Iconic Pose Detector
A real-time pose detection web app built with Flask, MediaPipe, and a custom trained ML model. Point your webcam at yourself, hold a pose for 1.5 seconds, and the app identifies it live in the browser.
About the Project
Iconic Pose Detector is a computer vision project that turns your webcam into a real-time body language classifier. Built around the idea that pose recognition should feel instant and interactive, it combines MediaPipe's landmark detection with a custom scikit-learn model and streams predictions directly to the browser — no page reloads, no lag.
Built for CV enthusiasts, ML learners, and anyone curious about human pose estimation, it provides a complete end-to-end pipeline that feels responsive and polished — from raw webcam feed to confirmed pose detection, all running locally in your browser.
Real-Time Detection, Built to Feel Alive
The core idea was to make pose detection feel like a live experience rather than a demo. Every technical decision — MJPEG streaming for the camera feed, Server-Sent Events for predictions, a 1.5-second hold timer to eliminate false positives — was made to keep the interaction tight and reliable. It even has a Target Mode where you're given a random pose to match within 10 seconds, with score and streak tracking, turning a CV project into something genuinely fun to use.
