MoveLens

Movement-analysis demo — Sam's uploads, analysed locally
Pose model: MediaPipe Pose Landmarker (Heavy) 33 landmarks/frame · 24 fps Standards: ISO 11228-2 (pulling) · NIOSH (lifting) Processed: on this machine, offline
What you're looking at. Each video below was uploaded by Sam (an Australian physiotherapist beta-tester) through the MoveLens app, then processed through Google's MediaPipe Pose Landmarker running locally to extract 33 body landmarks per frame. The annotated MP4 shows the skeleton overlay, current trunk-inclination angle, and risk zone (Green/Amber/Red against ISO 11228-2 thresholds) in real time. Below each video are the deterministic clinical metrics computed in Python from the raw landmark stream — same input always produces the same numbers — plus four annotated keyframes pinned to the moments a clinician would want to inspect.