Coral imaging AI reaches operational testing with NOAA

University of Miami researchers developed PICOGRAM, an open-source AI system that measures coral colonies from underwater photographs, and NOAA has tested it at an operational scale. NASA assessed the project at Application Readiness Level 8, indicating validation in an operational setting. The system generates its own training masks from image characteristics, reducing the need for scientists to outline corals manually. In tests using 137 photographs, its coral-cover estimates differed from expert measurements by an average of 1.2 percentage points. Overlap with expert outlines fell from 87.5% under familiar conditions to 82.5% at unfamiliar reef sites; scientists can review and correct uncertain predictions.