Federal Grant Award Summary

This collaborative research project, funded by the National Science Foundation's Division of Information and Intelligent Systems under the Computer and Information Science and Engineering program (CFDA 47.070), develops an artificial intelligence-driven framework enabling networked cameras to work cooperatively for improved video analytics. Awarded to George Mason University on July 15, 2026, with $269,000 obligated and completion targeted for June 30, 2029, the project addresses limitations in existing cross-camera video systems by creating a solution that functions under bandwidth and computing constraints. The research delivers three primary technical innovations: an asymmetric cross-camera video codec supporting hybrid offline and online training to adapt to fluctuating network bandwidth; a unified inference framework featuring networked submodels that share features, schedule appropriate cameras, and support diverse analytics tasks; and a reactive system adaptation mechanism through adaptive buffer management that determines which video frames to retain or discard at the server.

Conducted collaboratively between George Mason University (lead institution in Fairfax, Virginia) and the University of Illinois Urbana-Champaign, this initiative creates practical solutions for real-world deployments across transportation infrastructure, commercial facilities, buildings, and laboratory environments where multiple cameras must provide integrated situational awareness. Rather than relying on hand-crafted cross-camera rules, the project integrates learned cross-camera dependencies into an end-to-end system pipeline, enabling fast decision-making and improved accuracy even when underlying networks and analytics servers operate at capacity.