Back to Portfolio
Open Source Computer Vision

APT Fitness Assistant

Real-time pose estimation with MediaPipe for form correction, rep counting, and personalized workout recommendations — your AI gym partner, running locally from your webcam.

What is it?

Most people lift alone without feedback on their form — and bad form leads to injury. APT (Automated Personal Trainer) uses your webcam and MediaPipe's pose landmark detection to give you real-time, frame-by-frame feedback on movement quality.

It counts reps by tracking joint angles through the full range of motion, flags form breaks with visual overlays, and suggests adjustments. Everything runs locally — no cloud, no subscription, no data leaving your device. Built as a Streamlit app so setup is a single pip install.


What it does

🏋️

Pose Estimation

MediaPipe detects 33 body landmarks per frame in real time — shoulders, elbows, hips, knees, and more.

Form Correction

Compares your joint angles against reference ranges and flags deviations with on-screen cues.

🔢

Rep Counting

Tracks range-of-motion completion per exercise to count reps accurately without manual input.

💡

Workout Recommendations

Suggests follow-up exercises based on session data — progressive overload, muscle balance, recovery.

🖥️

Streamlit UI

Clean browser interface with live video feed, landmark overlay, and stats panel — no frontend code needed.

🔒

Fully Local

All processing happens on your machine. No video data sent anywhere — complete privacy by design.


Built with

Python MediaPipe OpenCV Streamlit NumPy

See the code

Full source and setup instructions on GitHub.