How francisocol/signar works
SignAR is a browser-based Filipino Sign Language recognition and visualization prototype. Its main workflow starts a camera, sends frames to MediaPipe Hands, recognizes gestures, and presents feedback while confirmed signs build a sentence that can be spoken. Separate capabilities include searchable sign learning, 3D practice, camera controls, and installable/offline PWA behavior. Recognition and sentence integration details are only partly shown in the sampled source; dashed edges mark that uncertainty.
Main entry points: sw.js Service Worker · script.js Gesture Recognition · manifest.webmanifest PWA Manifest · index.html App Assets
Components 16
Live Recognition
- Camera Capture
script.js - MediaPipe Hands
- Gesture Recognition
script.js - AR Feedback
script.js - Camera Controls
script.js
Learning and Output
- Sentence Builder
script.js - Sentence Translation
script.js - Browser Speech
script.js - Sign Learning
script.js - 3D Practice
script.js
PWA Support
- Service Worker
sw.js - Asset Cache
sw.js - PWA Manifest
manifest.webmanifest - App Assets
index.html
Other components
- User
- Speech Synthesis
Connections 16
- User → Camera Controls: starts or switches
script.js - Camera Controls → Camera Capture: starts capture
script.js - Camera Capture → MediaPipe Hands: sends frames
script.js - MediaPipe Hands → Gesture Recognition: provides landmarks
- Gesture Recognition → AR Feedback: shows recognition
script.js - Gesture Recognition → Sentence Builder: adds confirmed signs
script.js - Sentence Builder → Sentence Translation: formats output
script.js - User → Sentence Builder: edits sentence
script.js - User → Browser Speech: requests speech
script.js - Browser Speech → Speech Synthesis: speaks text
- User → Sign Learning: searches signs
script.js - User → 3D Practice: practices signs
- Service Worker → Asset Cache: stores assets
sw.js - Service Worker → Asset Cache: reads on failure
sw.js - Service Worker → App Assets: caches app files
sw.js - User → PWA Manifest: installs app