Watch · local
Watch a camera still
Drop a still and see what is there: a delivery, a package, a person or a car. Runs on this host, and the picture is not stored.
What Vision sees appears here
Drop a still from your camera. Vision names the things on your list, with a confidence for each, and says how long it took.
This is object detection, not face recognition. There is no template and no tracking identity.
Use this from the API
Cam calls POST /api/v1/watch with an owner token, an image and up to 20 classes; send previous_image for a scene change. The API returns detections with class, confidence and box, plus a receipt for the engine, time and cost.