The classical approach to artificial intelligence in vision, has faced many fundamental problems. Progress was made in very specific and strictly-controlled application domains only. On the other hand, during recent years a tremendous amount of new biological evidence has been brought about by visual neuroscience which could be used for developing artificial vision systems.
This project proposes a new 3D object recognition system that will reconcile state-of-the-art computer vision techniques with neurobiological and psychophysical evidence of visual perception. Active vision, i.e. including intentional observer motion in the visual process, will be implemented based on locally-autonomous and self-organising brain-style neural networks.
The aim of this project is to realise a working system which can recognize many (i.e. more than 100) types of natural or everyday 3D-objects, in a way which is invariant to changes in the presentation of the objects or to the presence of structural noise in the input.