The National Geospatial-Intelligence Agency (NGA) on Monday called on industry to provide data-labeling services to assist in training artificial intelligence (AI)-driven computer vision systems.
According to NGA officials, data labeling services procured under the Sequoia program’s indefinite delivery/indefinite quantity (ID/IQ) contract – valued at $708 million over a maximum of seven years – will enhance satellite imagery processing and improve the identification of critical targets.
AI-driven computer vision models help NGA analysts swiftly interpret the vast amounts of satellite imagery and geospatial data coming in from various sources, including commercial satellites. However, these models need to be trained first to accurately identify military targets and detect unusual activities.
Data-labeling – the methods used by analysts and experts to identify, tag, and categorize raw data – would provide context for training machine learning models.
During a Defense Writers Group meeting on Aug. 30, NGA Director Vice Adm. Frank Whitworth told reporters that in using these AI systems analyst face “this issue of distinction, guaranteeing to the best of our ability … the distinction between a combatant and non-combatant, an enemy and non-enemy, and that’s hard, and I will tell you, based on my 35-plus years of experience, especially in targeting, that’s one of the hardest things we do in targeting,” he said.
Whitworth added that the Sequoia contract is “the largest data labeling request for proposal in the U.S. government,” and “represents a significant investment in computer vision, machine learning and AI.”
According to NGA, the Sequoia IDIQ contract is intended to include data labeling activities to support geospatial intelligence (GEOINT) AI/ML capabilities across multiple programs and directorates within NGA, the National System for GEOINT, and the Department of Defense (DoD).
One of those programs is NGA’s Maven Program, which NGA took over from DoD in 2022. The Maven program applies AI capabilities to intelligence, surveillance, and reconnaissance sensors and platforms, primarily through computer vision.
“The NGA Maven [computer vision] algorithms, which rely heavily on robust data labeling, can perform a number of automated tasks, including but not limited to object detection, object tracking, object classification and pattern detection,” the announcement reads.
The contract also will support natural language processing, analytic models, and AI/ML models that support business process automation for the GEOINT mission. The contract will leverage commercial computer vision and AI capabilities to augment existing programs and will integrate directly into analytic workflows for operational use.
The full request for proposal is posted to the classified Intelligence Community Acquisition Research Center (ARC) website. The source selection process includes two phases, Phase One is an oral presentation and Phase Two is a written proposal. Notices of intent to participate in Phase one are due by 2 p.m. on Oct. 15 via the classified ARC. Dates for Phase Two will be announced following the completion of Phase One.
