Self-Supervised Representation Learning for Visual Behavior Analysis and Synthesis (Talk)
- Prof. Dr. Björn Ommer (Full Professor)
- Computer Vision Group at Heidelberg University
Understanding objects and their behavior from images and videos is a difficult inverse problem. It requires learning a metric in image space that reflects object relations in real world. This metric learning problem calls for large volumes of training data. While images and videos are easily available, labels are not, thus motivating self-supervised metric and representation learning. Furthermore, I will present a widely applicable strategy based on deep reinforcement learning to improve the surrogate tasks underlying self-supervision. Thereafter, the talk will cover the learning of disentangled representations that explicitly separate different object characteristics. Our approach is based on an analysis-by-synthesis paradigm and can generate novel object instances with flexible changes to individual characteristics such as their appearance and pose. It nicely addresses diverse applications in human and animal behavior analysis, a topic we have intensive collaboration on with neuroscientists. Time permitting, I will discuss the disentangling of representations from a wider perspective including novel strategies to image stylization and new strategies for regularization of the latent space of generator networks.
Biography: Björn Ommer is a full professor for Scientific Computing and leads the Computer Vision Group at Heidelberg University. He has studied computer science together with physics as a minor subject at the University of Bonn, Germany. His diploma (~M.Sc.) thesis focused on visual grouping based on perceptual organization and compositionality. After that he pursued his doctoral studies at ETH Zurich Switzerland in the Pattern Analysis and Machine Learning Group headed by Joachim M. Buhmann. He received his Ph.D. degree from ETH Zurich in 2007 for his dissertation "Learning the Compositional Nature of Objects for Visual Recognition" which was awarded the ETH Medal. Thereafter, Björn held a post-doc position in the Computer Vision Group of Jitendra Malik at UC Berkeley. He serves as an associate editor for the journal IEEE T-PAMI and previously for Pattern Recognition Letters. Björn is one of the directors of the HCI and of the IWR, principle investigator in the research training group 1653 ("Spatio/Temporal Graphical Models and Applications in Image Analysis"), and a member of the executive board and scientific committee of the Heidelberg Graduate School HGS MathComp. He has received the Outstanding Reviewer Award at ICCV'15, CVPR'14, ICCV'13, CVPR'11, and CVPR'10 and has served as Area Chair for ECCV'18. Björn has organized the 2011 DAGM Workshop on Unsolved Problems in Pattern Recognition.
