SuperADD: Training-free Class-agnostic Anomaly Segmentation -- CVPR 2026 VAND 4.0 Workshop Challenge Industrial Track

2026-05-14

AI Summary

This work addresses the challenge of distribution shifts in industrial visual anomaly detection caused by variations in acquisition conditions such as illumination. The authors propose a training-free, category-agnostic method for robust anomaly segmentation, building upon SuperAD with a unified architecture and shared hyperparameters across all object categories to eliminate class-specific design. Key innovations include the adoption of a DINOv3 backbone, an overlapping patch partitioning strategy, intensity augmentation, an improved memory bank sampling mechanism, and iterative morphological closing to produce spatially coherent anomaly maps. Evaluated on the MVTec AD 2 dataset, the method achieves F1 scores of 62.61%, 57.42%, and 54.35% on public test, private test, and mixed private test splits, respectively, substantially outperforming existing approaches and demonstrating strong suitability for rapid deployment in industrial settings.

Abstract

Visual anomaly detection (AD) for industrial inspection is a highly relevant task in modern production environments. The problem becomes particularly challenging when training and deployment data differ due to changes in acquisition conditions during production. In the VAND 4.0 Industrial Track, models must remain robust under distribution shifts such as varying illumination and their performance is assessed on the MVTec AD 2 dataset. To address this setting, we propose a training-free and class-agnostic anomaly detection pipeline based on the work of SuperAD. Our approach improves generalization through several modifications designed to enhance robustness under distribution shifts. These adaptations include using a DINOv3 backbone, overlapping patch-wise processing, intensity-based augmentations, improved memory-bank subsampling for better coverage of the data distribution, and iterative morphological closing for cleaner and more spatially consistent anomaly maps. Unlike methods that rely on class-specific architectures or per-class hyperparameter tuning, our method uses a single architecture and one shared hyperparameter configuration across all object classes. This makes the approach well suited for industrial deployment, where product variants and appearance changes must be handled with minimal adaptation effort. We achieve segmentation F1 scores of $62.61\%$, $57.42\%$, and $54.35\%$ on test public, private, and private mixed of MVTec AD 2 respectively, thereby outperforming SuperAD and other state-of-the-art methods. Code is available at https://github.com/LukasRoom/SuperADD.

Problem

• visual anomaly detection
• distribution shift
• class-agnostic
• industrial inspection
• anomaly segmentation

Innovation

• training-free
• class-agnostic
• distribution shift robustness
• DINOv3
• anomaly segmentation