MNIST Digit Recognition Model

Project Overview

This repository hosts a TensorFlow-based neural network model designed to classify handwritten digits from the MNIST dataset. The project demonstrates the application of deep learning techniques to recognize numerical digits, providing a foundation for further exploration into machine learning and image processing.

Features

• Data Visualization: Initial display of MNIST dataset images.
• Preprocessing: Normalization of images to prepare data for training.
• Neural Network Architecture: A sequential model with multiple dense layers.
• Training and Evaluation: Model training and accuracy evaluation on the MNIST test set.

Technologies Used

• Python
• TensorFlow
• Keras
• Matplotlib

Getting Started

To run this project, follow these steps:

1. Clone the repository:

2. Install required libraries:

3. Run the notebook: Navigate to the notebook directory and launch Jupyter Notebook:

Model Architecture

The model consists of:

• Input layer: Flatten the 28x28 image data.
• Hidden layers: Two layers with 128 nodes each, using ReLU activation.
• Output layer: A softmax layer with 10 nodes corresponding to the digit classes.

Results

• After training for 3 epochs, the model achieves an accuracy of approximately 97% on the test set.

Authors

• Youssef

License

This project is licensed under the MIT License - see the LICENSE file for details.