NASA’s Prithvi Becomes First AI Geospatial Foundation Model In Orbit

A team of researchers from Adelaide University and the SmartSat Cooperative Research Center deployed NASA and IBM’s Prithvi Geospatial AI foundation model on two in-orbit platforms: Kanyini satellite and IMAGIN-e payload aboard the International Space Station, making it the first geospatial foundation model to be deployed in orbit.
Prithvi, trained on 13 years’ worth of global geospatial data from NASA and ESA satellites, can facilitate tasks like flood and disaster monitoring.
The model’s performance for flood and cloud detection was tested across different orbiting platforms, with promising results shared in a preprint article.
The open-source nature of Prithvi saved researchers significant time and effort, exemplifying the benefits of making AI models available for collaborative advancements.
Foundation models like Prithvi, which are trained on vast amounts of unlabeled data, can be fine-tuned for specific tasks with less data, offering flexibility for diverse Earth observation applications.
The deployment showcases the potential for foundation models to revolutionize data analysis and interaction with instruments on satellites. NASA and IBM envision a future where operators can interact with satellites using natural language.
NASA continues to develop and release open-source foundation models, including a heliophysics model called Surya, and plans to expand to other scientific fields.

You may have missed