19/5 - Computational Problems Associated with the Production of 3D Animated Features Films

Speaker: Tasso Lappas
Computer Science Colloquium

Date: 19/05/08
Place: IFW A 36
Time: 16:15
Language: English
Host: Petros Koumoutsakos
 

Animated features have become an important part of the annual output of Hollywood Studios over the last fifteen years. If one includes the large number of live action films with extensive computer generated visual effects, then these films represent the majority of annual blockbuster movies and therefore the largest percentage of collected revenue that range in the billions of dollars.
A great deal has been publicized about the variety of computer graphic techniques used to produce the imagery in these films ranging from animated characters and their interactions with real actors to the more extensive physical simulations used to produce water, explosions, fire and many other realistic effects. A lot has also been said about the creative challenges faced by directors and producers.
What has not been talked about much is the computational challenge associated with the production process that generates full 3D animated features in their entirety. These films are typically features, at least 90 minutes long, and are entirely a product of pure computation on large computing clusters with little or no live action footage provided by standard photography. The production environment represents a complex adaptive system of the order of 350 users and 3000 computing nodes generating more than 50 TB of data over a period of 2-3 years.

In this talk, a brief description of a standard production pipeline is given as context, in order to motivate the variety of interesting computational problems present in such a complex production environment. These problems live at the juncture of many standard areas of interest such as Concurrency, Optimization, Control, and Complex Adaptive Systems. At every step of the way Computational Complexity and NP-hard type problems appear and thwart attempts to optimize and streamline this process. The ultimate goal is to reduce the cost or time of production while steadily increasing the quality of the imagery.

Bio
Tasso Lappas been working in Computer Graphics for the Animation and Visual Effects film industry since 1994. He has worked in large studios such as Walt Disney Pictures and Sony Imageworks as well as in smaller production facilities in supervisory positions, ranging from head of R&D to Director of Technology.
While at Disney he supervised the software development of the non-photorealistic rendering software known as "Deep Canvas". For this work he is a recipient of an Academy of Motion Pictures Arts and Sciences Technical Achievement Award for 2002.
He holds a PhD in Aeronautics with emphasis on Computational Fluid Mechanics from the California Institute of Technology (1993).
Currently he works as an independent consultant to production studios and technolgy companies developing products for such production facilities.