Return to RESEARCH

WP1. Shape representation

The objective of this work package is to revisit shape representations from a user-centered perspective, which is totally new in Computer Graphics: the goal is to design the best possible representation to satisfy the user needs in terms of natural behavior during design and animation. Contrary to main-stream graphics research, this leads us to focus on high-level representations. However, we need to totally depart from the existing models in the area: as we have shown section 2.a, we need representations that separately handle morphology versus placement of a 3D shape. Since they will bring a natural response to both design gestures and animation, we call them “responsive shapes”.

Scientific problem: Morphology stands for the intrinsic parameters of a shape. Some are extensive such as the amount of constitutive material, others are intensive such as the structural distribution of material in space; placement is the current embedding of a shape in the 3D space. This embedding, to be used in animation and possibly during design to put a shape into a desired posture, may be isometric (i.e. preserve the geodesic distances between all pairs of points) or not, depending if the constitutive material can stretch and compress. We are looking for representations for 1D, 2D and 3D shapes enabling to separately describe the extensive and intensive morphological parameters, plus to store the current placement of a shape.

Our first results :

The first part of our work was devoted to volumetric (3D) implicit shapes. We developed a methodology for adding geometric details to skeleton-based implicit surfaces [Zanni 2012], and introduced SCALIS, a skeleton-based shape representation where blending behaves the same at any scale [Zanni 2013]. We also contributed to a general method, called gradient blending, to solve many long-lasting problems in the blending of implicit surfaces [Gourmel 2013]. This method was used to develop a novel skinning method for virtual characters, called implicit skinning [Vaillant 2013]. Lastly, we started working on models for shape distributions over other shapes in the specific case of generating settlement distributions on arbitrary terrains, which is a good, practical example of responsive shapes [Emilien, 2012].