Depth in the Wild

Depth in the Wild is a dataset for single-image depth perception in the wild, i.e., recovering depth from a single image taken in unconstrained settings. It consists of images in the wild annotated with relative depth between pairs of random points.