Handwritten Digit Recognition

Brief Introduction

This project is an assignment for my undergraduate machine learning course. The project content includes:

Use PyTorch to build LeNet5 model;
Training LeNet5 model on MNIST dataset;
Make a handwritten digit recognition applet.

Demo

After running the main program, a whiteboard will appear. Press and hold the left mouse button to write on the whiteboard. After releasing the left mouse button, the program will automatically recognize the numbers on the whiteboard and display the recognition results below the whiteboard. Press the right button of the mouse to clear the whiteboard.

File Structure

Digits
├── docs              # Project Documents
    ├── demo.gif      # Demo
    └── README_cn.md  # Chinese Description Document
├── model             # LeNet Package
    └── __init__.py   # LeNet Model
├── install.sh        # Install program
├── LICENSE           # LICENSE
├── main.py           # Handwriting Digit Recognition Application
├── README.md         # English Description Document
├── requirements.txt  # List of Requirements
└── train.py          # Training Program

Quick Start

1.Clone

git clone https://github.com/Yue-0/Digits.git
cd ./Digits

2.Install requirements

sh install.sh

3.Train model

python train.py --lr 0.1 --epoch 10 --batch 64

Due to the simple network structure, this project uses CPU for training and inference.

4.Run the handwritten digit recognition application

python main.py