Normality-Preserving Continual Industrial Anomaly Detection via Orthogonal LoRA Banks
Researchers have developed a new framework for continual industrial anomaly detection using diffusion models. This method addresses the challenges of historical normality prior drift and catastrophic forgetting by employing orthogonal LoRA banks. The proposed system effectively isolates and protects category-specific normality priors during sequential adaptation, outperforming existing state-of-the-art methods on benchmark datasets. AI
IMPACT Introduces a novel approach to continual learning for anomaly detection, potentially improving industrial quality control systems.