On-device computer vision

Live detection

Point a camera at your space — or upload footage — and EyeAI segments every object and piece of equipment in view, tagging each with a mask and label. This runs entirely in your browser: the video never leaves your device.

Ready. Choose a source to begin.

Start the camera or upload a clip. The detection model (~15 MB) loads on first use, then runs on-device.

How this works. The feed is segmented on-device by a compact vision model (SegFormer, ADE20K classes) via WebAssembly — no server, no upload. Detected classes that are kitchen/retail equipment (stove, oven, sink, refrigerator, microwave, counter, shelving, table…) are highlighted; other objects are shown faintly. This is a browser demo of the perception layer; a production EyeAI deployment runs a model tuned to your exact equipment on the in-store edge box for full frame-rate and accuracy.