Learning Appearance Based Models: Mixtures of Second Moment Experts

Chris Bregler and Jitendra Malik

We devloped a new technique for object recognition based on learning appearance models. The image is decomposed into local regions which are described by a new texture representation derived from the output of multiscale, multiorientation filter banks. We call this representation ``Generalized Second Moments'' as it can be viewed as a generalization of the windowed second moment matrix representation used by Garding & Lindeberg. Class-characteristic local texture features and their global composition is learned by a hierarchical mixture of experts architecture (HME by Jordan & Jacobs). The technique is applied to a vehicle database consisting of 5 general car categories (Sedan, Van with back-doors, Van without back-doors, old Sedan, and Volkswagen Bug). This is a difficult problem with considerable in-class variation. The new technique has a 6.5 % misclassification rate, compared to eigen-images which give 17.4 % misclassification rate, and nearest neighbors which give 15.7 % misclassification rate.


Chris Bregler (bregler@cs.berkeley.edu)