Detect a face of Donal Trump on images while browsing the Web and block it. Kind of like this:
While browsing all the faces are extracted and compared to icons/detailed/Donald_Trump_official_portrait.jpg, if a match is significant enough - icons/detailed/Carrot-128a.png is drawn on top of the face.
Using FaceAPI.js, which is based on TensorFlow.js.
Download (or clone) the repository. Click the Chrome menu icon and select Extensions from the Tools menu. Ensure that the Developer mode checkbox in the top right-hand corner is checked. Then click Load Unpacked and select the downloaded repository.
Ish. It does use a lot of computing power. A dedicated GPU would not hurt (on Nvidia GTC 965 it takes about 200ms to detect faces on 1 image). You probably don't want to run it day to day, but this is a good POC of using ML models from within an extension.