Florence Bertails-Descoubes: From a real object to its virtual avatar

You are a member of the BiPoP project team at Inria Grenoble-Rhône-Alpes. What is your project all about?

At the present time, we have relatively realistic motion simulators for mechanical systems with similar complexities to clothing and hair. However, actually simulating any given real object is very difficult. The main difficulty relates to the configuration of the simulator. You have to take account of the stiffness of the object, its mass, its shape at rest, its coefficient of friction, and many more different parameters.

My project uses an innovative inverse method to estimate these parameters simply by observing the shape of the real object. The underlying idea is to extract information about the mechanical properties of the object from purely geometric data.

The results of this research will help me in my work on the modelling of complex structures, including the assemblies of flexible fibres and deformable surfaces that are found in hair and clothing. I am using image capture and 3D reconstruction tools to make accurate observations of the real object.

By combining a series of static poses, I believe that it should then be possible to deduce the mechanical properties of the object to a high degree of accuracy, taking full account of all contacts and friction. Using just the observed shape of the object, and without any invasive measurement processes, we should be able to obtain a complete characterisation of the object and predict its future movements as a function of the forces acting on it.

Finally, I want to carry out experiments on real models in order to confirm the validity of these virtual models.

Why are contacts and friction so important in the modelling process?

Simply because they play such an important role in achieving realistic simulations. In graphical computing, this realism equates to an increased level of visual richness. The volume of the objects is conserved and the threshold effect, typical of friction between solids, is captured. 

To take hair as an example, one of the most complex systems to model. If we want to create a scene in which the heroine’s hair slides over her shoulders before settling naturally into position while at all times maintaining its original volume, it is essential to take account of the contacts and friction between the individual hairs, and between each hair and the character’s body.

These problems have formed the basis of my work since I joined the BiPoP team in 2007.
Together with my students, we have developed new dynamic fibre models and efficient solving tools to simulate the frictional contact.

Applying our results to the creation of artificial images has also enabled me to combine two of my great interests; mathematics and graphic art.

I have been lucky enough to see our results taken up by a number of animation studios, including the French start-up Néomis Animation, a specialist in virtual hair styling, and the New Zealand company Weta Digital, one of the world leaders in the production of special effects. Weta Digital was responsible for creating the fur on some of the characters in The Hobbit using our solution algorithms for frictional contact.

I also regularly transfer the results of my research to leading cosmetic companies such as L’Oréal.

How are you going to make use of this ERC grant?

As well as the satisfaction of seeing my project recognised, the financial aspect is crucial. With a grant of 1.5 million euros over five years, I will be able to recruit three PhD students, two post-doctoral researchers and one engineer.

I will also be able to expand my contacts with experts in computer vision and 3D reconstruction, and with both theoretical and experimental physicists and mechanical engineers.

Without this grant, it would be difficult for me to set out on a multidisciplinary journey of that extent.