Geometric & Visual Computing (GeoViC) is a Computer Graphics and Computer Vision group tackling geometric analysis, modeling and computer animation research.
Note: GeoViC is the continuation of the STREAM team since November 2019
Our approach favors the introduction of efficient structures for high-level, possibly abstract, geometrical features of the studied shapes or phenomena. These structures can either be extracted from existing data or synthesized from prior knowledge and mathematical modeling.
Examples of such structures span latent space obtained from geometric-aware convolutional neural networks, distributions of shape arrangements and their associated metric, intrinsic properties on manifolds such as local-symmetry or Gaussian curvature, non-manifold skeletal structures representing volumetric implicit shapes, visual surface features associated with details or wrinkles on a parent shape, multi-resolution representations of complex shapes, as well as space-time features expressing user-guided or simulated interactions such as contacts and shape deformations.
These structures offer an efficient computational representation highlighting meaningful properties of the digital model and can be used to improve geometry processing algorithms, to provide support for expressive modeling or visualization methods, or to enable efficient and controllable animation.
The results of our research lead to fast and robust virtual content analysis and creation methods such as shape matching and feature retrieval, creative prototyping and modeling of detailed objects or characters that automatically adapt to the context and the user interactions, or either animated natural scenes composed of multiple layers coupling geometrical constraints with efficient simulation. The targeted applications include entertainment & artistic creation (animation cinema, video games, VFX), virtual prototyping & manufacturing, as well as modeling & visualization for natural sciences.
We got nominated for the Best Paper at ICCV 2019 for the article Unsupervised Deep Learning for Structured Shape Matching Jean-Michel Roufosse, Abhishek Sharma, Maks Ovsjanikov. There was 7 papers nominated out of 1075 accepted papers, and 4303 submissions.
We received the Best Presentation Award at MIG 2019 for the article Animation Synthesis Triggered by Vocal Mimics, Adrien Nivaggioli, Damien Rohmer.
- Damien Rohmer is program co-chair of the ACM SIGGRAPH / EUROGRAPHICS Symposium on Computer Animation - SCA - 2018, with Maud Marchal (IRISA INSA). SCA took place on 11-13th of July in Paris.
- Pooran Memari and Maks Ovsjanikov, are program co-chair of the EUROGRAPHICS Symposium on Geometry Processing - SGP - 2018, with Tamy Boubekeur (Télécom ParisTech, LTCI). SGP took place on 9-11th of July in Paris.