Fig. 1.

Perception beyond the visible. On the left is an image of a street scene. While some parts of the scene are occluded, such as the building behind the tree highlighted by the red box, humans have no trouble reasoning about the shape and appearance of such hidden parts. In this work we go beyond 2.5D shape representations and learn to predict layered scene representations from single images that capture more complete scenes, including hidden objects. On the right, we show our method’s predicted 2-layer texture and shape for the highlighted area: (a, b) show the predicted textures for the foreground and background layers respectively, and (c, d) show the corresponding predicted inverse depth. Note how both predict structures behind the tree, such as the continuation of the building. (Color figure online)