On-line Learning of Object Representations
Radial Basis Function (RBF) networks have been proposed as suitable representations for 3-D objects, in particular, since they can learn view-based representations from a small set of training views. One of the basic questions that arises in the context of RBF networks concerns their complexity, i.e., the number of basis functions that are necessary for a reliable representation, which should balance the accuracy and the robustness. In this paper we propose a systematic approach for building object representations in terms of RBF networks. We studied and designed two procedures: the “off-line” procedure, where the network is constructed after having a complete set of training views of an object, and the “on-line” procedure, where the network is incrementally built as new views of an object arrive.
KeywordsBasis Function Radial Basis Function Object Representation Radial Basis Function Network Stick Figure
Unable to display preview. Download preview PDF.
- H. H. Bülthoff and S. Y. Edelman and M. Tarr, How are three-dimensional objects represented in the brain?, A. I. Memo No. 1479, C. B. C. L. Paper No. 96, Massachusetts Institute of Technology, April 1994.Google Scholar
- S. Chandrasekaran, B.S. Manjunath, Y.F. Wang, J. Winkler, and H. Zhang, An eigenspace update algorithm for image analysis, Technical report, TR CS 96-04, Dept. of Computer Science, Univ. of California, Santa Barbara.Google Scholar
- Y. L. Le Cun and J. S. Denker and S. A. Solla. Optimal Brain Damage. Advances in Neural Information Processing Systems 2. Ed. D. S. Touretzky. San Mateo, CA: Morgan Kaufmann, pp. 598–605, 1988.Google Scholar
- B. Hassibi and D. G. Stork. Second Order Derivatives for network Pruning: Optimal Brain Surgeon. In Proceedings of NIPS 5. Ed. S. J. Hanson et al. Morgan Kaufmann, pp. 164–172, 1993.Google Scholar
- A. Leonardis and H. Bischof. Complexity Optimization of Adaptive RBF Networks. In Proceedings of the ICPR′96 Vol IV pages 654–658, IEEE Computer Society Press, 1996.Google Scholar
- D. Lowe. Adaptive radial basis function nonlin-earities, and the problem of generalisation. In 1st IEE Conference on Artificial Neural Networks pages 171–175. London U.K., 1989.Google Scholar
- S. Mukherjee and S. K. Nayar. Automatic Generation of GRBF networks for visual learning. In Proceedings of the ICCV′95 pages 794–800, 1995.Google Scholar
- J. Ponce and A. Zisserman and M. Hebert (Eds.), Object Representation in Computer Vision II, ECCV′96 International Workshop, Cambridge, U.K., April 1996, Springer Verlag, LNCS-1144.Google Scholar
- L. Sardo and J. Kittler, Complexity analysis of RBF networks for pattern recognition, In Proceedings of the CVPR′96 pp. 574–579, 1996.Google Scholar
- M. Stricker and A. Leonardis. Figure ground segmentation using tabu search. In Proceedings of the IEEE Int. Symposium on Computer Vision Coral Gables, Florida, November 1995.Google Scholar
- T. Werner, R. D. Hersch, and V. Hlaváč, Rendering real-world objects using view interpolation. In Proceedings of the ICCV′95, pp. 957–962, Boston, USA, June 1995. IEEE Press.Google Scholar