Real-time feedback
MediaPipe runs pose detection in the browser and the UI updates posture state at roughly 10 frames per second.
Student-built local AI project
The app detects upper-body landmarks in the browser, scores posture with a tiny custom neural network, and gives real-time feedback without accounts, uploads, or server inference.
What it demonstrates
MediaPipe runs pose detection in the browser and the UI updates posture state at roughly 10 frames per second.
A small dense neural network scores seven geometric posture features instead of shipping frames to an API.
Scores are charted after monitoring ends, with a CSV export for reviewing posture trends or presenting results.
How it works
The browser requests webcam access and keeps the video stream local.
MediaPipe extracts nose, ear, and shoulder landmarks with WebGL acceleration.
Distances, ratios, and angles describe head position and shoulder alignment.
A hardcoded JavaScript model returns a 0-100 posture score and status color.
Privacy and constraints
Inference happens inside the page. The app does not require an account or upload webcam frames.
The model was trained around a typical desk camera angle. Low light, occlusions, and steep laptop angles can reduce accuracy.
This is an educational AI demo, not medical advice or a clinical posture assessment tool.