Cinema - R&D
Tony DeRose : « France is attractive for its mathematical sophistication »
On the 10th of February, the Sophia Antipolis Inria teams had the pleasure to welcome the Lead of the Research Group at Pixar Animation Studios, the American Tony DeRose. He presented his work in the Jacques Morgenstern colloquium and visited Inria. Short summary of his job and his impressions about France.
Can you tell us about Pixar studios, especially your research team?
There are 1200 people working in Pixar studios, and Pixar’s Research team is composed of six researchers, plus internships and post-docs. As far as I know, we’re the only company in computer animation to have a research department. It has been formed in 2004 because we needed to take risk, but without compromising one particular project. We look for risky things inside Pixar, but also inside Disney (about 20% of our effort). We try to stay informed and develop technologies in advance. It can take four years before our work is taken in a film.
Can you give us an up-to-date overview of R&D in cinema?
Well, it depends on how far you want to go back! I spent 15 years in Pixar. When I started, we did a terrible job on hair, we couldn’t render the complexity of environment. We had to invent most of the technologies we work with now. Today, characters move in more and more compelling ways, and costs of production have gone down.
What are the challenges for R&D in cinema ?
High impact areas for the research group are ligthing, simulation and interfaces.
Lighting is the latest job on a film. The lighting artist places the virtual light sources, then render is run to create the final imagery for the film. In the days of the original Toy Story, 15 years ago, when the digital lighter moved a light, he had to wait 4 hours to see the final imagery. Today, our render force is about 1000 times faster, but we’re putting 1000 times more complexity, so it still takes 4 hours. Our goal is to have this lighting process go in real time. It represents a speed-up of about 10000 times. We’ll close this big gap with algorithmic, engineering, and hardware contributions.
Regarding simulation, we continue to work on complex phenomena such as the way hair, smoke or clothing moves, grass waving in the wind, splashing…
Another field of research is interfaces. In Pixar, we write all our software. Artists use them ten hours a day, and they basically work with a mouse and a keyboard, sometimes a cintiq (a direct drawing device with a stylus). It is not ideal. So a couple of years ago, we started using an exotic but emerging technology : multitouch. We built our own large multitouch work station. One potential benefit is that the gesture vocabulary is very large. We’ve tested it and we’re now to the point where it is capable of producing quality results with about 20% speed-up over the traditional mouse and keyboard interfaces.
Why did you come to France?
Georges Drettakis’ Inria team works on topics closely related and complementary to ours. It isn't exactly what we do in the area of level-of-detail control, but we think we can combine what they do with what we have done and are planning to do, for example develop automatic level-of-detail control, to gradually increase the detail as the model grows larger on screen.
And France is attractive: it offers an important level of mathematical sophistication. Quality of research is as good as in good universities in the US. Internships seem to be more integrated into European education. We generally have two Europeans at a time working with us. As American students generally have lots of course requirements, which means they are often only available during the summer, we hire European students for winter and spring and for longer periods.
George Drettakis, leader of the project-team REVES:
« I met Tony DeRose for the preparation of the international venues presenting new research results in computer graphics and interactive techniques: the SIGGRAPH and SIGGRAPH Asia 2010 conference. We were both chairpersons for the respective technical papers programs and worked together.
As we have similar scientific interests, the Inria organizing committee invited him to France for the colloquium at Sophia Antipolis, in February. His presentation was very highly attended, with more than 300 students, researchers and industrials.
His visit to our team was an enriching experience for everybody, including the students. In our research team Reves, we follow what all members of Pixar Research Group publish and we have been inspired by their publications, in particular concerning rendering problems. We have developed novel procedural texturing techniques based on Gabor Noise which were inspired by work developed previously at PIXAR research. Our scientific challenges, notably in the domain of rendering and texturing, are complementary.
We showed Tony most of our recent and ongoing projects. He was particularly interested in followup work which uses Gabor Noise. »