Suppose two moving robots surveying a landscape take pictures of two different scenes. Generically, we expect the pictures to be different, but it can happen that the images coincide. This unexpected phenomenon is an instance of ill-posedness in computer vision. To simplify the problem we ask the same question about robots moving in flatland: 1) when can two scenes have the same image, and 2) what are the cameras that can make this happen? In this talk we will see a complete answer to these questions using classical invariant theory of marked points on a projective line. No familiarity with computer vision is required.