Skill by nvidia

tao-generate-od-defects

Run AnomalyGenNext 1.1 inference from a native testcase or a prepared generation plan, then publish validated fine-grained and binary COCO labels. Use for synthetic object-detection defect generation, not model training.