Monday 24, April – 13:30-16:30

Room: Rhône 4

Motion synthesis

Author: Du, Han
Affiliation: Saarland University

Simulated reality is widely used in industry, entertainment and education. In many application areas, there is a growing demand for efficiently generating natural-looking character animation for simulated realities from high-level semantic description. For instance, the simulation of workers’ assembly operations in shop-floor based on semantic task description, or automatically translating text to realistic sign language animation for disabled people. Motion capture technology can provide high-quality movement data for animation. Therefore, data-driven motion synthesis approaches are prevailing for generating realistic motions. However, most data-driven approaches usually require considerable manual efforts for preprocessing motion capture data. It is also challenging to adapt motion data to different constraints or scenarios efficiently. The focus of this work is to create a general framework to model motion capture data and create new animations based on high-level semantic description.

Polycubes Simplification, Optimisation and Remeshing

Author: Cherchi, Gianmarco
Affiliation: University of Cagliari

Generating structured volumetric and surface representations of 3D objects is today a relevant problem in several applications. They are input for methods like FEM (Finite Element Methods) or IGA (IsoGeometric Analysis) but also for applications like Animation, Gaming or Biomedical fields. In the last years, important progresses in the meshing research community have been made. Several methods and algorithms have been proposed, including the recent parameterisation-based techniques. Among these, polycube-based approaches are particularly interesting. The meshing pipeline via polycube has a lot of highlights and it allows us to easily create well structured representations. This approach also presents some limits like singularities misalignments, self-intersections or poor control of the generated meshes quality. I think that, considering the interest of the community about this topics, could be worth to face and solve these problems. I solved some of them in the first period of my PhD course, while at the moment I am studying how to head and working out remaining problems.

Proceduralization of Urban Models

Author: Demir, Ilke
Affiliation: Purdue University

The technological developments in the last century have carried us from a few pixels per screen to infinite worlds. Consecutively, recent bottleneck in graphics industry has been changing from the means to the objects, in other words, from the technology to the content: creating the content generation problem. Procedural models are known to be an effective solution to create massive amount of content from powerful and compact parameterized representations. While procedural modeling provides compelling architectural structures, creating detailed and realistic models needs time and extensive coding as well as significant domain expertise. Observing the content creation problem and the challenges of procedural modeling, we identified and developed a set of proceduralization tools, that convert existing models into an easy to manipulate procedural form and allow quick synthesis of visually similar objects. The central idea of our research is that we can automate and assist modeling and creation tasks by proceduralization of existing models such as architectural meshes, building point clouds, or textured urban areas. Our research applies techniques from computer graphics, computer vision, and geometry processing to address the challenge of proceduralization. We develop solutions to fundamental problems such as segmentation, pattern discovery, and procedural representation for the aforementioned structures. These solutions enable useful tools such as a structure-aware procedural editing engine that can create diverse models with similar style, or a reconstruction system to exploit the grammar for having more complete models of components, or a localization framework that can organize a city for feature-based search.

Scalar Fields and Secondary Motion in Character Animation

Author: Roussellet, Valentin
Affiliation: IRIT

Producing convincing animated characters in computed-generated imagery ultimately depends upon the generation of realistic deformation of the character’s skin. Skin deformation, or skinning, is controlled by a set of intuitive handles known as a rig, which frequently takes the form of an animation skeleton.

Skinning is often understood by animators as the result of primary and secondary motions. Primary motion, directly created by the animation skeleton, is the main driver of the skin’s deformation. However, the production of compelling skin behavior requires the addition of convincing secondary motion: elements deforming as a reaction to the character’s movements such as body parts jiggling, muscle bulging, wrinkles forming on the skin, hair or cloth floating, etc.

Generating secondary motion for a character remains a challenging problem, given the variety of characters and scenarios in computer animation. Simplistic approaches fail to deliver the appropriate results, while computationally-intensive algorithms are too slow for interactive use and hamper the workflow of animators. Skin deformation methods which focus too much on realism may also conflict with artistic requirements to model believable imaginary creatures or exaggerated featuresThe goal of this research project is to provide a consistent framework which can produce believable secondary motion to animated characters. This framework should be able to integrate to existing animation pipelines, enabling animators to augment a character with secondary motion.

Data driven sketch interpretation

Author: Delanoy, Johanna
Affiliation: INRIA Sophia-Antipolis

Sketch-based modeling strives to bring the ease and immediacy of drawing to the 3D world. However, while drawings are easy to create by humans, they are very challenging to interpret by computers due to their sparsity and ambiguity. As a result, most existing systems tackle the problem of 3D reconstruction from drawings by making restrictive assumptions on shape and drawing techniques or by requiring extensive user interactions. We propose to use data driven techniques, which have the potential to adapt to a larger variety of styles, to create a sketch-based modeling tool capable of adapting to various drawing techniques and styles, rather than imposing a particular domain to users.

Fabrication of Simplified Shapes using Digital CNC Machines

Author: Muntoni, Alessandro
Affiliation: University of Cagliari

On my Ph.D. I am focusing my research on Fabrication of 3D shapes using subtractive techniques like 3-axis CNC milling or carving machines. These techniques cannot manufacture free-form geometries, and they have very precise constraints on the manufacturable models. My research consists of algorithmically processing free-form 3D models in order to obtain manufacturable shapes using these techniques. CNC Carving machines are able to work only on sheets of materials with a limited thickness, and for 3-axis milling machines every millable piece needs to be an height-field. On this Research Statement I’ll present a method for the simplification of 3D models in order to fabricate them with CNC Carving Machines, and I’ll introduce another project for the decomposition of 3D shapes in manufacturable blocks with 3-axis milling machines.

Semi Automatic Cage Construction

Author: Casti, Sara
Affiliation: University of Cagliari

Computer animators need simple tools and techniques that allow them to reduce their workload in order to realize animations. Simplified structures (points, skeletons and cages), called handles, have been designed in order to speed up animation pipeline. In cage-based deformation, the animators manipulate a coarse cage bounding the model to affect the space region inside the cage thus deforming the original model itself. Although points and skeletons design is quite simple, constructing a proper cage for the model is an expensive and time-consuming task because it is produced mainly by hand. We propose a semi-automatic approach to construct a bounding cage guided by its curve skeleton. We aim to generate an high quality cage, coarse enough to be manipulated easily, which catch all meaningful parts of the original model, thus providing a good abstraction of the shape for animation purposes.

Doctoral Consortium Chairs

Nicolas Bonneel (Univ. Lyon/LIRIS/CNRS, France)
Belen Masia (Univ. Zaragoza, Spain)