Prof. Xudong Jiang, Nanyang Technological University, Singapore

Speech Title: Vision and Image Recognition: from Subspace Approach and Sparse Coding to Deep Learning

Abstract: Vision and image recognition handles high-dimensional data that contains rich information. Fully utilizing the rich information in image undoubtedly increases the possibilities of solving difficult real world problems such as identifying people, object and understanding the behavior of people, objects and crowds. This, however, brings the difficulty for us to design a robust recognition system due to the complex characteristics of image and large variations of images taken under different conditions. Machine learning from the training database is a solution to extract effective features from the high dimensional image for classification. It is thus not a surprise that approaches of the learning-based methods emerge in various research journals and conferences. This speech reviews various research efforts and technologies developed in solving difficult real world vision and image recognition problems. The first attempt is subspace approaches that extract features or reduce the data dimensionality. The speech will be far more than just PCA and LDA. The sparse representation-based classifier (SRC) significantly differentiates itself from the other classifiers in three aspects. One is the utilization of training samples of all classes collaboratively to represent the query images and another is the sparse representation code that coincides with the general classification target. The last is the L1-norm minimization of the representation error that enables SRC to recognize query images heavily corrupted by outlier pixels and occlusions. The analysis of these three merits of SRC pave the way for us to investigate how the recent developments solve these problems and overcome the limitations of SRC, which bring the sparse representation-based image classification to a significantly higher level. Finally, deep learning in vision and image recognition, CNN, is explored and its merits and limitations are investigated.

Biography: Xudong Jiang received the B.Eng. and M.Eng. degree from the University of Electronic Science and Technology of China, Chengdu, China in 1983 and 1986, respectively, and received the Ph.D. degree from the Helmut Schmidt University Hamburg, Germany in 1997, all in electrical and electronic engineering.

From 1986 to 1993, he worked as Lecturer at the University of Electronic Science and Technology of China where he received two Science and Technology Awards from the Ministry for Electronic Industry of China. He was a recipient of the German Konrad-Adenauer Foundation young scientist scholarship. From 1993 to 1997, he was with the Helmut Schmidt University Hamburg, Germany as scientific assistant. From 1998 to 2002, He worked with the Centre for Signal Processing (CSP),Nanyang Technological University, Singapore, first as Research Fellow and then as Senior Research Fellow, where he developed a fingerprint verification algorithm that achieved the fastest and the second most accurate fingerprint verification in theInternational Fingerprint Verification Competition (FVC2000). From 2002 to 2004 he worked as Lead Scientist and appointed as the Head of Biometrics Laboratory at theInstitute for Infocomm Research, A*Star, Singapore. From 2002 to 2004 he was an Adjunct Assistant Professor. and joined NTU as a full time faculty member in 2004. Currently, Dr Jiang is an Associate Professor (tenured) of School of Electrical and Electronic Engineering, Nanyang Technological University and is appointed as Director of Centre for Information Security (CIS).

Dr Jiang has published over seventy research papers in international refereed journals and conferences. He is also an inventor of one PCT patent application, three Singapore patents and three United States patents, some of which were commercialized. Dr Jiang is a senior member of IEEE and has been serving as Editorial Board Member, Guest Editor and Reviewer of multiple international journals, and serving as Program Committee member, Keynote Speaker and Session Chair of multiple international conferences. His research interest includes pattern recognition, computer vision, image and signal processing, biometrics, face recognition and fingerprint recognition.

Keynote Speaker II

Prof. Dr. Q. M. Jonathan Wu, Chair, Computer Vision and Sensing Systems Laboratory, University of Windsor, Canada

Speech Title: Generalized ELM-Deep Network Framework for Representation Learning

Abstract: Most of actual images such as human face images, industrial images and MRI images are high-dimensional data. The feature representation is mainly for the purpose of extracting useful information and of using this information to build non-supervised classifier/supervised classifier or other types of predictor because the image processing performance is often closely related to the feature data extracted and used. In this talk, we propose a generalized ELM-Deep learning framework which is intended to extract the optimized features. Then, we extend and apply this method for such application fields as dimension reduction, image identification, and image reconstruction, etc. Compared with other feature representation methods, the experimental results show that the generalization performance of the proposed generalized learning framework is very advantageous. A brief overview of other related research activities in the presenter's laboratory related to computer vision and machine learning is also provided. Applications have been extended towards intelligent transportation systems, surveillance and security, face and gesture recognition, vision-guided robotics, and bio-medical imaging, among others.

 Biography: Q. M. Jonathan Wu (M’92–SM’09) received the Ph.D. degree in electrical engineering from the University of Wales, Swansea, U.K., in 1990. He was with the National Research Council of Canada for ten years from 1995, where he became a Senior Research Officer and a Group Leader. He is currently a Professor with the Department of Electrical and Computer Engineering, University of Windsor, Windsor, ON, Canada. He has published more than 300 peer-reviewed papers in computer vision, image processing, intelligent systems, robotics, and integrated microsystems. His current research interests include machine learning, 3-D computer vision, video content analysis, interactive multimedia, sensor analysis and fusion, and visual sensor networks.
Dr. Wu holds the Tier 1 Canada Research Chair in Automotive Sensors and Information Systems. He was Associate Editor for IEEE Transactions on Systems, Man, and Cybernetics Part A, and the International Journal of Robotics and Automation. Currently, he is an Associate Editor for the IEEE Transaction on Neural Networks and Learning Systems and the journal of Cognitive Computation. He has served on technical program committees and international advisory committees for many prestigious conferences.

Keynote Speaker III
Prof. Guo-Neng LU, Claude Bernard Lyon I Univeristy, France

Guo-Neng Lu received the B.S. degree from South-China University of Technology in 1981, the DEA (French equivalent B.S.) degree from Central Engineering School of Lyon in France in 1984, and the PhD degree from Paris-Sud University (Paris 11) in 1986. He obtained HDR (Habilitation à Diriger des Recherches – French research supervision certification) from Paris Diderot University (Paris 7) in 1998.
From 1988 to 1999, he was an associate professor at Paris Diderot University (Paris 7). Since 1999, he has been a full professor at Electrical Engineering Department of Claude Bernard Lyon 1 University.
For his research activities, he has been working at Lyon Institute of Nanotechnology, on integrated sensors and associated electronics. His current researches mainly focus on CMOS image sensors, photodetectors and dosimetric devices and systems, with industrial and academic collaborations. He has supervised 23 PhD students and has authored and co-authored more than 150 papers in international journals and conference proceedings.
eech Title: Development of small-sized pixel structures for high-resolution CMOS image sensors
Abstract: We present our studies on small-sized pixel structures for high-resolution CMOS image sensors. To minimize the number of pixel components, single-transistor pixel and 2T pixel architecture were proposed. To deal with crosstalk between pixels, MOS capacitor deep trench isolation (CDTI) was integrated. CDTI-integrated pixel allows better achievements in dark current and full-well capacity in comparison with the configuration integrating oxide-filled deep trench isolation (DTI). To improve quantum efficiency (QE) and minimize optical crosstalk, back-side illumination (BSI) was developed. Also, vertical photodiode was proposed to maximize its charge-collection region. To take advantages of these structures/technologies, we developed two pixel options (P-type and N-type) combining CDTI or DTI, BSI and vertical photodiode. All the presented pixel structures were designed in 1.4µm-pitch sensor arrays, fabricated and tested.