handtrack.js

Processes all data locally in the browser using TensorFlow.js, ensuring no server calls and enhancing privacy for sensitive or offline applications.

Can be added via a script tag or npm with a straightforward API for model loading and detection, enabling rapid prototyping without deep learning expertise.

Classifies hand gestures into categories such as open, closed, pinch, point, and face, allowing for diverse interaction models without custom training.

Offers quantized model variants (large, medium, small) with different precision levels, providing options to balance download size, though inference speed remains similar.