Rock Paper Scissors: Machine Learning Edition

Overview

This project uses a Rock Paper Scissors hand sign dataset and a custom-trained machine learning model to let you play real-time Rock Paper Scissors against your computer.
You can play either via the command line (rpscli.py) or a modern web interface.
The project is designed for the Nvidia Jetson Orin Nano running JetPack 6.x, but may work on other Linux systems with compatible dependencies.

• Backend: Python (Flask), custom API (rpsapi)
• Frontend: TailwindCSS, HTML, JavaScript
• Model: YOLO, trained on hand sign images

Setup

Important

This project is designed for Linux. Windows is not supported.

1. Clone the Repository

git clone https://github.com/milesmuehlbach/rockpaperscissors-ml.git
cd rockpaperscissors-ml

2. Install Dependencies

scripts/installdependencies.sh

Tip

If you encounter errors during installation, carefully read the error messages and install any missing system packages as instructed.

3. Train the Model

Warning

No pre-trained model is provided.
Training will take 30+ minutes and requires a CUDA-capable GPU (Jetson recommended).

scripts/trainmodel.sh

After the model is trained, you should be good to start using the app. Follow the instructions under Running to proceed.

Running

Command Line Interface (CLI)

(Assuming you're in the root directory of the project)

python3 ./rpscli.py

Web Application

(Assuming you're in the root directory of the project)

python3 -m webapp.app

The server should now be running on port 5000. If you want to access it from localhost, you can visit it at http://localhost:5000 in your browser.

Note

The development server is accessible over the network, but most browsers block webcam access for non-HTTPS or non-localhost sites.
For best results, run the webapp on localhost or configure HTTPS.
You may need to set special flags in your browser to allow webcam access on local network addresses.

Project Structure

rockpaperscissors-ml/
├── rpscli.py           # Command-line interface
├── webapp/             # Flask web application
│   ├── app.py
│   ├── templates/
│   └── static/
├── rpsapi/             # API and ML model logic
│   ├── fetchsign.py
│   └── rps.py
├── scripts/            # Setup and training scripts
│   ├── installdependencies.sh
│   ├── trainmodel.sh
│   └── runtailwinddev.sh
└── README.md

Troubleshooting

• Webcam not working?

  • Make sure you are using localhost or HTTPS.
  • Check browser permissions for camera access.
  • See browser console for errors.

• Model not found?

  • Ensure you have run scripts/trainmodel.sh and the model file exists in rpsapi/model/.

• Other issues?

  • Check the Issues section on GitHub or open a new issue with details.

Tip

Contributions, suggestions, and feedback are welcome!
Please open an issue or submit a pull request on GitHub.