Form Tracker is a real-time web-based feedback tool that helps users track and improve their exercise form using their webcam. It uses MediaPipe Pose for pose estimation to count reps and provide simple feedback on movement quality.
💡 This is an early-stage prototype focused on client-side pose tracking and feedback delivery.
- ✅ Track basic body movements in real time using a webcam
- ✅ Count reps based on pose and joint angle changes
- ✅ Display feedback to encourage proper form
- ⚙️ Build toward a full AI-assisted fitness form corrector
-
Webcam Access The browser captures video using native WebRTC APIs.
-
Pose Detection (MediaPipe Pose) The video feed is processed through MediaPipe Pose, which identifies 33 key body landmarks in real time.
-
Movement Analysis Custom JavaScript logic analyzes joint angles and position deltas to:
- Detect motion patterns
- Count exercise reps
- Provide live feedback (e.g., “Go lower”, “Keep your arms straight”)
-
UI Overlay Feedback and rep count are shown on a simple UI layered on top of the video.
- Frontend: HTML, CSS, JavaScript
- Pose Detection: MediaPipe Pose
- Rendering: HTML5
<canvas>for overlay + pose drawing - Deployment: Local testing (live hosting planned)
- Webcam + video stream capture
- Real-time pose detection with MediaPipe
- Exercise rep counter
- Dynamic feedback text (e.g. posture correction)
- Exercise selection (e.g. push-ups, squats)
- Form grading / scoring
- User session export or tracking
This project is currently client-side only. No backend or build system needed.
- Clone the repository:
- Open
index.htmlin a browser (preferably Chrome). - Allow camera access when prompted.
- Click “Start” and begin your exercise.
- Add multiple exercise types with toggle or detection
- Improve pose-based accuracy for different body types
- Add audio or visual form correction feedback
- Save user session stats (e.g. total reps, form quality)
- Support mobile / tablet cameras