This demo uses a TensorFlow.js pre-trained COCO-SSD model. COCO-SSD stands for 'Common Objects in COntext - Single Shot multibox Detection'. More pre-trained models are available in: TensorFlow.js models
Please note that this may not work with Chrome. Try Edge instead.
It will take a short while (some 15 seconds for first use, 3 seconds after that) for the model to load, after which 'Enable Webcam' button becomes active and you can click it. The browser will then request for access to you camera - you need to give the permission for this demo to work.
The COCO-SSD model will then try to identify some common objects in your webcam's view. A box is drawn around a shape when identification confidence is high enough - for demo purposes, I've set the threshold quite low.
There are 90 different object classes identified, a list is provided here
This demo is based on the tutorial by Jason Mayes found here: Make a smart webcam in JavaScript with a TensorFlow.js pre-trained Machine Learning model