Learning by a generation approach to appearance-based object recognition

W e propose a methodology f o r the generation of learning samples an appearance-based object recognition. In m a n y practical si tuations, it i s no t easy to obtain a large number of learning samples. T h e proposed method learns object models f r o m a large number of generated samples derived from a small number of actually observed images. T h e learning algorithm has two steps: (1) generation of a large number of images by image interpolation, or image deformation, and (2) compression of the large sample sets using parametric eigenspace representation. W e compare our method with the previous methods that interpolate sample points in eigenspace, and show the performance o f our method to be superior. Experiments were conducted f O T 432 amage samples for 4 objects to demonstrate the effectiveness of the method.