Géométrie Riemannienne pour l’analyse de la forme des visages 3D et les trajectoires des actions

Le jeudi 3 mars à 14h en salle 201, le Professeur Mohamed Daoudi, membre du Laboratoire d’Informatique Fondamentale de Lille, viendra faire un séminaire

Abstract :

In computer vision, shapes have been represented in many different ways: point clouds, surfaces, images or skeleton are only some examples. The difficulty comes from nonlinearity of these shape spaces. Indeed, these shape spaces are not Euclidean and one cannot perform classical statistics. One way to overcome this difficulty is to introduce a Riemannian structure on the shapes space. This enables us to exploit the geometry of these shape spaces and to develop efficient statistical tools. In this talk, I will show some recent works from our group on Riemannian geometry and its application in a variety of problems including recognition of faces, expressions and actions.