Researchers in computer vision have been developing techniques for recovering the three-dimensional shape and appearance of objects in imagery. We now have reliable techniques for accurately computing a partial 3D model of an environment from thousands of partially overlapping photographs. There are two major themes in the computer vision literature: 3D geometry and recognition. The first theme is about using vision as a source of metric 3D information: given one or more images of a scene taken by a camera with known or unknown parameters, how can we go from 2D to 3D, and how much can we tell about the 3D structure of the environment pictured in those images? The second theme is all about vision as a source of semantic information: can we recognize the objects, people, or activities pictured in the images, and understand the structure and relationships of different scene components just as a human would?