Assistant Professor, National Taiwan University of Science and Technology
# 43, Sec. 4, Keelung Rd., Taipei 106, Taiwan Tel: +886-2-2730-3664, Fax: ++886-2-2730-1081
Email: hua@mail.ntust.edu.tw (台灣科技大學資訊工程系花凱龍教授)
ABSTRACT
Clothing image analysis has shown its potential for use in a wide range of applications such as personalized clothing recommendation. Given a consumer photo, this talk addresses the problem of finding clothes and recognizing the genre of that clothes. This problem is very challenging due to large variations of uncontrolled realistic imaging conditions. In order to tackle these challenges, we formulate a novel framework by integrating local features of multimodality as the instances of the price-collecting Steiner tree (PCST) problem to discover clothing regions, and exploiting visual style elements to discover the clothing genre. The experimental results show that our fully automatic approach is effective to identify irregular shape of clothing region, and it significantly improves the accuracy of clothing genre
recognition for images taken in unconstrained environment.
BIOGRAPHY
Kai-Lung Hua received the B.S. degree in electrical engineering from National Tsing Hua University in 2000, and the M.S. degree in communication engineering from National Chiao Tung University in 2002, both in Hsinchu, Taiwan. He received the Ph.D. degree from the school of electrical and computer engineering, Purdue University, West Lafayette, IN, in 2010
Since 2010, he has been with National Taiwan University of Science and Technology, Taipei, Taiwan, where he is currently an assistant professor in the department of computer science and information engineering. His current research interests include digital image and video processing, computer vision, and multimedia networking.
Prof. Hua is a member of Eta Kappa Nu and Phi Tau Phi, as well as a recipient of MediaTek Doctoral Fellowship. He served as a guest editor for IEEE MMTC R-Letters in 2012. He also chaired the Multimedia Computing and Communications (MCC) symposium at IEEE International Conference on Computing, Networking and Communications (ICNC) in 2013. He co-organized a special session at Asia-Pacific Signal and Information Processing Association (APSIPA) on Emerging Technologies in Multimedia Communications in 2013. He contributed as a tutorial speaker on Recent Advances in Sensing Techniques of Visual Semantics at The Pacific-Rim Conference on Multimedia (PCM), 2013.
134
Technical Session D2-W3-T3: New Media/Multimedia, Machine Learning, Web, and Entertainment Technology
Introduction to Iris Recognition Yung-Hui Li
Assistant Professor, Department of Computer Science and Information Engineering National Central University
Jhongli City, Taoyuan County, 32001, Taiwan (R.O.C.) Tel: +886-3-4227151 ext 35204, Fax: +886-3-4226062
Email: yunghui@csie.ncu.edu.tw (中央大學資訊工程學系栗永徽教授)
ABSTRACT
In the last few decades, biometric recognition has drawn significant attention due to the vast applications in the field of law enforcement, surveillance, border control and national security.
In 2011, IBM predicted that biometric technology would be one of the five big things that impact human lives in the next five years. In September 2013, Apple ships the first iPhone that has a built-in fingerprint sensor which is able to verify the user’s identity without using
password. Such device, combined with the emergence of mobile e-commerce, may greatly enhance the user experience and further bring a revolution on commercial behavior of both customer and retailer.
There exists many modalities of biometric recognition, including face, fingerprint, voice, signature …etc. Among all the usable characteristics for biometric recognition, the pattern of iris texture is one of the few characteristics believed to be the most distinguishable among different people. In fact, according to what Prof. J. Daugman published in 2004 on IEEE Transaction, the probability of false matching rate of iris recognition can achieve to the order of 10-10. Such high precision makes iris recognition useful for large scale deployment for real world problem. In fact, it has been deployed in immigration checkups in airport in one of the middle-east countries for several years.
Most of the iris recognition system follows the basic framework built by Prof. J. Daugman in 2004. In this talk, I will briefly review the standard process (stages) of iris recognition.
Currently, less constrained and long range iris recognition starts to get more attentions in both academia and industry. There are new challenges that the researchers and engineers are facing today for less constrained iris recognition system. The optical design of such system needs to be re-considered. New image processing technique has been developed. Such issues will also be addressed in this talk.
BIOGRAPHY
Yung-Hui Li received his BS degree from Department of Electrical Engineering in National Taiwan University (NTUEE) in Taipei, Taiwan (R.O.C) in 1995, MS degree from Department of Computer and
Information Science in University of Pennsylvania (UPenn) in 1998, and PhD degree from Language Technology Institute, School of Computer Science in Carnegie Mellon University (CMU) in 2010. He is currently an assistant professor in Department of Computer Science and
Information Engineering, National Central University, Taoyuan, Taiwan (R.O.C).
His research focused on issues about long-range iris recognition, including extended DOF for iris image acquisition, blur iris image segmentation, iris mask generation, and iris image
super-resolution. Other research topics include face recognition, computer vision, pattern recognition and machine learning. He has authored one paper in IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), more than 10
conference papers and co-authored 5 book chapters in this field.
136
Technical Session D2-W3-T3: New Media/Multimedia, Machine Learning, Web, and Entertainment Technology