Learning spatial relations in object recognition Export

@article{pham06, abstract = {This paper studies two types of spatial relationships that can be learned from training examples for object recognition. The first one employs deformable relationships between object parts with a Gaussian model, while the second one describes pairwise relationships between pixel intensity values using Bayesian networks. We perform experiments on a human face dataset and a horse dataset, imposing the same amount of annotation of training data, which can be seen as sending knowledge to the learning algorithms. The result indicates that the Bayesian network method compares favorably to the deformable model, as it can capture long-distance stable relations in the object appearance. We also conclude that both methods are superior to strictly spatial matching by template and strictly non-spatial classifiers.}, address = {New York, NY, USA}, author = {Pham, Thang V. and Smeulders, Arnold W. M.}, citeulike-article-id = {827918}, citeulike-linkout-0 = {http://portal.acm.org/citation.cfm?id=1195784}, citeulike-linkout-1 = {http://dx.doi.org/10.1016/j.patrec.2006.03.016}, citeulike-linkout-2 = {http://linkinghub.elsevier.com/retrieve/pii/S0167865506001310}, citeulike-linkout-3 = {http://www.sciencedirect.com/science/article/B6V15-4KCHD02-1/2/36135aed7be7922bc8d43f4abe10f561}, day = {15}, doi = {10.1016/j.patrec.2006.03.016}, issn = {0167-8655}, journal = {Pattern Recogn. Lett.}, keywords = {by-parts, detection, object, onur2009}, month = {October}, number = {14}, pages = {1673--1684}, posted-at = {2009-06-27 00:00:13}, priority = {2}, publisher = {Elsevier Science Inc.}, title = {Learning spatial relations in object recognition}, url = {http://dx.doi.org/10.1016/j.patrec.2006.03.016}, volume = {27}, year = {2006} }

TY - JOUR ID - pham06 L3 - citeulike-article-id:827918 N2 - This paper studies two types of spatial relationships that can be learned from training examples for object recognition. The first one employs deformable relationships between object parts with a Gaussian model, while the second one describes pairwise relationships between pixel intensity values using Bayesian networks. We perform experiments on a human face dataset and a horse dataset, imposing the same amount of annotation of training data, which can be seen as sending knowledge to the learning algorithms. The result indicates that the Bayesian network method compares favorably to the deformable model, as it can capture long-distance stable relations in the object appearance. We also conclude that both methods are superior to strictly spatial matching by template and strictly non-spatial classifiers. IS - 14 JF - Pattern Recogn. Lett. SN - 0167-8655 AD - New York, NY, USA EP - 1684 TI - Learning spatial relations in object recognition PB - Elsevier Science Inc. VL - 27 SP - 1673 KW - by-parts KW - detection KW - object KW - onur2009 AU - Pham, Thang AU - Smeulders, Arnold PY - 2006/10/15/ UR - http://dx.doi.org/10.1016/j.patrec.2006.03.016 ER -