Learning View Generalization Functions - Computer Science > Computer Vision and Pattern RecognitionReport as inadecuate




Learning View Generalization Functions - Computer Science > Computer Vision and Pattern Recognition - Download this document for free, or read online. Document in PDF available to download.

Abstract: Learning object models from views in 3D visual object recognition is usuallyformulated either as a function approximation problem of a function describingthe view-manifold of an object, or as that of learning a class-conditionaldensity. This paper describes an alternative framework for learning in visualobject recognition, that of learning the view-generalization function. Usingthe view-generalization function, an observer can perform Bayes-optimal 3Dobject recognition given one or more 2D training views directly, without theneed for a separate model acquisition step. The paper shows that viewgeneralization functions can be computationally practical by restating twowidely-used methods, the eigenspace and linear combination of views approaches,in a view generalization framework. The paper relates the approach to recentmethods for object recognition based on non-uniform blurring. The paperpresents results both on simulated 3D ``paperclip- objects and real-worldimages from the COIL-100 database showing that useful view-generalizationfunctions can be realistically be learned from a comparatively small number oftraining examples.



Author: Thomas M. Breuel

Source: https://arxiv.org/







Related documents