Handwritten Digit Recognition with CNN:-

This project implements a Convolutional Neural Network (CNN) for handwritten digit classification using the Modified National Institute of Standards and Technology (MNIST) dataset. It is built with PyTorch framework and includes features like training, saving/loading model checkpoints, and visualizing training metrics.

Here are some key features of the project:

1. CNN Architecture: A simple yet effective CNN for image classification.

2. Training & Checkpoints: Train the model on the MNIST dataset and save checkpoints for easy restoration.

3. Visualizations: Plot training loss and accuracy over epochs. During the training process, the model's loss and accuracy are plotted to help visualize the training progress, these graphs show how the model's performance improves over time.

4. Model Evaluation: Evaluate model performance on the test dataset.

Training process:-

The model is trained on the MNIST dataset, which contains 70,000 handwritten digits (0-9). In the first session, the model is trained for a set number of epochs, and the model and optimizer states are saved in a checkpoint file (checkpoint1.pt). In the second session, the checkpoint is reloaded, and training continues for additional epochs. After each training session, the model is evaluated on the test dataset to track its performance.

There are 2 python files in repository: model.py and train.py

model.py: this file defines the CNN model architecture and includes functions for training and testing the model. train.py: this file handles the entire training workflow, from loading the dataset to saving and loading checkpoints. This file uses the double_plot function to visualize training loss and accuracy.

Made by Sanad Nassar