow and Keras.

Project Overview

This project implements a CNN model to recognize handwritten digits with high accuracy. Two different approaches are included:

1. MNIST.ipynb: Uses the standard MNIST dataset (60,000 training samples, 10,000 testing samples)
2. 47_99accuracy_pro_players.ipynb: Uses custom hand-drawn digits stored in the new_digits_2 folder

CNN Architecture

The model uses a multi-layer CNN architecture:

Input → Conv2D → ReLU → MaxPooling → Conv2D → ReLU → MaxPooling → Conv2D → ReLU → Flatten → Dense → ReLU → Dense → Softmax

Results

• The model trained on the MNIST dataset achieves high accuracy on standard test data
• The model trained on custom hand-drawn digits achieves 99.47% accuracy, demonstrating excellent generalization capabilities

Files Description

• MNIST.ipynb: Jupyter notebook implementing CNN with standard MNIST dataset
• 47_99accuracy_pro_players.ipynb: Jupyter notebook implementing CNN with custom hand-drawn digits
• new_digits_2/: Folder containing custom hand-drawn digit images
• Hand_Digit.pdf: Detailed documentation with project insights, methodology, and results analysis

Requirements

• TensorFlow 2.x
• NumPy
• Matplotlib
• scikit-learn

How to Run

1. Clone this repository
2. Install the required dependencies: pip install tensorflow numpy matplotlib opencv-python scikit-learn seaborn
3. Run the Jupyter notebooks

Further Information

For more detailed insights into the project methodology, implementation details, and results analysis, please refer to the Hand_Digit.pdf document included in this repository.