Tuning SVM-Based Relevance Feedback for the Interactive Classification of Images

In the context of SVM-based interactive image classificatio n with relevance feedback, we put forward a new active learning selection cri terion that minimizes redundancy between the candidate images shown to the user at every round. We also show that insensitivity to the scale of the dat a is an important quality of the learning machine and we propose the use of spec ific kernel functions to achieve this. Experiments performed on severa l image databases confirms the atractiveness of our suggestions.

[1]  Bernhard Schölkopf,et al.  The Kernel Trick for Distances , 2000, NIPS.

[2]  Daphne Koller,et al.  Support Vector Machine Active Learning with Application sto Text Classification , 2000, ICML.

[3]  F. Fleuret,et al.  Scale-Invariance of Support Vector Machines based on the Triangular Kernel , 2001 .

[4]  Lei Wang,et al.  Bootstrapping SVM active learning by incorporating unlabelled images for image retrieval , 2003, 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings..

[5]  G. Medioni,et al.  Content-based image retrieval: an overview , 2004 .

[6]  Daphne Koller,et al.  Support Vector Machine Active Learning with Applications to Text Classification , 2000, J. Mach. Learn. Res..

[7]  C. A. Murthy,et al.  A probabilistic active support vector learning algorithm , 2004, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[8]  Thomas S. Huang,et al.  Relevance feedback in image retrieval: A comprehensive review , 2003, Multimedia Systems.

[9]  Ingemar J. Cox,et al.  An optimized interaction strategy for Bayesian relevance feedback , 1998, Proceedings. 1998 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (Cat. No.98CB36231).