Research
TensorFlow Gesture Controls
A TensorFlow.js version of the gesture photo-stack
experiment.
Same interaction model, different runtime.
What This Is
This project mirrors the existing gesture photo stack, but swaps the hand-tracking engine to TensorFlow.js so the runtime can be studied and tuned separately.
What TensorFlow Changes
- Model choice can be tuned for speed or accuracy
- Gesture thresholds and smoothing can be customized in code
- Custom classifiers can be added later without changing the interface
What Can Be Customized Later
- Pinch sensitivity, hold timing, and release behavior
- Lite vs full hand model selection
- Future tap, hold, or symbolic gesture classifiers