FAST-EO 🌍 Foundation Models for Earth Observation

FAST-EO is a European project funded by ESA, focused on developing open-source multimodal foundation models for Earth Observation. Our models are designed to handle multispectral and mulitmodal data, enabling new capabilities in geospatial AI. Find more information on our webiste: https://www.fast-eo.eu.

🧠 Featured Models

• TerraMind: A generative, any-to-any multimodal foundation model for EO, co-developed with ESA Φ-lab.
• MS-CLIP: A contrastive vision language model for multispectral EO data.

📚 Publications

• TerraMind – ICCV 2025
• MS-CLIP – ECML 2025
• TerraMesh dataset – CVPR EarthVision 2025
• SmallMinesDS dataset – IEEE Geoscience and Remote Sensing Letters
• HyBiomass dataset – IEEE Geoscience and Remote Sensing Letters
• HYPERVIEW dataset – IEEE Geoscience and Remote Sensing Magazine

see all

🤝 Partners

FAST-EO is funded by ESA and supported by a consortium of leading organizations:

• German Aerospace Center (DLR)
• ESA Φ-lab
• IBM Research
• Forschungszentrum Jülich
• KP Labs

🌐 Open Collaboration

We are committed to open science, transparent benchmarking, and community-driven innovation. Explore our models, datasets, and tools — and help shape the future of geospatial AI.