ECCV 2008 Tutorial

Date: Sunday October 12th - Afternoon

Venue: Palais des Congres Parc Chanot, in Marseille.

Motion Capture in Practice

Speakers: Andrew J Stoddart (Vicon), Steve Caulkin (Image Metrics), Clement Menier (4D View Solutions), Steve Sullivan (Lucasfilm/Industrial Light & Magic)


Tracking human motion has been a topic of intense interest in the computer vision community for at least 20 years. It has also been an active area commercially for about the same period.

This tutorial will present an introduction to the challenges faced when delivering products and services into this market and how they are being met today.

1. [1400-1445] Optical motion capture. Andrew J. Stoddart. [45 mins]

o “It’s the camera, stupid”, the anatomy of a modern mocap camera. Why is custom hardware still so hard to beat?

o Visual Geometry, calibration, reconstruction, trajectories and labeling. It’s not as easy as it looks...

o How to combine automatic & manual tools. The algorithm isn’t always right!

o Kinematic modeling and the VSK markup language. What does the customer actually want?

2. [1445-1530] Performance-driven facial animation, Steve Caulkin, ImageMetrics. [45 mins]

o Overview of performance-driven facial animation

o Goals of facial capture

o Traditional facial motion capture

o Markerless facial capture

o Capture techniques

§ Witness camera

§ Head-mounted camera

§ Multiple cameras/stereo

o Simultaneous facial and body mocap

o Driving facial animation from performance data

*. [1530-1550] Coffee

3. [1550-1635] Capturing Reality, Clement Menier, 4D View Solutions, [45 mins]

o Recording a scene: the challenge of synchronized multiple-view color acquisition.

o 3D modeling of a real dynamic sequence. Robustness: the key point.

o Offline versus Real-time modeling: two different worlds.

o Where and how can multiple-view recording or surface modeling be helpful?

§ In academic research.

§ In industrial processes.

o Delivering it to the customer.

4. [1635-1720] Motion capture at Industrial Light and Magic: current practice and remaining challenges, Steve Sullivan, Lucasfilm [45 mins]

o Traditional marker-based mocap

o Onset body mocap

o Facial mocap “in the chair” and on-set

o Camera capture

The speakers

· Andrew Stoddart is head of Research at Vicon Motion Systems. He has a Ph.D. in theoretical particle physics from the University of Cape Town. In 1995 he became a lecturer in the Centre for Vision, Speech and Signal Processing at Surrey. In 1999 he joined OMG and was instrumental in the formation of the 2d3 Ltd. He was part of the boujou team that was awarded an Emmy in 2002 by the Academy for Television Arts and Sciences. Andrew currently leads several research projects within Vicon.

· Steve Caulkin is head of Research at Image Metrics. He completed his PhD in computer vision applied to digital mammography in 2001 and joined Image Metrics, working on a range of application areas before focusing on facial animation. He is currently responsible for research on Image Metrics’ performance-driven facial animation technology and related projects.

· Clement Menier is head of Research and Development at 4D View Solutions. He obtained his Ph.D. in vision systems in 2007 from the French INRIA Institute in Grenoble. His work on real time modeling and interactivity was selected for the Emerging Technology section of SIGGRAPH'07. Since 2007, he has been working on developing the technical offering of 4D View Solutions. Clement is currently in charge of the different research and development projects at 4D View Solutions.

· Steve Sullivan is Director of Research and Development at Lucasfilm. Steve joined Industrial Light & Magic in 1998 to develop computer vision techniques for visual effects production. Working within ILM's R&D group, he led a team creating systems for matchmoving, photogrammetry, image-based rendering, and motion capture. He has worked on over 50 films, and in 2002 received a Technical Achievement Award from the Academy of Motion Picture Arts and Sciences for the MARS motion and structure recovery system. Later in 2002, he became Director of Research and Development, and in 2004 initiated an effort to unify ILM film technologies and LucasArts game technologies in a common framework for all the Lucas companies. In 2007, Steve received a Scientific and Engineering Award from the Academy of Motion Picture Arts and Sciences for ILM’s Image-based Modeling system. He currently leads the R&D effort for all the Lucas companies, creating advanced artist technology for film, games, animation, and television. Prior to ILM, Sullivan received his PhD in Electrical and Computer Engineering from the University of Illinois at Urbana-Champaign in 1996, with an emphasis on automatic object modeling, recognition, and surface representations. After graduation, Sullivan spent two years with Rhythm & Hues Studios in Los Angeles, developing animation and 3D tracking software before joining ILM.