Seminar zu Machine Learning and Data Analytics

Üblicherweise erster Donnerstag im Monat, 16:00 Uhr, FZ.02.06, Gebäude FZ.

Bergische Universität Wuppertal, Campus Freudenberg, 42119 Wuppertal

Im Seminar werden aktuelle Forschungsthemen aus Wissenschaft- und Technikforschung in den Themengebieten maschinelles Lernen und Datenanalyse diskutiert.

Nächster Termin bereits am 30.06.2022.

Datum   Referent   Vortrag
07.04.2022   Thomas Thiele, Data Intelligence Center - House of AI, Deutsche Bahn AG   Heavy Metal meets AI – Künstliche Intelligenz in Mobilität und Logistik auf der Schiene
05.05.2022   Arnulf Jentzen, Chinese University of Hong Kong, Shenzhen & Universität Münster   Overcoming the Curse of Dimensionality: from nonlinear Monte Carlo to Deep Learning
02.06.2022   Siniša Šegvić, University of Zagreb   Elements of Learning Algorithms for Natural Scene Understanding
30.06.2022   Tim Fingscheidt, Technische Universität Braunschweig   From Synth to Real: An Overview of Domain Generalization and (Continuous) Unsupervised Domain Adaptation Methods for Semantic Segmentation
07.07.2022   Sebastian Stober, Otto-von-Guericke-Universität Magdeburg   Cognitive neuroscience inspired techniques for eXplainable AI (CogXAI)
         
Stand:   20.06.2022    
         
         

Abstract for the current seminar:

From Synth to Real: An Overview of Domain Generalization and (Continuous) Unsupervised Domain Adaptation Methods for Semantic Segmentation

Deep neural networks for automotive environment perception require training data, synthetic or real, and are then applied in various different real domains, i.e., in “reality”. Synthetic training data is attractive since it comes with free labels. There are several concepts to achieve robustness against such domain mismatch, including domain generalization (DG) and unsupervised domain adaptation (UDA). The talk selects semantic segmentation as example environment perception method and provides some bird's eye view onto common DG and UDA concepts, including source-free and continuous/continual UDA, followed by example simulations and some practical findings in the literature about such methods.

Weitere Infos über #UniWuppertal: