Nothing in the post or the videos indicate any facial recognition or any Machine Learning application. Why would he put the simple object detection overlays in the video at all as if that were relevant for anything other than creating the illusion of AI. I imagine he looked up some popular spots with public camera feeds, found posts with those coordinates and then manually matched the video from the image timestamp. The most impressive thing here is the artist realizing the idea of this would be provocative and garner attention.