Learning kernels on extended Reeb graphs for 3d shape classification and retrieval

Abstract

This paper addresses 3D shape classification and retrieval in terms of supervised selection of the most significant features in a space of attributed graphs encoding different shape characteristics. For this purpose, 3D models are represented as bags of shortest paths defined over well chosen Extended Reeb graphs, while the similarity between pairs of Extended Reeb graphs is addressed through kernels adapted to these descriptions. Given this set of kernels, a Multiple Kernel Learning algorithm is used to find an optimal linear combination of kernels for classification and retrieval purposes. Results are comparable with the best results of the literature, and the modularity and flexibility of the kernel learning ensure its applicability to a large set of methods.