------------------------------------------------------------------------------ B O S T O N U N I V E R S I T Y Computer Science Department C O L L O Q U I U M Thursday, May 1, 1997 3:00 pm (Coffee served at 2:45 pm) Room MCS 148 ------------------------------------------------------------------------------ ANIMATION FOR THE REST OF US Michael Gleicher (formerly of Apple Research Laboratories) With computer animation, anything is possible: desk lamps jump, tigers change into cars, and dinosaurs rip open sport-utility vehicles. However, the time, skill, and talent required to create animation means that very few can express themselves in the media. Just as a better paintbrush does not eliminate the need for art classes, simply refining existing computer animation tools and techniques is not enough. In this talk, I will describe some projects that aim to make it easier for the rest of us to create animation and special effects. In each, we recast some part of an animation task as a non-linear constrained optimization problem, so we can use the machine to do some of the work. All of these projects were done in the graphics research group at Apple Computer: + Motion Adaptation and Editing: animated motion tends to be very special purpose and not reusable: it almost always is a specific character a specific action. We developed methods for repurposing motions, with the intent that it could enable clip-motion libraries. I will describe a constraint-based approach to motion adaptation/editing that attempts to preserve as much of the original motion as possible while creating a new motion that meets new needs. This requires solving a single (large) constrained optimization problem over the entire motion. However, with a bit of care, it is possible to solve these spacetime constraints problems fast enough to provide real-time motion editing, even on 3D motion-captured data. + Generating animation from performance: given observations of an actor (and not just a human actor) performing some motion, how do we make a graphical model perform the same action? This project developed ways to more automatically process motion capture data. + Projective tracking and registration: this work developed a method to watch groups of pixels as they moved from frame to frame in a video sequence. Because the technique determines a proper projective transformation between frames, the motion can be reconstructed. I will show how this can be used to create "virtual graffiti," where we can paint on one frame of a video, and have the changes propagated to later frames. Host: Prof. Stan Sclaroff ------------------------------------------------------------------------------ For colloquium info, including directions, see http://www.cs.bu.edu/colloquium For more information contact Prof. David Yates ------------------------------------------------------------------------------