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