Visual illusions and learning algorithms

How well do learning algorithms perform on images with visual illusions?