Stack - noam-hoze/sfinx GitHub Wiki
MoveNet + BlazeFace + FaceMesh (via TensorFlow Lite) Platform: Android, iOS via TFLite
Tracks: Full-body pose (MoveNet), facial landmarks (BlazeFace, FaceMesh)
Depth?: Estimations possible using Z-coordinates of keypoints
SDK: Use with TensorFlow Lite – portable and customizable
Open Source: ✅
✅ Great for low-latency performance on mobile using Google’s lightweight models
See full discussion Here