Zero-Shot Anomaly Detection via Batch Normalization
Anomaly detection (AD) plays a crucial role in many safety-critical application domains. The challenge of adapting an anomaly detector to drift in the normal data distribution, especially when no training data is available for the "new normal," has led to the development of zero-shot AD techniques. In this paper, we propose a simple yet effective method called Adaptive Centered Representations (ACR) for zero-shot batch-level AD. Our approach trains off-the-shelf deep anomaly detectors (such as deep SVDD) to adapt to a set of inter-related training data distributions in combination with batch normalization, enabling automatic zero-shot generalization for unseen AD tasks. This simple recipe, batch normalization plus meta-training, is a highly effective and versatile tool. Our theoretical results guarantee the zero-shot generalization for unseen AD tasks; our empirical results demonstrate the first zero-shot AD results for tabular data and outperform existing methods in zero-shot anomaly detection and segmentation on image data from specialized domains. Code is at https://github.com/aodongli/zero-shot-ad-via-batch-norm
PDF Abstract NeurIPS 2023 PDF NeurIPS 2023 Abstract
MNIST
MVTecAD
AnoShift
| Task | Dataset | Model | Metric Name | Metric Value | Global Rank | Benchmark |
|---|---|---|---|---|---|---|
| Unsupervised Anomaly Detection | AnoShift | ACR-NTL (zero-shot, test anomaly ratio=1%) | ROC-AUC FAR | 62.5 | # 1 | |
| Unsupervised Anomaly Detection | AnoShift | ACR-DSVDD (zero-shot, anomaly ratio=20%) | ROC-AUC FAR | 59.1 | # 4 | |
| Unsupervised Anomaly Detection | AnoShift | ACR-NTL (zero-shot, test anomaly ratio=20%) | ROC-AUC FAR | 62 | # 2 | |
| Unsupervised Anomaly Detection | AnoShift | ACR-DSVDD (zero-shot, anomaly ratio=1%) | ROC-AUC FAR | 62 | # 2 | |
| Anomaly Detection | MVTec AD | ACR (zero-shot) | Detection AUROC | 85.8 | # 84 | |
| Segmentation AUROC | 92.5 | # 77 |
