For a long time, computer scientist didn't believe it was possible to generate 3-D models from a 2-D image source. Some very bright people out of Carnegie Mellon University have been making some very impressive progress in the area. By identifying objects and regions image (such as the sky, the ground, cars or people) as reference points, their < a href="http://www.cs.cmu.edu/~dhoiem/projects/popup/index.html">software is able to extrapolate a 3-D model. Their results speak for themselves.

Via Techeblog and HipTechBlog