EHuM2: 2-nd Workshop on Evaluation of Articulated Human Motion and Pose Estimation

(Workshop at CVPR 2007)

Home Schedule Registration HumanEva Dataset Submissions

Frontpage Template Resources


- CFP (pdf, ps)

- Register to obtain dataset.



Organizing Committee:

Leonid Sigal (Brown U)

Michael J. Black (Brown U)

Horst Haussecker (Intel)


Program Committee:

Ankur Agarwal (Microsoft Research)

Stefan Carlsson (KTH)
Trevor Darrell (MIT)
James Davis (UC Santa Cruz)
Larry Davis (U of Maryland)
David Fleet (U of Toronto)
David Forsyth (UIUC)
Pascal Fua (EPFL)
Horst Haussecker (Intel)
Daniel Huttenlocher (Cornell U)
Ram Nevatia (USC)
Deva Ramanan (TTI-C)
James Rehg (Georgia Tech)

Stan Sclaroff (Boston U)
Cristian Sminchisescu (TTI-C)

Philip Torr (Oxford Brookes)
Bill Triggs (INRIA)
Ying Wu (Northwestern U)

Ming-Hsuan Yang (Honda)


Invited Speakers:

David Fleet (U of Toronto)






Morning Session (9:00am - 12:30pm)
9:00 am Opening remarks
Workshop Organizers
9:10 am Keynote Talk in Compter Vision
David Fleet
University of Toronto, Canada
10:10 am coffee break  
10:30 am HumanEva datasets and evaluation metrics
Leonid Sigal
Brown University, RI, USA
  Paper Session  
10:45 am Evaluating Example-based Pose Estimation: Experiments on the HumanEva Sets
Ronald Poppe
University of Twente, Netherlands
11:15 am

Articulated Body Pose Estimation from Voxel Reconstructions using Kinematically Constrained Gaussian Mixture Models: Algorithm and Evaluation
Shinko Y. Cheng and Mohan M. Trivedi
University of California, San Diego, CA, USA

11:45 am Evaluation of a Hierarchical Partitioned Particle Filter with Action Primitives
Zsolt L. Husz, Andrew M. Wallace and Patrick R. Green
Heriot-Watt University, UK
  Spotlight results presentations  
12:15 pm 3D Human Motion from Fusing Multiple Percepts
Jane Mulligan
University of Colorado at Boulder, CO, USA
Lunch (12:30pm - 2:00pm)
Afternoon Session (2:00pm - 5:00pm)
2:00 pm Keynote Talk in Biomechanics: Markerless motion capture and Biomechanics
Stefano Corazza

Stanford University, CA, USA
3:00 pm coffee break  
  Invited Technical Sketches  
3:30 pm Cloth X-Ray: Clothing Models for Markerless Motion Capture
Bodo Rosenhahn
Max-Planck-Center Saarbruecken, Germany


3:55 pm Boosted Multiple Deformable Trees for Parsing Human Poses (Abstract)
Greg Mori
Simon Fraser University, BC, Canada
4:20 pm Bottom-Up Recognition and Parsing of the Human Body
Jianbo Shi
University of Pennsylvania, PA, USA
4:45 pm Closing remarks and Best Paper Award
Workshop organizers


Accepted Papers Not Being Presented
Recognition-Based Motion Capture and the HumanEva II Test Data
Nicholas Howe
Smith College, MA, USA
David Fleet

Bio: David Fleet is professor of computer science at the University of Toronto. He received the PhD in Computer Science from the University of Toronto in 1991. From 1991 to 2000 he was on faculty at Queen's University, Canada, in the Department of Computing and Information Science, with cross-appointments in Psychology and Electrical Engineering. In 1999 he joined the Palo Alto Research Center (PARC) where he managed the Digital Video Analysis Group and the Perceptual Document Analysis Group. He returned to the University of Toronto in October 2003.

In 1996 Dr. Fleet was awarded an Alfred P. Sloan Research Fellowship for his research on biological vision. His 1999 paper with Michael Black on probabilistic detection and tracking of motion boundaries received Honorable Mention for the Marr Prize at the IEEE International Conference on Computer Vision. His 2001 paper with Allan Jepson and Thomas El-Maraghi on robust appearance models for visual tracking was awarded runner-up for the best paper at the IEEE Conference on Computer Vision and Pattern Recognition. In 2003, his paper with Eric Saund, James Mahoney and Dan Larner won the best paper award at ACM UIST '03. He has been associate editor of IEEE Transactions on Pattern Analysis and Machine Intelligence (2000-2004), and program co-chair for the IEEE Conference on Computer Vision and Pattern Recognition in 2003. He is currently Associate Editor-In-Chief for IEEE Transactions on Pattern Analysis and Machine Intelligence, and a Fellow of the Canadian Institute of Advanced Research.

His research interests include computer vision, image processing, visual perception, and visual neuroscience. He has published research articles and one book on various topics including the estimation of optical flow and stereoscopic disparity, probabilistic methods in motion analysis, 3D people tracking, modeling appearance in image sequences, non-Fourier motion and stereo perception, and the neural basis of stereo vision.


Stefano Corazza


Today most common methods for accurate capture of three-dimensional human movement require a laboratory environment and the attachment of markers or fixtures to the body's segments. These laboratory conditions can cause
unknown experimental artifacts. Modern biomechanical and clinical applications require the accurate capture of normal and pathological human movement without the artifacts associated with standard marker-based motion capture techniques such as soft tissue artifacts and the risk of artificial stimulus of taped-on or strapped-on markers. The need for accurate
information on the characteristics of normal and pathological human is motivated in part by the introduction of new clinical approaches for the treatment and prevention of diseases that are influenced by subtle changes in the patterns movement. The need for markerless human motion capture methods is discussed and the advancement of markerless approaches is
considered in view of accurate capture of three-dimensional human movement for biomechanical applications. The role of choosing appropriate technical equipment and algorithms for accurate markerless motion capture is critical. The implementation of this new methodology offers the promise for simple, time-efficient, and potentially more meaningful assessments of human movement in research and clinical practice. The feasibility of accurately and precisely measuring 3D human body kinematics using a markerless motion capture system is demonstrated.

Bio: The primary mission of the Biomotion Research Group is to study normal and pathological function which can be ultimately applied to the improved evaluation and treatment of musculoskeletal disease and injury. The goals are addressed by studying normal subjects and patients with injury or disease that influence the normal function of the musculoskeletal system. In addition, the BioMotion Research Group is also committed to the development of improved methods for the measurement and analysis of human movement and is working on a markerless system using multiple optical sensors that will efficiently and accurately provide 3D measurements of human movement for application in clinical practice.

Boosted Multiple Deformable Trees for Parsing Human Poses
Greg Mori
Abstract: We present a method for estimating human pose in still images. Tree-structured models have been widely used for this problem. While such models allow efficient learning and inference, they fail to capture additional dependencies between body parts, other than kinematic constraints. In this paper, we consider the use of multiple tree models, rather than a single tree model for human pose estimation. Our model can alleviate the limitations of a single tree-structured model by combining information provided across different tree models. The parameters of each individual tree model are trained via standard learning algorithms in a single tree-structured model. Different tree models are combined in a discriminative fashion by a boosting procedure. We present experimental results showing the improvement of our model over previous approaches on a very challenging dataset.


Copyright Leonid Sigal. All Rights Reserved. Brown University