NTT DATA Partners with Bifrost AI to Revolutionize Satellite AI Training

NTT DATA has embarked on a groundbreaking collaboration with U.S.-based startup Bifrost AI to transform AI model training for satellite imagery analysis. This partnership focuses on leveraging synthetic satellite data to overcome the traditional challenges of costly and time-consuming real-world data collection.

Using Bifrost AI’s synthetic data generation technology, NTT DATA trained detection models hundreds of times faster and approximately 70 percent cheaper compared to conventional methods. This accelerated timeline resulted in a nearly 10-point improvement in accuracy within weeks instead of months.

Synthetic image data—computer-generated yet highly realistic—allows AI models to “see” diverse scenarios, including rare or previously unseen environmental conditions, addressing gaps in actual satellite image datasets. This approach removes the dependency on large volumes of annotated real-world images, which are often expensive or difficult to obtain.

The proof-of-concept demonstrated efficacy in multiple use cases such as object detection and change detection on high-resolution satellite images, achieving increased mean average precision and F1 scores. Importantly, Bifrost AI’s platform produces annotated synthetic images requiring minimal additional manual labeling efforts.

NTT DATA envisions this technology powering “physical AI,” where machines interpret the world through sensors and cameras, comparable to human learning by observing visual reference materials. Applications include remote infrastructure monitoring, agricultural optimization, and disaster risk reduction without the need for frequent physical inspections.

The collaboration is part of NTT DATA’s broader AI strategy to enhance intelligent systems like satellites and sensors, unlocking new efficiencies and capabilities in Earth observation and beyond. By integrating synthetic data training, NTT DATA aims to scale AI deployments faster, more cost-effectively, and with higher accuracy, shaping the future of satellite-based AI solutions.