Skip to main content


Model of shape classification

Research Achievements

Model of shape classification

Trainee Wilder used his training in human perception and computational models of perception to develop and test a model of shape classification based upon the statistics of natural shapes and focusing on the generation of shape skeletons from contour information. The model was used to develop and train a Bayesian classifier to distinguish animal and leaf shapes. Human subjects' performance was well predicted by the Bayesian classifier. This suggests that the shape skeleton contains a sufficient amount of information to allow for accurate classification, and may capture some of the mechanisms that allow humans to find objects in natural images, as well as improve the algorithms used by automated shape recognition systems. (Wilder, Feldman, & Singh, 2011, Cognition, 119, 325-340.)