← All demos
Human vs machine · project 14

Can you beat the AI?

Depth models read a single photo very well on ordinary surfaces, then fail on mirrors, glass and painted illusions. Humans fail on some of those too, but not always on the same ones.

Two points, A and B. Which one is closer to the camera? 12 images: ordinary rooms, but also mirrors, glass, flat pictures and illusions, where depth models are known to fail.

Mode

Image credits