Meta's SAM 3 and DINOv3 Drive Automated Image Analysis at DOE National Labs

Meta's open-source models SAM 3 and DINOv3 power the SYNAPS-I pipeline at Lawrence Berkeley, Argonne, Brookhaven, and Oak Ridge. The system automates scientific image segmentation — separating objects within images — that once took experts weeks per dataset. Upgraded detectors now capture 100,000 images per second, generating tens of petabytes yearly.