PAIDF AnomalyGen generates labelled synthetic defect imagery for industrial inspection from a handful of real examples. It adapts the NVIDIA Cosmos generator into an anomaly-inpainting model: given a clean product image and a mask marking where a defect belongs, it paints a realistic defect of the requested class into that region. Only a small parameter subset is trained on top of a frozen base network, so fine-tuning is few-shot and the resulting checkpoints are small. The container covers automatic mask placement, fine-tuning, generation, KPI evaluation, quality refinement, and pseudo-labeling, producing training data for downstream defect detectors and segmenters.

Governing terms

This container is governed by the NVIDIA Software License Agreement and Product-Specific Terms for NVIDIA AI Products.