Fast multi-view graph kernels for object classification

  • Authors:

  • Luming Zhang;Mingli Song;Jiajun Bu;Chun Chen

  • Affiliations:

  • College of Computer Science, Zhejiang University, China;College of Computer Science, Zhejiang University, China;College of Computer Science, Zhejiang University, China;College of Computer Science, Zhejiang University, China

  • Venue:
  • AI'11 Proceedings of the 24th international conference on Advances in Artificial Intelligence
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

Object classification is an important problem in multimedia information retrieval. In order to better objects classification, we often employ a set of multi-view images to describe an object for classification. However, two issues remain unsolved: 1) exploiting the spatial relations of local features in the multi-view images for classification, and 2) accelerating the classification process. To solve them, Fast Multi-view Graph Kernel (FMGK), is proposed. Given a set of multi-view images for an object, we segment each view image into several regions. And inter- and intra- view linkage graphs are constructed to describe the spatial relations of the regions between and within each multi-view image respectively. Then, the inter- and intra- view graphs are integrated into a so-called multi-view region graph. Finally, the kernel between objects is computed by accumulating all matchings' of walk structures between corresponding multi-view region graphs. And a SVM [11] classifier is trained based on the computed kernels for object classification. The experimental results on different datasets validate the effectiveness of our FMGK.