Handwritten Digit Recognition using CNN

Project Overview

This project implements a Handwritten Digit Recognition system using a Convolutional Neural Network (CNN) trained on the MNIST dataset. A Streamlit web application allows users to upload handwritten digit images ( 0-9 ) and receive predictions in real time.

Features

• MNIST dataset preprocessing
• CNN-based digit classification
• Model accuracy evaluation
• Saved trained model
• Streamlit web interface
• Real-time digit prediction

Technologies Used

• Python
• TensorFlow / Keras
• NumPy
• Streamlit
• Pillow

Project Structure

handwritten-recognition-system/

├── model/

│ └── mnist_cnn.h5

├── app.py

├── train_model.py

├── requirements.txt

├── README.md

└── .gitignore

Model Performance

Test Accuracy: ~98-99%

Installation

pip install -r requirements.txt

Train Model

python train_model.py

Run Application

streamlit run app.py

Dataset

MNIST Handwritten Digits Dataset

Classes: 0–9

Author

Mounika Golusula