MNIST Digit Classification using PyTorch

This project implements a handwritten digit classifier using a fully connected neural network built with PyTorch.
The model is trained and evaluated on the MNIST dataset, which contains grayscale images of handwritten digits (0–9).

---

Project Overview

• Task: Handwritten digit classification
• Dataset: MNIST
• Framework: PyTorch
• Model: Fully Connected Neural Network
• Output: Digit class (0–9)

---

Model Architecture

Input (28 × 28 = 784)
→ Linear (128) + ReLU
→ Linear (64) + ReLU
→ Linear (10)

---

Dataset

The MNIST dataset contains:

• 60,000 training images
• 10,000 test images
• Image size: 28 × 28 (grayscale)

The dataset is automatically downloaded using torchvision.

---

Requirements

Install the required libraries using:

pip install torch torchvision matplotlib

---

Training Details

• Loss Function: CrossEntropyLoss
• Optimizer: Adam
• Learning Rate: 0.001
• Epochs: 5

---

Results

The model achieves approximately 97% accuracy on the MNIST test dataset after training.

A sample image is displayed along with:

• Actual label
• Predicted label

---

Model Saving

After training, the model is saved as:

mnist_model.pth

---

Future Improvements

• Use a Convolutional Neural Network (CNN)
• Add validation accuracy plots
• Support custom image input

---