I am a researcher in the 3D Computer Graphics area. My research plans focus on developing high level 3D geometrical models to manipulate virtual contents. I am especially interested in using 3D models to help the comprehension of complex phenomenon.
I am performing my research in CG as a member of the Imagine team, a join team of Inria Grenoble and Laboratoire Jean Kuntzmann (LJK), CNRS, and as a member of CPE Lyon. I am also a member of the Hevea project in collaboration with Institut Camille Jordan, GIPSA Lab and LJK on complex mathematical surface visualization. I am a former member of the Lawrence Berkeley National Lab (LBNL) on medical data visualization.
I have applied 3D models for Content Creation, Animation and Visualization.
1. Model for Content Creation
Design: Sketching developable surface
In order to help artists to enable digital garment design using traditional sketching pipeline, we investigate the generation of a 3D developable model and its 2D patterns from sketches depicting the object. In 2015, we proposed a method enabling to generate 3D developable surface with folds from a set of 2 orthogonal side views and currently investigate similar modeling from a single view.
Medical: Modeling human embryology
This project aims at helping understanding the 3D human embryology development using a 3D animated model. An anatomical ontology stores the organ architecture, their approximate geometry, and the rules guiding their development. During the visualization, the 3D model is procedurally generated in interpreting the rules queried in the ontology.
CG: Deforming complex objects
Complex 3D objects, such as assemblies, liquids, or model with fine details are particularly tedious to deform as the consistency of them may depend on complex parameters, and on different layers interacting together. We proposed different methods able to ease such deformation. First, we developed a method enabling to stretch and compress a 3D surface containing geometrical details. We also proposed an approach enabling to synthesize new assemblies from existing one in analyzing the graph of connection of the different subparts. More generally, we proposed a general framework, called Deformation Grammar, enabling to deform an object defined hierarchically, while maintaining its consistency. Finally, we investigated fluid animation and proposed an algorithm enabling to copy and past part of the animation in space and time.
- - CGF 2017, Deformation Grammar: Hierarchical Constraints Preservation under Deformation.
- - MIG 2016, Space-time sculpting of liquid animation.
- - SMI 2015, Real-Time Continuous Self Replicating Details for Shape Deformation.
- - EUROGRAPHICS 2015, Replaceable Substructures for Efficient Part-Based Modeling.
CAD: Generating functional model from scans.
Within the PhD of P. Coves, we develop a method to generate a usable CAD model from laser scan acquisitions of real mechanical pieces.
2. Model for Animation
Garments: Geometrical wrinkles
We developed a method adding geometrical wrinkles on top of an existing coarse cloth animation. This method enable to keep a very fast simulation, while preserving the fine cloth details during the animation.
The goal is to help digital artists to create efficiently a variety of detailed 3D shapes from an original mesh.
Characters: Improving skinning deformation
We developed several ways of improving the skinning deformation method. One is to ensure a constant volume of the shape through the deformation creating bulges when contracted. Another one is the use of the implicit skinning improving the articulation shape and ensuring a precise contact when the surface self-collide.
The goal is to ease the task of the digital artists animating surfaces using skinning techniques.
Developable surfaces: Animating crumpled paper interactively
We developed a new geometrical model to efficiently interact with paper-like surface. The animation is a mix between a simulation on a very sparse mesh and a set of geometrical rules ensuring the developability of the surface and the creation of singular points when needed. This is the first interactive crumpled paper surface, and it may be applied to efficiently model paper surfaces in CG. We further extended this model with interactive sound generation, and surface tearing.
3. Model for Visualization
Mathematics: Paradoxal surfaces
We developed a new geometrical model to efficiently interact with paper-like surface. The animation is a mix between a simulation on a very sparse mesh and a set of geometrical rules ensuring the developability of the surface and the creation of singular points when needed. This is the first interactive crumpled paper surface, and it may be applied to efficiently model paper surfaces in CG.The images of the flat torus have been used for instance in the following publications
- - Front Cover of PNAS 2012, Flat torus in 3D.
- - Pour la Science 2013, Les fractales lisses, un nouvel objet mathématique.
- - La Recherche 2012 & 2013, Le tore plat carré visualisé grace à l'informatique.
- - Science et Vie 2012, Le tore plat n'a plus de secrets.
Medical: Cardiac fibers
We proposed a visualization of the 3D fiber and sheet structure within the normal and diseased heart. This visualization aims at helping physicians and surgeons to understand the organization of the heart and reorganization of the fiber structure after infarct.