HKBU  |  SCI  |  BUniPort  |  Library  |  Alumni  |  Job Vacancies  |  Intranet  |  Sitemap        
Undergraduate Admissions
Taught Postgraduate Admissions
Research Postgraduate Admissions
Job Vacancies
News & Achievements
Events
Video
Research Highlights
Contact & Direction
International Exchange and Internship Programmes
 
HONG KONG BAPTIST UNIVERSITY
FACULTY OF SCIENCE

Department of Computer Science Colloquium
2005 Series

Kernel Learning for Semantic Image Retrieval and Classification

Prof. Kap Luk CHAN
Nanyang Technological University
Singapore

Date: December 12, 2005 (Monday)
Time: 3:00 - 4:00 pm
Venue: RRS905, Sir Run Run Shaw Building, Ho Sin Hang Campus

Abstract
The problem of semantic gap in image retrieval and classification is receiving considerable attention currently as evident in recent top conferences in computer vision and multimedia information retrieval. In recent years, kernel methods, and in particular, the Support Vector Machines (SVMs), have been used to address this problem by establishing a nonlinear relationship between the low level visual features and the high level human perception. The kernel used in SVM is therefore of crucial importance to encapsulate this relationship. Kernel learning then becomes an important task. The existing model selection methods are often not really useful and applicable here due to the real-time response requirement and the problem of small sample size. In this talk, a general framework of kernel learning is proposed to deal with these problems. We represent the available semantic knowledge in the form of a target kernel, and a criterion is further proposed to minimize the difference between the employed kernel and the target kernel. The optimal kernel is achieved by minimizing the criterion through gradient-based search methods, incurring very little computational overhead to real-time applications. This allows for a dynamically refreshed kernel to be used in an iterative retrieval process to maximize the information available in the training samples. The proposed framework facilitates both kernel learning from sample labels and sample similarity information. This framework has been applied to content-based image retrieval and textured image classification, showing promising results.

Biography
Associate Professor Kap Luk Chan obtained his PhD degree in Robot Vision from Imperial College of Science, Technology and Medicine, University of London, London, U.K. in 1991. He is now an associate professor in the School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore. His research interests are in image analysis and computer vision, particularly in texture analysis, statistical image analysis and perceptual grouping, image and video retrieval, machine learning in computer vision, and biomedical signal and image analysis. He is on joint appointment in the division of Bioengineering, School of Chemical and Biomedical Engineering and currently appointed as a deputy director of the Biomedical Engineering Research Centre of the Nanyang Technological University. He has also been a consultant to local and multinational companies in Singapore. He is a member of the IEEE and IEE.

********* ALL INTERESTED ARE WELCOME ***********
(For enquiry, please contact Computer Science Department at 3411 2385)

http://www.comp.hkbu.edu.hk/v1/?page=seminars&id=95
Copyright © 2017. All rights reserved.Privacy Policy
Department of Computer Science, Hong Kong Baptist University
Hong Kong Baptist University