This custom Raspberry Pi project lets you play rock, paper, scissors versus AI
Table of Contents
If you’re looking for another Raspberry Pi 5 project, here’s a simple one based around one of the most universally known games: rock, paper, scissors (or whatever you may call it where you’re from). Combining a camera, display, the Pi, and some coding, user Citrus has created an AI-driven rock-paper-scissors game on the small computer.
Previously, such projects would require powerful graphics cards to run – just look at the massive GPUs AI machines utilize these days. However, thanks to the AI module onboard, it’s actually possible to do this all onboard a Raspberry Pi.
In the project description, the creators discuss how this was created and made possible. They claim the experiment results have been excellent, with the gesture classification accuracy and recall consistently above 95%. Since they showcase the library of images they use, it’s a good mix of hands, shapes, and skin tones to make it as broadly working as possible.
We developed a highly optimised AI model using BrainBuilder software, which was trained using a dataset of real captured images and synthetic augmented data generated through advanced 3D modelling techniques.
Citrus on Hackster.io
Raspberry Pi 5 AI rock paper scissors game draw, Source Citrus, Hackster io
Raspberry Pi 5 AI rock paper scissors game win, Source Citrus, Hackster io
Raspberry Pi 5 AI rock paper scissors game loss, Source Citrus, Hackster io
What you need for this project, and why you might make it
The rock, paper, scissors game is just meant to be a showcase of what the device can do. Gesture recognition can go beyond the game to something like even translating sign language or other projects you might want to create. With the creators saying they will work on expanding the data set and optimizing the neural network, it will look to expand its applications and commercial uses.
Below is a list of components and tools you need to complete the project. With only a couple of items ot buy, and code available to download, the sticking point might be the 3D printer requirement. Well, unless you send the STL file to a printer to do it for you, I’m sure there are other workarounds to position the camera in the right place for now.
- Raspberry Pi 5
- Raspberry Pi Touch Display
- Raspberry Pi AI Camera
- Raspberry Pi Raspbian
- Sony Brain Builder
- 3D printer
Video of the game in action
About the Author