From mechanical friction simulation to special effects for the cinema
The modelling and dynamic visual representation of frictional mechanical contacts: this is the subject of the thesis completed by Gilles Daviet within the Bipop project team - work for which he has just received the 2017 thesis first prize awarded by French Association for Computer Graphics, the French Association for Virtual Reality and theEurographicsFrench Chapter. Today, the young researcher is working on simulation tools aimed at special effects for the cinema, in New Zealand.
You have just received an award for the thesis you did within the Bipop team at Inria. What was the subject of this thesis?
I started off at Inria as a research engineer with Florence Bertails-Descoubes (Bipop team), in computer graphics. There I worked on the simulation of virtual hair. The aim was mainly to simulate the effects due to contacts between hairs, and in particular frictional phenomena. Processing these interactions in such an in-depth way was quite new at the time, and by taking them into account we were able to truly turn a corner with regard to the realism of the simulations produced.
This is the subject I wanted to develop further during my thesis. However, rather than simulating these contacts one by one, we wanted to find a corresponding macroscopic model that we could process within the scope of the mechanics of continuous environments. The idea being that as the complexity of the simulation no longer depended on the number of components, we could achieve significant gains in computing time. My thesis mainly focused on the simulation of granular materials, such as, for example, sand. These materials are also widely governed by contacts and friction between the grains; moreover, they have the advantage of being relatively homogeneous, thereby making them suitable for simulation using continuous models.
Since this thesis, you have been working in New Zealand, atWeta Digital, a company specialising in special effects for the cinema. What does your work consist of?
I work as a researcher, I do research and development. More specifically, we are developing simulation software that will be used to create realistic special effects. For my part, I do not therefore directly produce images, but the software programs I am working on are intended for use by the artists who will apply them in order to produce certain film scenes.
How did you end up at Weta Digital?
I've known of this company by name for a long time. TheLord of the Ringstrilogy - whose special effects were done byWeta- had left its mark on me. After my engineering studies, I got into computer graphics somewhat by accident - I was mainly looking to work on the simulation of physical phenomena, and the subject offered by the Bipop team at Inria fitted in perfectly. So I applied and I was successful. After a year in New Zealand, I came back to do my thesis at Inria Grenoble. I then returned toWetain 2016.
Since your first post as research engineer at Inria, has the applicative focus of your work always been the cinema and special effects?
Yes, it's something that I like. I enjoy getting an immediate visual result, compared to other fields that are a bit more theoretical, and seeing your work on the big screen is really satisfying. However, we can distinguish between two types of applications of computer graphics for the cinema: on the one hand, animation films, where the physical side is very controlled, a caricature even, and contributes primarily to the narration. On the other hand,live actionfilms, where we try to have a physical aspect that is very close to reality, as invisible as possible. This is the side of special effects that interests me in particular, and during my thesis we tried to work on methods that are not solely aimed at entertainment; the idea was that the mechanics community could also use our simulators to make predictions.
You have been back and forth between the world of business and Inria several times. Do you foresee a return to academic research?
I like the industrial side, notably because we do not need to look for problems, they genuinely exist. We are in a concrete situation with users, feedback, etc. And we are in cycles that are quite fast-moving. I do however like to be able to reflect on a subject in a slightly more fundamental way, with fewer constraints with regard to time or success. So a return to research afterwards - why not?
This year Inria is celebrating its 50 years. What, in your opinion, will be the major challenges in computer graphics over the next 50 years?
For the time being, in order to carry out simulation in computer graphics we start with a model and we try to reproduce it by computer. The result is that we are restricted not only by the available computing power, but also by the hypotheses intrinsic to the model, and the quality of the parameters we provide it with. Recently, a new approach has been developing. It consists in no longer programming the model itself, but working it out directly using machines - along with its parameters. This is something that is still in its infancy in our field, but which should continue to grow quite rapidly, especially as we have a very simple metric to assess the quality of a simulation for the cinema: feedback from the supervisor. As it result, I think it is likely that in the future we will depend less on pre-designed models and will be working more based directly on observations.