Object class recognition by unsupervised scale-invariant learning-英文文献.pdfVIP

Object Class Recognition by Unsupervised Scale-Invariant Learning R. Fergus P. Perona A. Zisserman Dept. of Engineering Science Dept. of Electrical Engineering University of Oxford California Institute of Technology Parks Road, Oxford MC 136-93, Pasadena OX 1 3PJ, U.K. CA 91125, U.S.A. fergus,az@robots.ox.ac.uk perona@ Abstract in the background of the object, scale normalization of the training sample) should be reduced to a minimum or elimi- We pr esent a method to learn and recognize object class nated. models fr om unlabeled and unsegmented cluttered scenes The problem of describing and recognizing categories, in a scale invariant manner. Objects are modeled as fle xi- as opposed to specific objects (e.g. [6, 9, 11]), has re- ble constellations of parts. A p robabilistic representation is cently gained some attention in the machine vision litera- usedfor all aspects of the object: shape, appearance, occlu- ture [1, 2, 3, 4, 13, 14, 19] with an emphasis on the de- sion and relative scale. An entropy-based featur e detector tection of faces [12, 15, 16]. There is broad agreement is used to select regions and their scale within the image. In on the issue of representation: object categories are rep- learning the par ameters of the scale-invariant object model resented as collection of features, or parts, each part has a are estimated. This is done using ex