A mobile app that lets you play Rock–Paper–Scissors using your front camera. The app recognizes your hand gesture (rock, paper, or scissors) in real time and plays against a computer opponent.
- Expo (React Native)
- Expo Camera
- TensorFlow.js + @tensorflow/tfjs-react-native
- @tensorflow-models/handpose (lightweight hand keypoint model)
- react-native-svg for landmark overlay
- Real-time gesture detection via front camera
- Automatic AI opponent selection
- Instant round resolution and running score (Wins/Draws/Losses)
- Simple, clean UI
- Node.js LTS
- Expo CLI (optional) – you can use
npx expodirectly - iOS: Xcode + iOS Simulator, or Expo Go on device
- Android: Android Studio + Android emulator, or Expo Go on device
npm install
If you cloned the repo and are opening for the first time, also install iOS/Android build deps as needed by Expo.
npx expo start
- Press i to run on iOS simulator (macOS only)
- Press a to run on Android emulator
- Or scan the QR with Expo Go on your device
On first run, TensorFlow RN backend will initialize which can take a moment.
- The app uses
@tensorflow/tfjs-react-nativeto run TF.js on device with the RN WebGL backend. - The front camera stream is wrapped with
cameraWithTensors, which yields tensors in a render loop sized to the model input. @tensorflow-models/handposeestimates 3D hand landmarks and finger annotations.- A simple rules-based classifier infers the gesture:
- Rock: no extended fingers
- Paper: index+middle+ring+pinky extended
- Scissors: index+middle extended only
- A gesture must remain stable for ~600ms before a round is locked in, then the AI randomly picks its move and the result is shown briefly. Scores update automatically.
- Ensure good lighting and hold your hand within the camera frame
- Paper requires all four non-thumb fingers extended; thumb state is ignored for robustness
- Performance varies by device
MIT