Handwritten Digit Recognizer

This Git repository contains a simple handwritten digit recognition application built using Python and TensorFlow. It allows you to draw a digit on the canvas, and it will predict the digit using a neural network model trained on a dataset of handwritten digits.

Requirements

Before running the application, make sure you have the following dependencies installed:

• image (version 1.5.20)
• numpy (version 1.24.1)
• tensorflow
• tensorflow-datasets
• matplotlib

You can install these dependencies using pip with the following command:

pip install -r requirements.txt

How to Run

To run the Handwritten Digit Recognizer application, execute the following command:

python app.py

Usage

1. Launch the application using python app.py.
2. A canvas will appear where you can draw a digit using your mouse.
3. Click the "Predict" button to make a prediction of the drawn digit.
4. The predicted digit will be displayed in the console.

You can also click the "Erase All" button to clear the canvas and start over.

Code Structure

• app.py: This file contains the main application code, including the GUI using Tkinter, drawing on the canvas, and making predictions using the neural network model.

• model.py: This file defines functions for loading data, preprocessing images, training the neural network model, making predictions, and displaying errors.

• data : This folder contains X.npy and y.npy which are the datasets used by the Model.

Neural Network Model

The neural network model used for digit recognition consists of three layers:

1. Input Layer: 400 units
2. Hidden Layer 1: 25 units with ReLU activation
3. Hidden Layer 2: 15 units with ReLU activation
4. Output Layer: 10 units with linear activation

The model is trained on a dataset of handwritten digits and compiled with the Adam optimizer and sparse categorical cross-entropy loss.

Displaying Errors

The application can also display errors by comparing the predicted digits to the actual labels for a subset of the dataset. The errors are displayed in a grid, showing the misclassified digits and their predicted labels.

Acknowledgments

This application was created as a learning project and uses the MNIST dataset for training the neural network model.