● Research· Healthcare· Vision·2025Demo-ready
PT-Assist
Real-time AI physical therapy assistant using YOLO11 pose estimation for live form analysis and rep counting.
Overview
PT-Assist is a real-time physical therapy assistant that uses YOLO11 pose estimation to monitor exercise form, count repetitions, and provide live feedback. Built to help patients perform PT exercises correctly without a therapist physically present.
What it does
- Live pose tracking — extracts joint keypoints from webcam video in real time
- Joint-angle analysis — computes knee, hip, and elbow angles per frame against target ranges
- Rep counting with form gating — only counts reps performed within the correct angle envelope
- HUD feedback — overlays current rep count and an in-range / out-of-range form indicator
Stack
- Python · YOLO11 for pose estimation
- OpenCV for real-time video capture and overlay rendering
- NumPy for joint-angle vector math
Why it matters
Physical therapy adherence drops sharply when patients exercise at home unsupervised. PT-Assist demonstrates that a single laptop camera plus a modern pose model is enough to give patients the corrective feedback loop they lose between sessions — without specialized hardware.