Abstract

In this talk, I will present an overview of our long-standing research on digital twinning for complex indoor environments. I will trace the development of our indoor mapping systems—from early research prototypes to robust commercial solutions—and show how these technologies enable accurate, scalable, and efficient digitization of real-world spaces. Building on this foundation, I will introduce our recent advances in semantic segmentation, data completion, and 3D visual grounding. These methods transform raw, unstructured 3D scans into semantically enriched, interactive digital twins that support a wide range of downstream tasks. I will further highlight the role of digital twins in 6G radio-frequency propagation simulation, discuss how radar sensing can enhance digital twin fidelity, and outline how haptic sensing solutions can complement visual data by providing material-level information.

Bio

Eckehard Steinbach is a Professor at the Technical University of Munich (TUM), where he leads the Chair of Media Technology. His research spans visual and haptic information processing, machine learning, and robot perception. Over the past two decades, his group has made significant contributions to large-scale indoor mapping and digital twinning, with core research results successfully transferred into commercial solutions. He is also a co-founder and Chief Scientist of Olive Robotics GmbH. Prof. Steinbach is a Fellow of the IEEE.

Abstract
Emotions are often considered reactions to events, which are then appraised by people who live through this event. At the same time, emotions also constitute events. This raises two perspectives on events and emotions. One is language-processing specific and builds on top of semantic role labeling, namely emotion role labeling, where the task is to detect who is described to feel an emotion, which emotion that is, and why. The other requires the interpretation of events. Here, I discuss the computational use of appraisal theories that explain the parameters that lead to a particular emotion. The use cases include argument and multimodal social media post interpretations.

Bio

Roman Klinger is a professor at the Faculty for Information Systems and Applied Computer Science (WIAI) at the University of Bamberg. He studied computer science with a minor in psychology, holds a Ph.D. in computer science from TU Dortmund University (2011), and received the venia legendi in computer science in Stuttgart (2020). Before moving to Bamberg, he worked at the Institute for Natural Language Processing in Stuttgart, at the University of Bielefeld, at the Fraunhofer Institute for Algorithms and Scientific Computing, and the University of Massachusetts Amherst. Roman Klinger’s vision is to enable computers to understand and generate text regarding both propositional and non-propositional information. This finds application in interdisciplinary research, including biomedical text mining, digital humanities, modelling psychological concepts (like emotions) in language, and social media mining. These topics often constitute novel challenges to existing machine learning methods. Therefore, he and his group also contribute to the fields of probabilistic and deep machine learning.