Rating is available when the video has been rented.
This feature is not available right now. Please try again later.
Published on Jun 26, 2014
Ecole Polytechnique Fédédale de Lausanne (EPFL), Lausanne, Switzerland Institute of Electrical Engineering (EE) EE Distinguished Speaker Seminar
Title: Reconstruction meets recognition Prof. Marc Pollefeys, Dept. of Computer Science, ETH Zurich
Abstract: The two main goals of computer vision are to recognize the different elements of a scene and to reconstruct their 3D shape and spatial arrangement from images. In the past the reconstruction and recognition problems have mostly been approached separately. In recent years a few approaches have been proposed that use recognition to help reconstruction or vice-versa. In this talk we present the first approach that jointly performs 3D scene reconstruction and recognition. Our approach is formulated as a volumetric multi-class segmentation problem and solved using a convex relaxation method. A key element of our approach is that the interface between classes (e.g. building, ground, vegetation, air) have different anisotropic smoothness priors. We will show how our joint approach significantly improves the results.
Short bio: Marc Pollefeys is a full professor in the Dept. of Computer Science of ETH Zurich since 2007. Before that he was on the faculty at the University of North Carolina at Chapel Hill. He obtained his PhD from the KU Leuven in Belgium in 1999. His main area of research is computer vision, but he is also active in robotics, machine learning and computer graphics. Dr. Pollefeys has received several prizes for his research, including a Marr prize, an NSF CAREER award, a Packard Fellowship and a European Research Council Starting Grant. He is the author or co-author of more than 200 peer-reviewed publications. He is the general chair of ECCV 2014 in Zurich.