Three dimensional representation and reasoning for indoor scene understanding

摘要: When addressing the problem of scene understanding from a single image, we want our system to understand not only where objects are in but also they 3D world. Segmenting and labeling regions 2D image plane does achieve this goal. We need representation that inherently encodes properties scene. In addition location 3D, make use physical knowledge about valid configurations world by rejecting violate constraints, such as two occupying same volume. geometric can aid detecting identifying certain classess well characterized their geometry. thesis, will demonstrate benefits using for indoor understanding. show models provides natural way represent inject have perform reasoning.