Welcome to the MLVMA 2008 website!

Vision-based motion analysis aims to detect, track and identify objects, and more generally, to understand their behaviors, from video sequences. This exciting research area has received growing interest in recent years due to a wide spectrum of proposing applications such as visual surveillance, human-machine interface, virtual reality, and motion analysis. Statistical machine learning algorithms have been recently successfully applied to address challenging problems involved in this area. Novel statistical learning technologies have strong potential to contribute to the development of robust yet flexible vision systems. The process of improving the performance of vision systems has also brought new challenges to the field of machine learning. Solving the problems involved in object motion analysis will lead to the development of new machine learning algorithms. In return, new machine learning algorithms are able to address more realistic problems in object motion analysis and understanding.

This one-day workshop seeks to present and highlight the latest developments in vision-based object motion analysis and understanding from a machine learning perspective. It aims to bring together worldwide researchers from related disciplines, to provide a forum for the dissemination of significant research work and innovative practice, and to encourage exchanges, interactions and possible collaboration between participants.

What’s new:

  • August 20th: Technical program is available
  • August 7th: Paper acceptance notification was sent.
  • July 15th: The submission server was closed.
  • July 1st: The submission deadline was extended to 13 July
  • June 16th: Information on invited speeches was updated
  • June 16th: ViHASi: Virtual Human Action Silhouette Data>
  • June 16th: Second call for papers
  • May 5th: Invited speakers were confirmed
  • April 28th: Online submission was already open>
  • April 1st: First call for papers
  • March 15th: MLVMA08 website>

Welcome to the MLVMA 2008 website!

Vision-based motion analysis aims to detect, track and identify objects, and more generally, to understand their behaviors, from video sequences. This exciting research area has received growing interest in recent years due to a wide spectrum of proposing applications such as visual surveillance, human-machine interface, virtual reality, and motion analysis. Statistical machine learning algorithms have been recently successfully applied to address challenging problems involved in this area. Novel statistical learning technologies have strong potential to contribute to the development of robust yet flexible vision systems. The process of improving the performance of vision systems has also brought new challenges to the field of machine learning. Solving the problems involved in object motion analysis will lead to the development of new machine learning algorithms. In return, new machine learning algorithms are able to address more realistic problems in object motion analysis and understanding.

This one-day workshop seeks to present and highlight the latest developments in vision-based object motion analysis and understanding from a machine learning perspective. It aims to bring together worldwide researchers from related disciplines, to provide a forum for the dissemination of significant research work and innovative practice, and to encourage exchanges, interactions and possible collaboration between participants.

Program

08:50-08:55 Opening Address

09:00-09:50 Invited Talk 1

From learning individual actions to 3D animation of team sports
Prof. Stefan Carlsson, KTH (Royal Institute of Technology), Sweden

09:50-10:30 Session 1: Motion Tracking

Learning Bayesian tracking for motion estimation
Michael Felsberg and Fredrik Larsson, Linkoping University, Sweden

Human motion tracking using a color-based particle filter driven by optical flow
Tony Tung and Takashi Matsuyama, Kyoto University, Japan

10:30 - 11:00 Coffee Break

11:00 - 12:20 Session 2: Action and Behavior analysis

Flexible dictionaries for action classification
Michalis Raptis, Kamil Wnuk and Stefano Soatto, UCLA VisionLab, USA

A framework for indexing human actions in video
Kaustubh Kulkarni, Srikanth Cherla, Amit Kale and V Ramasubramanian,Siemens - Corporate Technology, India

Independent viewpoint silhouette-based human action modeling and recognition
Carlos Orrite, Francisco Martinez, Elias Herrero, University of Zaragoza, Spain
Hossein Ragheb and Sergio A. Velastin, Kingston University, UK

From local temporal correlation to global anomaly detection
Chen Change Loy, Tao Xiang and Shaogang Gong, Queen Mary, University of London, UK

12:20 - 14:00 Lunch Break

14:00-14:50 Invited Talk 2

Linear and Non-Linear Models for Monocular 3D Motion Capture
Prof. Pascal Fua, EPFL, Switzerland

14:50-16:00: Session 3: Posters

15:30 - 16:00 Coffee Break

16:00 - 17:20 Session 4: Learning Methods

Super-resolved digests of humans in video
Dong Seon Cheng, Marco Cristani and Vittorio Murino, University of Verona, Italy

Learning Pullback metrics for linear models
Fabio Cuzzolin, INRIA Rhone-Alpes, France

Spatio-temporal motion pattern modeling of extremely crowded scenes
Louis Kratz and Ko Nishino, Drexel University, USA

Simultaneous learning of motion and appearance
Karel Zimmermann, Tomas Svoboda and Jiri Matas, Czech Technical University, Czech Republic

Posters:

Approximate RBF kernel SVM and its applications in pedestrian classification
Hui Cao, Takashi Naito and Yoshiki Ninomiya, Toyota Central R&D LABS.,INC. Japan

Combination of supervised and unsupervised methods for navigation path mining
Naoya Ohnishi and Atsushi Imiya, Chiba University, Japan

Self-similar regularization of optic-flow for turbulent motion estimation
Patrick Héas and Dominique Heitz, Cemagref, France
Etienne Mémin, INRIA Rennes France

Facial motion analysis using clustered shortest path tree registration
David Cristinacce, Natalie Butcher and Tim Cootes, University of Manchester, UK

Optimizing trajectories clustering for activity recognition
Monique Thonnat, Francois Brémond, Guido Pusiol and Jose Luis Patino, INRIA, Sophia Antpolis, France

Spatio-temporal feature recognition using randomised Ferns
Olusegun Oshin, Andrew Gilbert, John Illingworth and Richard Bowden, University of Surrey , UK

Unsupervised learning of behavioural patterns for video-surveillance
Nicoletta Noceti, Matteo Santoro and Francesca Odone, DISI , Italy

Capturing video structure with mixture of probabilistic index maps
Alessandro Perina, Marco Cristani, Vittorio Murino, University of Verona, Italy
Nebojsa Jojic, Microsoft Corporaton, USA

A new spatio-temporal MRF framework for video-based object segmentation
Rui Huang, Vladimir Pavlovic and Dimitris Metaxas, Rutgers University, USA

partners MLVMA08