Emotion Detection
A live in-browser demo that detects faces, tracks 68 landmarks, and classifies expressions in real time — all inference runs locally via face-api.js.
A real-time face and emotion detection demo running entirely in the browser. The app pulls your webcam feed, detects faces using a tiny YOLO-style detector, then chains a 68-point landmark model and an expression classifier to overlay bounding boxes, facial landmarks, and per-emotion probability bars — all at ~10 fps on modern hardware.
Built with face-api.js, which wraps a small collection of pre-trained TensorFlow.js models. Nothing leaves your device: the video stream stays local and inference runs on your browser’s WebGL backend.
Originally built as a companion piece to the Object Detection and Pose Estimation demos — pushing on the same theme: how much useful ML can we squeeze into a static webpage with no server behind it? Turns out: quite a lot.