Persistent and Autonomous Learning for Robot Vision
Place: E:1406, E-building, Ole Römers väg 3, LTH, Lund University
Contact: volker [dot] krueger [at] cs [dot] lth [dot] se
Save event to your calendar
Topic: Persistent and Autonomous Learning for Robot Vision
Speaker: Dr. Rudolph Triebel, robotics lab of DLR, the German Space Agency
Abstract: Building on the recent achievements in Machine Learning many powerful tools have been developed, mainly in computer vision but also many others. From the perspective of a researcher in robot perception, this is a unique opportunity to tackle challenging tasks that rely on vision such as object manipulation, semantic mapping and SLAM, or human-robot collaboration. However, it turns out that robotics applications establish at least three specific challenges to the vision system that are not solved by standard (deep) learning methods. These challenges can be summarized as real-time capability, autonomy from the human, and trustworthiness. To address these challenges, in my department at DLR we develop visual learning algorithms that learn persistently in the sense that offline and online learning methods are combined to achieve an efficient inference. Furthermore, we aim at autonomous learning methods that require less human supervision and enable decisions on what to learn from. And finally, we investigate introspective learning techniques that reduce the overconfidence of the prediction. In the talk, I will show examples of these ideas applied to object pose estimation, grasp planning, and object classification, and I will demonstrate how these design principles help to overcome the special challenges of robot vision tasks.
Bio: Dr. Triebel is the leader of the Department for Perception and Cognition at the Robotics Institute of the German Aerospace Center (DLR). He received his PhD in 2007 from the University of Freiburg, Germany, with a PhD thesis on “Three-dimensional Perception for Mobile Robots”. From 2007 to 2011, he was a postdoctoral researcher at ETH Zurich, where he worked on machine learning algorithms for robot perception within several EU-funded projects. Then, from 2011 to 2013 he worked in the Mobile Robotics Group at the University of Oxford, where he developed unsupervised and online learning techniques for detection and classification applications in mobile robotics and autonomous driving. Since 2013, Dr. Triebel works as a lecturer at TU Munich, where he teaches master level courses in the area of Machine Learning for Computer Vision.