2003 Workshop on Multi-Object Tracking (WOMOT)

Co-Chairs

John Krumm

Microsoft Research, Redmond, WA, USA

David Beymer

IBM Almaden Research, San Jose, CA, USA

Program Committee

J. K. Aggarwal

University of Texas at Austin, Austin, TX, USA

Dorin Comaniciu

Siemens Research

Princeton, NJ, USA

Trevor Darrell

MIT, Cambridge, MA, USA

James W. Davis

Ohio State University, Columbus, OH, USA

Larry Davis

University of Maryland, College Park, MD, USA

James Ferryman
University of Reading, Reading, UK

Gaile Gordon

Tyzx, Inc., Palo Alto, CA, USA

Neil J. Gordon

Defence Science and Technology Organisation, Adelaide, AUSTRALIA

Vera Kettnaker

Rennselaer Polytechnic Institute, Troy, NY, USA

T. Kirubarajan

McMaster University, Hamilton, Ontario, CANADA

John MacCormick

HP Labs, Palo Alto, CA, USA

Ronald P. Mahler

Lockheed Martin Tactical Systems, Eagan, MN, USA
 

Simon Maskell

QinetiQ and Cambridge University, ENGLAND

James Rehg

Georgia Institute of Technology, Atlanta, GA, USA

Tieniu Tan

National Laboratory of Pattern Recognition, Beijing, CHINA

The 2003 Workshop on Multi-Object Tracking (WOMOT) is scheduled as part of CVPR 2003 to be held in Madison, Wisconsin in June. This will be the second WOMOT. The first one was held as part of ICCV 2001 and had 87 registered attendees.

WOMOT is intended to highlight problems and solutions unique to the task of tracking multiple objects simultaneously. The range of objects is wide, including airplanes, missiles, vehicles, people, animals, insects, and microorganisms. Tracking multiple such objects is complicated by the fact that the objects can touch and occlude each other, enter and leave the sensors' fields of view, and interact and intertwine. Sometimes the objects are coupled, either strongly (e.g. body parts) or weakly (e.g. swarms). Sometimes the objects are distinguishable, sometimes not, and sometimes they slowly differentiate themselves over time. These unique challenges invite interesting approaches in state representation, dynamic modeling, and sensor processing. Besides the multi-object tracking problems, these algorithms must deal with all the usual problems of single-object tracking, including object deformation, articulation, occlusion, illumination changes, and background changes.

Since CVPR is largely a computer vision conference, we expect an emphasis on video tracking at WOMOT. However, the same fundamental problems of multi-object tracking exist for many other sensors, and we invite and encourage participation from outside the vision community in an effort to help spread applicable ideas between fields.

WOMOT03 will consist of one day of oral presentations, including a half day of short tutorials. The program is here.