About cookies on this site Our websites require some cookies to function properly (required). In addition, other cookies may be used with your consent to analyze site usage, improve the user experience and for advertising. For more information, please review your options. By visiting our website, you agree to our processing of information as described in IBM’sprivacy statement. To provide a smooth navigation, your cookie preferences will be shared across the IBM web domains listed here.
Publication
JpGU 2024
Poster
An Introduction to Geospatial Foundation Model
Abstract
The increasing global concerns regarding climate change and environmental sustainability highlight the urgent necessity for effective solutions. The development of a foundational model for analyzing geospatial satellite images emerges as one such solution. We are currently engaged in the collaborative creation of an open-source geospatial foundational model with NASA. This endeavor entails pre-training large transformer models using multispectral remote-sensing image datasets sourced from diverse channels. Our foundational model demonstrates state-of-the-art performance, requiring fewer labeled training examples and exhibiting robust generalization capabilities across geographical regions. During the session, we will present recent advancements in the geospatial foundational model initiative, including techniques for global region data sampling, the latest experiments in fusing multisatellite images, and downstream applications such as flood detection in Japan.