Exp. 01Computer vision2025
ProPosing
A pose detection project using computer vision techniques for bodybuilding.
What it does
- Evaluates four classic bodybuilding poses in real time (Double Biceps, Most Muscular, Side Chest and Side Triceps).
- A FastAPI backend runs MediaPipe on each camera frame and returns landmarks, a status and feedback.
- A Flutter interface for macOS and the web draws the skeleton overlay and the live evaluation.
- Geometric rules per pose (joint angles, elbow height), combined with scikit-learn models trained on collected data.
- Python
- MediaPipe
- OpenCV
- FastAPI
- Flutter
- scikit-learn
Adjust
- Right arm angletarget 30–80°60°
- Left arm angletarget 30–80°104°
- Right elbow at shoulder heightPose OK
- Left elbow at shoulder heightPose OK
Left arm angle out of range (30–80°, now 104°).
Drag the elbows and wrists, or focus a joint and use the arrow keys.
Logic from pose_evaluator.py
