Daan de Geus

About

Daan de Geus is a researcher focused on advancing computer vision for autonomous systems, with a particular emphasis on semantic segmentation in challenging, real-world environments. His work bridges the critical gap between controlled training data and the unpredictable conditions encountered in the wild, addressing the fundamental problem of domain shift. In his highly cited 2023 study, "Empirical Generalization Study: Unsupervised Domain Adaptation vs. Domain Generalization Methods for Semantic Segmentation in the Wild," de Geus systematically compares two leading approaches—unsupervised domain adaptation and domain generalization—to determine which strategies best enable scene understanding models to perform reliably across diverse, unseen scenarios. This work, which has already garnered 19 citations, provides essential guidance for developing safer autonomous vehicles and mobile robots. By empirically evaluating how well these methods generalize beyond their training distributions, de Geus has made a notable contribution to making vision models more robust and practical for deployment in the unpredictable conditions of the real world.

Research Focus

Key Achievements

Top Papers

1. 1
   Empirical Generalization Study: Unsupervised Domain Adaptation vs. Domain Generalization Methods for Semantic Segmentation in the Wild

   19 citations · 2023

Key Collaborators

Contact & Links