The scenario
SignBridge uses hand-landmark detection and a lightweight classifier to recognise signs live. You will collect your own data, train the model, and build a simple app that assembles recognised signs into words.
System architecture
01 Webcam Live video
02 Hand landmarks MediaPipe
03 Classifier Sign → letter/word
04 Text output Sentence builder
High-level flow of SignBridge, from input to output.
What you will learn
- Landmark-based features
- Real-time inference loops
- Collecting your own dataset
What you will build
- Data collector
- Sign classifier
- Live translation app
Build plan
- Collect data
- Train classifier
- Real-time loop
- Sentence builder
- Deploy