More Ways to Help Computers See

Last month, we talked about an effort at MIT to help computers understand more of what their electronic eyes can see. The Carnegie Mellon School of Computer Science is engaged in a similar project, which it calls Geometrically Coherent Image Interpretation.

Using this technology, a computer can take an photographic image and construct a 3D model of what it sees, based on depth perception and the ability to identify types of objects. In this way, the computer can gauge relative distances of the objects it sees -- critical for such tasks as robotic navigation -- and can distinguish classes of objects.

Source: Smart Mobs