Heh, people detection, gaze detection and people hotspots detection (i.e. where people mostly stand, say at an intersection) have been a thing for 15-20 years now, I consulted on such a project in the 2000s, using simple mobile cameras that were placed in a van, the van moved, and they'd make "maps" and sell that data to advertisers who chose where to place fixed/changeable ads banners. They eventually got that data from cell providers, and didn't need the cameras anymore.