Skip to main content


Detection and recognition of shapes

Trainee Achievements

Detection and recognition of shapes

The detection and recognition of shapes is a critical perceptual operation by humans or machines. Trainee Wilder compared how well contour and global shape-generating models accounted for the detection of complex shapes in noise. Contour models define shape complexity based on the turning angles. Skeleton (global) models define complexity based on the probability of the boundary conditioned on an estimated skeleton. Psychophysical results showed that although shapes could be detected from any small segment of the contour, without detecting the entire shape, shape detection nevertheless deteriorated as a function of skeletal complexity. This shows that detection is not entirely local, instead there is a global shape representation that constrains detectability.