ISM TRECVID2008 High-level Feature Extraction

We studied a method using support vector machines (SVMs) with walk-based graph kernels for the high-level feature extraction (HLF) task. In this method, each image is first segmented into a finite set of homogeneous segments and then represented as a segmentation graph where each vertex is a segment and edges connect adjacent segments. Given a set of features associated with each segment, we then obtain a positive definite kernel between images by comparing walks in the respective segmentation graphs, and image classification is carried out with an SVM based on this kernel. We submitted six runs using this method with several combinations of the values of the kernel and SVM parameters.

• Type: Conference
• Year: 2008
• Where: TRECVID
• Authors: Tomoko Matsui, Jean-Philippe Vert, Shin'ichi Satoh, Yuji Uchiyama
• Keywords: Internet Technology, Segmentation Graphs, Support Vector Machines, TRECVID 2008, Walk-based Graph Kernels