C-UAS DETECTION MODELING WITH MINIMAL REAL DATA

See how engineers used Rendered.ai’s Platform as a Service (PaaS) to train a Detectron2 model to accurately identify drones in scenes with complex backgrounds and lighting.  Only 10% of a real training dataset was needed to establish a reliable C-UAS detection model.

• Model accuracy jumped to 85% from 77% in just 14 days.
• 90% less reliance on real data to train a performant C-UAS algorithm.
• Difficult-to-acquire training imagery was simulated effectively in just a few days at a much lower cost.