Gemma 3n Webcam Analysis Application
A modern web application featuring real-time webcam analysis powered by Google's Gemma 3n AI model. The frontend provides an intuitive interface for webcam capture, AI-powered image analysis, and persona-based interactions, with a backend API for model management.
🚀 Features
- Interactive Frontend: Modern React-based interface with real-time webcam capture
- AI-Powered Analysis: Image analysis using Google Gemma 3n model
- Persona System: Customizable AI personalities for different analysis scenarios
- Webcam Integration: Live camera feed with instant capture and analysis
📋 Development requirements
- Python 3.11+
- Node.js 22+ (for frontend development)
- CUDA-compatible GPU (optional, but recommended)
- Webcam/Camera device
- 64GB+ RAM
- WSL (Windows Subsystem for Linux) for frontend builds
This was my submission to the Gemma 3n Kaggle contest which I previously blogged about!
https://github.com/grctest/g3n-fastapi-webcam-docker
https://hub.docker.com/r/grctest/gemma3n_webcam_app
It enables Gemma 3n 2b-it AI analysis of webcam footage using either your CPU or GPU, either manually or regularly on a scheduled basis!
You can provide your own custom system and user prompts to experiment with Gemma3n's AI capabilities, as long as you have sufficient RAM and optionally a compatible GPU!
There were around 609 entries to the Kaggle contest, and I'm only eligible for 4 potential prizes so the odds of me winning are slim, but it was fun to throw together this web app, refreshing my skills in Python and Docker as well as continuing to use React and Astro with AI assisted development, so it's not the end of the world if I don't win, lol.
If you've got a compatible computer you should try it out! It's fully open source so feel free to fork the project and built upon it if you like the functionality and can think of an use case!
Cheers :)