L2S Talk by Matthias Neuwirth-Trapp (Bosch Research/ETH Zürich): Lifelong Detection – Enabling Models to Continuously Learn New Objects

Mon 27 April 2026, 16.00 h in H-B 6414

Object detection is usually studied in a static regime, although deployed systems evolve over time. As environments shift and new categories become relevant, detectors must adapt sequentially while preserving prior knowledge. In the presentation we dive into why this matters for Bosch, where the difficulties are, what current approaches are and the insight we gained into the problem.

Matthias Neuwirth-Trapp is a PhD student at Bosch Research and ETH Zürich). His thesis is supervised by Luc Van Gool (INSAIT, former ETH Zürich) and Konrad Schindler (ETH Zürich).

Tags
Jan Söhlke
Jan Söhlke

Dr. Jan Söhlke is the head of communication and staff photographer at ZESS, as well as the Scientific Coordinator for the DFG Research Unit 'Learning to Sense' (FOR 5336).

Following his doctoral studies at LMU Munich, he moved into science communication and the visual documentation of research environments. His work focuses on photographing complex scientific setups and high-tech infrastructure - translating engineering and academic projects into clear visual assets. In addition, he works as a freelance photographer for industrial and research-driven organizations. You can find his portfolio at https://jansoehlke.com/.

Articles: 128

Related Posts