This course is typically taught in winter semesters and develops theory and methods for dealing with uncertain observations and uncertain actions in the real world. It is based on the "Probabilistic Robotics" book by Sebastian Thrun, Wolfram Burgard and Dieter fox and discusses the fundamental concepts employed e.g. in self-driving cars.

This course is typically held in summer semesters. The goal is to understand the key concepts of geometric machine vision. How can we mathematically describe a camera, how can we relate images taken from different positions? As humans, the different perspectives of our two eys provide us with depth information, and even when moving with only one eye open, humans can perceive 3D scene layout. This class provides the key concepts employed for visually scanning objects for 3D printing, for AR glasses such as HoloLens or for 3D vision in robotics.

Classical rule- and logic-based computing paradigms are very successful in reasoning about well-defined settings that can be expressed in formal mathematical language. However, formal problem analysis can struggle with real world tasks such as distinguishing cats and dogs, a job that even small children can do easily. In this lecture we will investigate a data-centric perspective that starts from a huge number of examples rather than from formal rules and look into ways to represent, fuse, cluster, classify, analyze real-world data. Additionally, we will investigate caveats and problems and look at metrics that can characterize the performance of a method.

This master seminar is typically offered in the semester after the lecture 3D Computer Vision and also after Mobile Robotics. It provides the opportunity to discuss selected topics in vision and robotics more in depth. At the semester start students select one topic from a list (typically a seminal or recent publication), study this topic and provide a presentation. Please get in touch with us in case you are interested to participate.

In this master project students will gain practical experience in applying concepts from the 3D Computer Vision and Mobile Robotics classes (you should have attended one of them before). Typically, projects are addressed in teams of two with different subtopics to select for people who prefer (not) working with hardware. And: Nobody has to go on a ship or dive or even get their hands wet.  

We are glad to provide thesis topics, please get in touch! In the best case you will work on a topic that connects to a thesis of a PhD student of our group. Check out the team website, look at the publications and feel free to contact anybody. Ideally, you have already attended some of the above courses beforehand, such that you know already that you like the field and have some understanding. We encourage (and try to enable) students to also contribute to or create a publication from their thesis. Example thesis topics in the past semesters looked into theoretical and practical aspects of "Volumetric 3D Reconstruction of Plankton", "New Fiducial Markers for Absolute Pose Estimation", "Endoscope to CT Registration", "Inverse Rendering for Underwater Vision", "Evaluating Learned SLAM Approaches", "Equivariant Image Features", " Noise in Monte-Carlo-based Rendering", "Light Calibration and Tracking in Robotics", ....