Mattéo Couplet

I am a postdoctoral researcher in the Computer Graphics Lab at Boston University, working with Prof. Edward Chien. I am interested in the generation of structured meshes for engineering and graphics applications, and, lately, how it connects to computational fabrication problems like knitting and 3D printing.

I did my PhD at UCLouvain in Belgium, where I was advised by Prof. Jean-François Remacle.

I am grateful to be supported by the Belgian American Educational Foundation and Wallonie-Bruxelles International.

Email  /  CV  /  Scholar

Publications

We place knit singularities with semidiscrete optimal transport of a curl measure. With principled control over the singular foliation, we produce globally optimal knit graphs in dense settings.

The synchronization problem minimizes the deviation of a field relative to a connection. It models a variety of problems in geometry processing, and we propose a unified method to solve them.

Integrability can be formulated for 3-dimensional odeco frames. This enables a surface quadrilateral mesher with control over size, orientation, and distortion metrics.

Extending our integrable frame fields work, an end-to-end quad mesher for planar geometries that complies with sizing prescriptions on boundary and feature curves.

Vector field curl indicates where to place knit singularities. With this insight we develop a fast knit graph generation method, with guaranteed manufacturability.

Odeco tensors are a natural representation for frame fields. We find new energies that make these fields integrable, with automatic creation and placement of singularities.

Building upon Gmsh's quasi-structured quadrilateral mesher, we compute coarse quad layouts through integer optimization.

Physically accurate 3D representations for porous media can be reconstructed from 2D slices by training deep neural networks on texture synthesis tasks.

Website template from Jon Barron.