Hand Landmark Detection With Gesture Recognition

Google's cross-platform on-device ML framework with pre-built solutions.

Unique: Provides 21-point hand skeleton with built-in multi-hand tracking and left/right hand classification in a single unified API, using a two-stage detection-then-landmark approach optimized for mobile devices; includes gesture recognition foundation (raw keypoints) without requiring separate gesture classification models.

vs others: More accurate and faster than OpenPose for hand tracking on mobile devices, and includes native multi-hand support unlike some single-hand-focused alternatives, but requires post-processing for actual gesture classification unlike specialized gesture recognition systems.