Visual Intelligence Online Seminar #78: FM4CS: A Versatile Foundation Model for Earth Observation Climate and Society Applications
Presenter: Arnt-Børre Salberg, Chief Research Scientist, Dept. for Image Analysis, Machine Learning and Earth Observation, Norwegian Computing Center
As deep learning transforms earth observation (EO) analysis, foundation models offer a promising alternative to traditional supervised learning by addressing data labeling challenges through large-scale, self-supervised learning.
The FM4CS model, developed for the European Space Agency, is a versatile multimodal foundation model tailored for climate and society EO applications. It supports four different Sentinel sensors: Sentinel-1 SAR, Sentinel-2 MSI, Sentinel-3 OLCI, and Sentinel-3 SLSTR, with resolutions ranging from 10 m to 1000 m. Evaluations across various benchmark EO tasks demonstrate FM4CS's robustness and adaptability, establishing it as a strong foundation for diverse EO applications.