GeoAI at the CVPR Conference

Denver, June 2026

In June 2026, researchers from the project team at the University of Ljubljana, Faculty of Computer and Information Science (UL FRI)—Filip Wolf, Blaž Rolih, and Luka Čehovin Zajc—presented a new approach to building foundation models for Earth observation at CVPR (IEEE Conference on Computer Vision and Pattern Recognition 2026). CVPR is the leading conference in computer vision and machine learning, and it features an increasing number of papers on Earth observation.

The paper is open access:

• Wolf, F., Rolih, B., Čehovin Zajc, L. (2026). Brewing Stronger Features: Dual-Teacher Distillation for Multispectral Earth Observation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026, pp. 27815–27826. https://openaccess.thecvf.com/content/CVPR2026/html/Wolf_Brewing_Stronger_Features_Dual-Teacher_Distillation_for_Multispectral_Earth_Observation_CVPR_2026_paper.html (and on GitHub: https://github.com/wolfilip/DEO-FM)

At the conference, the researchers also participated in the MORSE workshop (https://sites.google.com/view/cvpr-morse), where they also presented new deep-learning approaches to change detection.