Nos tutelles

CNRS Dauphine PSL *



12 octobre 2018: 1 événement


  • Séminaires du Pôle 3 : "Sciences des données"

    Vendredi 12 octobre 10:00-11:00 - Benjamin Perret - ESIEE

    Graph based hierarchical clustering for image analysis

    Résumé : Hierarchical clustering is a natural approach in image analysis : it provides a decomposition of a scene into objects and iteratively refines the objects into their parts. In this presentation, we first review how hierarchical clustering is used in practice on selected image analysis problems. In a second time, we focus on a particular hierarchical clustering approach, the watershed hierarchy, which corresponds to the solution of a combinatorial optimization problem on edge weighted graphs related to the minimum spanning tree problem. We will see that watershed hierarchies can be computed efficiently and achieve good performance in large computer vision benchmark. Finally, I will present some ongoing works and perspectives, like the application of hierarchal watershed to various graph data or the fusion of hierarchical clusterings.

    Lieu : C108

    En savoir plus : Séminaires du Pôle 3 : "Sciences des données"