最新消息

2013年5月20日学术报告:Vision, Learning and Applications,悉尼科技大学张健教授


添加时间:2013-05-15 12:58:03



报告人:悉尼科技大学张健教授

报告题目:Vision, Learning and Applications

时间:2013520日(星期一)下午2:30

地点中心教学楼1003

报告摘要:

In this talk, I will give a comprehensive overview of our research outcomes in vision and learning. I will show our work of feature detection, selection and extraction for object detection. Boosting, SVM, structured learning, inference and other machine learning tools are used in our research. Building upon the findings of our experiments for pedestrian detection, we propose a new, simpler pedestrian detector using the covariance features. Unlike the work in other references, where the feature selection and weak classifier training are performed on the Riemannian manifold, we select features and train weak classifiers in the Euclidean space for faster computation. A set of demo will be shown to link our research to many applications including vehicle, boat and human detection. We will then extend our talk to a higher semantic level of human action classification and recognition. In particular, localizing when and where a specific action happens in realistic videos is a prohibitive computation task. After a brief overview the state-of-the-art techniques applied into the human action recognition task, we propose a fast human action localization framework by employing sparse coding techniques, which is towards building an efficient human action recognition strategy. In our recent research, we focus on 3D motion estimation on RGB-D data of deformable surfaces. We address the challenging problem of motion estimation of 3D range points on deformable objects captured by an RGB-D camera. We introduce a hierarchical motion estimation scheme that leverages both color and depth information to derive the motion vectors. We formulate motion vector MVs constraints into an energy cost function. Through an optimization process, the optimized MVs are derived. Our algorithm is robust when range images contain holes or occlusions. Finally, I would like to share some progress about our recent work on large scale image retrieval across different social media domains. 

------------------------------------------------------------------------------------------------------------------

张健教授简历

Jian Zhang is currently an Associate Professor in Faculty of Engineering and Information Technology at University of Technology, Sydney (UTS), where he is a research leader of Multimedia and Media Analytics Program. He was a Conjoint Associate Professor in the School of Computer Science and Engineering at the University of New South Wales.

From January 2004 - July 2011, Prof. Zhang was a Principal Researcher with National ICT Australia (NICTA), where he was a research leader of Multimedia and Video Communications Research at NICTA Sydney Lab in UNSW Kensington campus. He led several NICTA research projects in the areas of computer vision, multimedia content analysis and management, and multimedia content indexing and query.



作者:杨蛟龙