The smart camera uses object-tracking software and learns from its errors.
Zdenek Kalal, a PhD student at the University of Surrey in England, has created an object tracking algorithm called Predator. As he demonstrates in his video, the software can recognize and then track objects, even if they move further away, are viewed from a different angle, or disappear and reappear. Furthermore, it learns and improves the more it is used.
Kalal also shows how this can be used as a HCI interface, instructing the software to track his fingers and then moving them across the screen, drawing like a virtual mouse.
Predator has a number of useful applications, including face-recognition. Kalal demonstrates that the algorithm is able to track his movements on-camera and when he disappears and holds up a piece of paper with thumbnail photographs, it is able to recognize his face and pick him out from the crowd.