/semiDiscreteCurvling

This toolbox allows to compute the 2-Wasserstein distance between a 2D images and a set of Dirac masses. Moreover it encompasses curvling constraints w.r.t. the Dirac masses position.

Primary LanguageJupyter NotebookMIT LicenseMIT

semi-discrete optimal transport for Curvling


DOI

This codes are the implementation of the following paper. It allows to compute the exact optimal transport between a continuous measure (with density) and measure carried by a set of Dirac masses. This toolbox allows the calculation in 2D of the optimal Transport distance. Moreover it includes Curvling constraints, that is imposing to the Diracs masses to be taken along a curve with a bounded speed and acceleration. We provide hands-on tutorials on the wiki.

The codes are released for Linux platforms (tested for Mint 18 Cinnamon 64-bit and Ubuntu 19.04 Disco Dingo).

The back-end computations are coded in C++ and make use of the computational geometry library CGAL and the linear algebra library Eigen3. We provide a python 3.7 interface by using the wrapper swig.

Authors

This software was developed by:

  • Frédéric de Gournay
  • Jonas Kahn
  • Alban Gossard
  • Pierre Weiss
  • Léo Lebrat

All members of the Toulouse institute of Mathematics France, granted by the ANR-17-CE23-0013.

License

This software is open source distributed under license MIT.