Jane Edwards
Microsoft has received an other transaction agreement from the National Geospatial-Intelligence Agency to develop a trained foundational neural network model that can ingest ground-level and overhead imagery, vector data and terrain data and can answer geospatial questions with accuracy and precision.
According to an award notice published Thursday on SAM.gov, the potential $14 million OTA for the GEOINT Foundational Model development initiative was awarded through the Boosting Innovative GEOINT – Science and Technology broad agency announcement.
NGA issued a solicitation for the BAA in December.
The research topic covers GFM, geospatial artificial intelligence, multimodal machine learning, masked autoencoding, contrastive machine learning, cross-view image matching, vision transformers, neural networks, geographic information system and remote sensing.
The agency said the research’s goal is to deliver a geographically aware foundational model that can form components of a virtual assistant that could help analysts geolocate objects or media or answer questions requiring specific geospatial understanding.
The award has a one-year base period and three option years.
Register here to join the Potomac Officers Club’s 2024 Intel Summit on Sept. 19 and hear top U.S. intelligence community officials and industry executives discuss the challenges, innovation initiatives, opportunities and technologies shaping the future of American intelligence.
