Deriving information from images

A system, comprising:

• a camera configured to capture a depth map of a scene;
• and,
• a processor configured to:
• receive a semantically-labeled 3D voxel representation of a scene;
• identify planes in the semantically-labeled 3D voxel representation;
• filter planes which do not contribute to a manifold that represents the scene;
• extend unfiltered planes to identify intersections;
• and,
• infer incomplete areas of the manifold based at least in part from the intersections of the unfiltered planes.

A computer-readable storage medium storing instructions which, when executed by a processing device, cause the processing device to perform acts comprising:

• identifying planes in a semantically-labeled 3D voxel representation of a scene;
• inferring missing information by extending individual planes associated with structural elements of the scene;
• and,
• generating a watertight manifold representation of the scene at least in part from the inferred missing information.