This Article 
 Bibliographic References 
 Add to: 
Improved Object Categorization and Detection Using Comparative Object Similarity
Oct. 2013 (vol. 35 no. 10)
pp. 2442-2453
Gang Wang, Adv. Digital Sci. Center, Nanyang Technol. Univ., Singapore, Singapore
D. Forsyth, Dept. of Comput. Sci., Univ. of Illinois at Urbana-Champaign, Urbana, IL, USA
D. Hoiem, Dept. of Comput. Sci., Univ. of Illinois at Urbana-Champaign, Urbana, IL, USA
Due to the intrinsic long-tailed distribution of objects in the real world, we are unlikely to be able to train an object recognizer/detector with many visual examples for each category. We have to share visual knowledge between object categories to enable learning with few or no training examples. In this paper, we show that local object similarity information--statements that pairs of categories are similar or dissimilar--is a very useful cue to tie different categories to each other for effective knowledge transfer. The key insight: Given a set of object categories which are similar and a set of categories which are dissimilar, a good object model should respond more strongly to examples from similar categories than to examples from dissimilar categories. To exploit this category-dependent similarity regularization, we develop a regularized kernel machine algorithm to train kernel classifiers for categories with few or no training examples. We also adapt the state-of-the-art object detector to encode object similarity constraints. Our experiments on hundreds of categories from the Labelme dataset show that our regularized kernel classifiers can make significant improvement on object categorization. We also evaluate the improved object detector on the PASCAL VOC 2007 benchmark dataset.
Index Terms:
support vector machines,image classification,knowledge management,learning (artificial intelligence),object detection,object recognizer,intrinsic long-tailed object distribution,visual knowledge,local object similarity information,knowledge transfer,similar object categories,dissimilar object categories,category-dependent similarity regularization,regularized kernel machine algorithm,kernel classifier training,object detector,object similarity constraints,Labelme dataset,object categorization,improved object detector,PASCAL VOC 2007 benchmark dataset,Training,Kernel,Detectors,Adaptation models,Support vector machines,Visualization,Object detection,sharing,Comparative object similarity,object categorization,object detection,kernel machines,SVM,deformable part model,PASCAL VOC
Gang Wang, D. Forsyth, D. Hoiem, "Improved Object Categorization and Detection Using Comparative Object Similarity," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 35, no. 10, pp. 2442-2453, Oct. 2013, doi:10.1109/TPAMI.2013.58
Usage of this product signifies your acceptance of the Terms of Use.