Campus Event Calendar
Event Entry
What and Who
Generalizations of the Multicut Problem for Computer Vision
Evgeny Levinkov
Max-Planck-Institut für Informatik - D2
Promotionskolloquium
AG 1, AG 3, AG 4, RG1, MMCI, AG 2, AG 5, SWS
Public Audience
English
Note: We use this to send email in the morning.
Date, Time and Location
Monday, 8 April 2019
11:30
60 Minutes
E1 4
024
Saarbrücken
Abstract
Abstract: Graph decompositions have always been an essential part of computer vision as a large variety of tasks can be formulated in this framework. One way to optimize for a decomposition of a graph is called the multicut problem, also known as correlation clustering. Its particular feature is that the number of clusters is not required to be known a priori, but is rather deduced from the optimal solution. On the downside, it is NP-hard to solve.
In this thesis we present several generalizations of the multicut problem, that allow to model richer dependencies between nodes of a graph. We also propose efficient local search algorithms, that in practice find good solution in reasonable time, and show applications to image segmentation, multi-human pose estimation, multi-object tracking and many others.
Contact
Connie Balzert
0681 9325 2000
--email hidden
passcode not visible
logged in users only
Connie Balzert, 04/01/2019 09:40
Connie Balzert, 03/26/2019 12:13 -- Created document.
Connie Balzert, 03/26/2019 12:13 -- Created document.