Tensorflow.js and pretrained model are run back-end by the video comes from WebRTC. It detects objects and mask them. Tensorflow.js and Coco-SSD it recognises up to 90 objects When you run html fle, it uploads pretrained model to your computer, mobile phone or tablet.After it finishes uploading pretrained model at background, Camera button activates itself.When you press open your camera button, your camera will be opened via your permission.And then it will start detection of objects. You can see accuracy and object name on green frame