I think I would much rather they secretly capture multiple frames and compute structure from motion or use flash-based segmentation to avoid the presented real

I think I would much rather they secretly capture multiple frames and compute structure from motion or use flash-based segmentation to avoid the presented real world failures. The given "wave their phone in the air for several seconds" dismissal is a pretty strong exaggeration. For one thing, my hands are moving anyway, because I, like everyone else, open the app before holding up the phone to take the shot. And seconds? What is this, 1995?