Groups
Our research group is working on a range of topics in Computer Vision, Machine Learning and Artificial Intelligence. Computer Vision is about interpreting images. More specifically the goal is to infer properties of the observed world from an image or a collection of images. Our work combines a range of mathematical domains including statistical inference, differential geometry, continuous (partial differential equations) and discrete (graph-theoretic) optimization techniques.
Research and application areas include Visual SLAM, Image-based 3D Reconstruction, Shape Analysis, Deep Learning, Biomedicine, Robot Vision, RGB-D Vision, Image Segmentation, Convex Relaxation Methods.
The Dynamic Vision and Learning (DVL) group at TUM is working on key problems in Computer Vision and Machine Learning. The main goal of the research team is dynamic scene understanding, i.e., to allow robots to see and understand the world around them from visual (video) input. The team works with cutting-edge deep learning methods and embeds classic computer vision knowledge such as 3D geometry or graph theory.
Research areas include multiple object tracking, video object segmentation, pedestrian trajectory prediction, visual localization, change detection, image retrieval, video forensics and video anonymization.
The Smart Robotics Lab (SRL) focuses on enabling technologies for mobile robots operating in a potentially unknown environment. This includes localisation (without infrastructure such as GPS), mapping, and 3D scene understanding with a suite of sensors, most importantly cameras. Respective algorithms ranging from computer vision and machine learning to motion planning and control need to be processed efficiently on-board to yield accurate results in real-time. The aim is to empower the next generation of intelligent mobile robots to plan and execute complex tasks safely in potentially cluttered, and dynamic environments, possibly close to people.
Research areas include Visual SLAM, Multi-Sensor-Fusion, 3D Reconstruction, Machine Learning, Deep Learning, 3D Scene Understanding, Robot Navigation, Drones, (Aerial) Manipulation, Model-Predictive Control; and these are used in a range of application areas, e.g. Industrial Inspection, Robotic Construction, Disaster Response, and Environmental Monitoring.
Our research group works on computer vision and multi-modal learning. Specifically, we are interested in learning from multiple data modalities, such as vision, sound, and text.
Research areas include audio-visual learning, video understanding, cross-modal retrieval (text-audio, text-image, and text-video retrieval), self-supervised learning, and low-shot learning.
Machine Vision is an interdisciplinary field of research and engineering at the intersection of informatics, mathematics, physics, and electrical and mechanical engineering. It is used in dozens of application areas whenever there is a need to automate processes. Our research focuses on developing new methods that are suitable for the machine vision industry and on extending research results in such a way that the meet the stringent demands of the machine vision industry, e.g., with respect to robustness, accuracy, precision, and real-time capability. Application areas include alignment, metrology, inspection, identification, object recognition, and robot vision for industrial robots.
Research areas include modeling and calibration of imaging sensors, close-range photogrammetry, 3D reconstruction, 2D and 3D object detection, 6-DoF object pose estimation, deep learning, anomaly detection, and optical character recognition.