For his thesis project, University of Surrey researched Zdenek Kalal has just made a visual tracking system straight out of the wildest dreams of sci-fi. Called Predator, his system offers an on-the-fly object recognition system, allowing a user to arbitrarily define objects within the camera's field of view. Once objects are defined, the system learns about them, recognizing them in more and more positions, at different scales and angles.
With remarkable speed, Kalal's software can recognize objects -- such as a human face -- when in profile, even if the system was told about the face while it was looking directly at the camera. The applications for this technology are quite broad, ranging from assisted-use applications, to comprehensive tracking of numerous people in a security setting, to many, many others.
It's hard to explain how futuristic this feels, so we invite you to check out Zdenek's video below.