Homepage of Zhe Lin^-^

Zhe Lin

I am a PhD Student in the Department of Electrical and Computer Engineering, University of Maryland at College Park, supported by Fellowship and Graduate Assistantship. I am currently working as a full time Graduate Research Assistant in Language and Media Processing Laboratory, Institute for Advanced Computer Studies (UMIACS), supervised by Prof. Larry Davis, Dr. David Doermann and Dr. Daniel DeMenthon.

Office: RM3348 A.V.Williams Bldg., Dept. of ECE, Univ. of Maryland, College Park, MD 20742
Home: 6100 Westchester Park Dr., Apt. 512, College Park, MD, 20740
E-mail: zhelin@umiacs Homepage: http://www.glue.umd.edu/zhelin/
Phone: 1-240-687-8440

Education

1997.9~2002.7
B.Eng. Dept. of Automation, University of Science and Technology of China (USTC), Advisor: Prof. Zengfu Wang
2002.9~2004.8
M.S. Dept. of EECS, Korea Advanced Institute of Science and Technology (KAIST), Advisor: Prof. In So Kweon
2005.8~present
PhD Student, Dept. of ECE, University of Maryland at College Park, Advisor: Prof. Larry S. Davis

Current GPA: 4.0/4.0, PhD Qualifying Written and Oral Exam (Pass, May 2006)
PhD Proposal Defense (Pass, Jan 2008), Expected Graduation (Summer 2009)

Research

Areas of Interests

Object Detection, Recognition and Categorization
Pattern Classification and Machine Learning
Feature Extraction, Image and Object Segmentation
Image/Video Retrieval, Tracking and Motion Analysis
Vision-based Robot Localization and Navigation

Current Research

Appearance Matching and Recognition
Developed pairwise discriminative learning-based techniques to appearance-based person recognition to better handle the scalability problem, i.e. recognition of large number of categories. Also developed algorithms for matching appearances under occlusion. The approach is tested on 61 person Honeywell appearance dataset, and 50 person multiview, multipose Keck lab (UMD) dataset.
Related papers: Z. Lin and L. S. Davis, Some paper, submitted to ECCV'08 and Z. Lin, L. S. Davis, D. Doermann and D. DeMenthon, Simultaneous appearance modeling and segmentation for matching people under occlusion, ACCV'07.
Part of this work is delivered to Panasonic Inc. Princeton, NJ
Source codes and datasets are available later.
Multiple Occluded Human Detection, Segmentation and Tracking
Developed algorithms for detecting and segmenting humans under occlusion. The detection is performed by a part-template shape matching method and the segmentation is done by pose-assisted appearance-based segmentation which simultaneously estimates human poses and segmentation. We are currently applying these approaches to multi-target tracking system for visual surveillance applications.
Vision-based Path Planning, Localization and Navigation System
Feature-based Object Recognition System
Development of Robust Visual Features

Work Experience

Research Intern, Microsoft Live Labs Research, Redmond, WA, May 2008-
- Scalable object detection in still and motion images
Part-time Researcher, Software Development at Hanool Robotics Corp., 2003-2004
- Object recognition algorithm porting and optimization on Strong-ARM embedded system.
- Development of fast object recognition system on Hanuri-RD mobile robot.

Professional Activities

Visiting Student
- Electronics and Telecommunications Research Institute (ETRI), Daejeon, Korea, Aug. 2004
Reviewer for
- International Conference on Pattern Recognition (ICPR), 2006, 2008
- Computer Vision and Image Understanding (CVIU), 2008
Membership
- IEEE Student Member, 2004~Present
- Korea Robotics Society (KROS) Student Member, 2004~Present

Publications

Thesis
- "Development of robo-cup soccer robot vision system," Bachelor's Thesis, 2002.6.
- "Robust local invariant features for real-time object recognition," Master's Thesis, 2004.6.
Conference
- Zhe Lin and Larry S. Davis, "A Pose-Invariant Descriptor for Human Detection and Segmentation", 10th European Conference on Computer Vision (ECCV 2008), Marseille, France, 2008, accepted.
- Zhe Lin, Larry S. Davis, David Doermann, Daniel DeMenthon, "An Interactive Approach to Pose-Assisted and Appearance-based Segmentation of Humans," ICCV Workshop on Interactive Computer Vision (ICV 2007), Rio de Janeiro, Brazil, 2007.pdf
  Annotated 404 pedestrian silhouettes (selected from INRIA person dataset): download
- Zhe Lin, Larry S. Davis, David Doermann, Daniel DeMenthon, "Simultaneous Appearance Modeling and Segmentation for Matching People under Occlusion," Asian Conference on Computer Vision (ACCV 2007), Tokyo, Japan, 2007.pdf
- Zhe Lin, Larry S. Davis, David Doermann, Daniel DeMenthon, "Hierarchical Part-Template Matching for Human Detection and Segmentation," IEEE International Conference on Computer Vision (ICCV 2007), Rio de Janeiro, Brazil, 2007.pdf
  Part-template-tree model: download
  Dataset (200 frames from Caviar data and 200 frames from Munich Airport data) with annotation: download
- Son Tran, Zhe Lin, David Harwood, Larry Davis, "UMD_VDT, an Integration of Detection and Tracking Methods for Multiple Human Tracking," Classification of Events, Activities and Relationships Workshop (CLEAR 2007). pdf
- Zhe Lin, Sungho Kim, In So Kweon, "Object recognition and indoor topological navigation using robust invariant features," 2005 IEEE/RSJ Conference on Intelligent Robots and Systems (IROS 2005), Edmonston, Canada, 2005. pdf
- Zhe Lin, Sungho Kim, In So Kweon, "Robust invariant features for object recognition and mobile robot navigation," 2005 IAPR Conference on Machine Vision Applications (MVA 2005), Tsukuba Science City, Japan, 2005. pdf
- Zhe Lin and In So Kweon, "Robust invariant feature extraction for object recognition and natural landmark based autonomous navigation," First International Conference on Ubiquitous and Intelligent Robot (ICUIR 2004), The Korea International Robot Week, COEX, Seoul, Nov. 2004.
- Sungho Kim, Zhe Lin, Wangheon Lee and In So Kweon, "3D object recognition system based on the robust properties of human visual system," First International Conference on Ubiquitous and Intelligent Robot (ICUIR 2004), The Korea International Robot Week, COEX, Seoul, Nov. 2004.
- Zhe Lin and In So Kweon, "A novel scale-invariant feature for object recognition and mobile robot navigation," Proceedings of the 11th IEEE International Conference on Mechatronics and Machine Vision in Practice (M2VIP 2004), Macau, 2005.
Journal
- Zhe Lin and In So Kweon, "Robust invariant features for object recognition, pose estimation and topological navigation," International Journal on Human Friendly Welfare Robotic Systems (IJHWRS), Volume 6, No. 1, Mar. 2005.

Graduate Courses

Fall, 2006

CMSC828J Image Segmentation, Prof. David Jacobs -------- A-
ENEE698A Pattern Recognition from Temporal Data, Prof. Rama Chellappa -------- A
ENEE698C Controls Seminar: Dynamics of Small World Networks, Prof. Eyad H. Abed, -------- S

Spring, 2006

CMSC828J Approaches to Representing and Recognizing Objects, Prof. David Jacobs -------- A-
ENEE698A Graduate Seminar: Communications, Prof. Richard J. La, -------- A-
CMSC733 Computer Processing of Pictorial Information, Prof. Yiannis Aloimonos, Course Projects -------- A+