U.S. Central Command and the Defense Department’s Chief Digital and Artificial Intelligence Office (CDAO) have partnered on a solicitation seeking to making it easier for users to apply artificial intelligence to imagery and video to identify threats.
The May 31 Commercial Solutions Opening (CSO) solicitation seeks demonstrations to solve two challenges related to the use of computer vision (CV) technology to identify threats in imagery, one being to update the models at the pace of the changing threat and the second to allow users to retrain the models, Schuyler Moore, chief technology officer at CENTCOM, posted on LinkedIn on Monday. She calls the solution “User-Driven Computer Vision.”
The CDAO wants submissions by June 30. A CSO allows an agency to quickly acquire capabilities—typically for prototyping—from non-traditional defense contractors.
“The goal of this efforts is to enable DoD warfighters to create performant CV models in 7 days or less via an integrated, self-service platform,” the solicitation says. “If successful, this process will complement existing CV development programs of record focused on model performance optimization.”
Moore says that CV can be important in sorting through large datasets of imagery and video but the command “has sometimes struggled to see results” because the models take too long to retrain and revalidate in response to shifting threats, and because when models are built without keeping operators’ capabilities and networks in mind, application can fall short.
User-driven CV tools should be “low-code/no-code, self-service” so that the models easily integrate into workflows, she says, leading to “greater user adoptions and more technically literate user community that can better articulate gaps and needs going forward.”
CENTCOM and the CDAO are interested in CV models applied to imagery and video collected by spaceborne, airborne, and land- and sea-based sensors.