14th Workshop "Theoretical Foundations of Computer Vision"
Statistical and Geometrical Approaches to Visual Motion Analysis
Schloß Dagstuhl/Wadern (near Saarbrücken/Germany)
July 13 - 18, 2008
and Alan Yuille (USA)
The preliminary programme is available now! Please check for changes every now and then (last update: 9th July ).
Note: Each slot is 30 minutes. To allow for long discussions we recommend to prepare a 20 minutes presentation to allow for 10 minutes discussion!
Motion analysis is central to both human and machine vision. It involves the interpretation of image data over time. It is crucial for a range of motion tasks such as obstacle detection, depth estimation, video analysis, scene interpretation, video compression and other applications. Motion analysis is difficult because it requires modeling the complicated relationships between the observed image data and the motion of objects and motion patterns (e.g. falling rain) in the visual scene.
This workshop is focused on critical aspects of motion analysis, including motion segmentation and the modeling of motion patterns. The aim is to gather researchers who are experts in the different motion tasks and in the different techniques used. These techniques include variational approaches, level set methods, probabilistic models, graph cut approaches, factorization techniques, and neural networks. All these techniques can be subsumed within statistical and geometrical frameworks. We will also involve experts in the study of human and primate vision. Primate visual systems are extremely sophisticated at processing motion so there is much to be learnt from studying them. In particular, we want to relate the computational models of primate visual systems to those developed for machine vision.
Another important component of the workshop is to develop datasets of image sequences with the associated motion groundtruth. These datasets can be used as benchmarks to compare the performance of motion analysis models. They can also be used as data to train statistical models of motion analysis. Datasets with groundtruth are being increasingly used in other aspects of machine vision but, at present, there are only very limited motion datasets (e.g. the Yosemite sequence).
We intend to make the seminar very interactive with plenty of time for discussion. The participants will be encouraged to exchange different modeling techniques and research experiences. We will also intend to identify the outstanding unsolved problems in motion analysis and determine strategies for solving them.
This seminar is intended to focus on these aspects, and to support establishments of future research collaborations. There will be an edited book following the seminar, and all seminar participants will be invited to contribute with chapters. The deadline for those chapters will be in September 2008 (allowing to incorporate results or ideas stimulated by the seminar), and submissions will be reviewed (as normal).
Previous Workshops "Theoretical Foundations of Computer Vision":
Human Motion - Understanding, Modeling, Capture and Animation: proceedings published in 2008 by Springer.
(ed. B. Rosenhahn, R. Klette and D. Metaxas)
MPI Informatics: last update: 06 Feb 2007