Recent Advances in Geospatial Machine Learning

Recent Advances in Geospatial Machine Learning 21 January 2026 by by Loïc Landrieu École des ponts ParisTech Geospatial machine learning has emerged as a compelling testing ground for modern computer vision and machine learning algorithms. In this talk, I presented recent progress in massively multimodal self-supervised learning for Earth observation, including large-scale pretraining across heterogeneous sensor modalities, unified representations for optical, radar, and elevation data, and scalable training strategies. I then discussed our latest work on visual geolocation, covering both standardized benchmarking and diffusion-based generative models that output spatial probability maps. Webinar sponsored by IEEE GRSS Image Analysis and Data Fusion (IADF) Technical Committee Learn more: https://www.grss-ieee.org/event/recent-advances-in-geospatial-machine-learning/