Abstract:In the industrial scenario, anomaly detection could locate but cannot classify anomalies. To complete their capability, we study to automatically discover and recognize visual classes of industrial anomalies. In terms of multi-class anomaly classification, previous methods cluster anomalies represented by frozen pre-trained models but often fail due to poor discrimination. Novel class discovery (NCD) has the potential to tackle this. However, it struggles with non-prominent and semantically weak anomalies that challenge network learning focus. To address these, we introduce AnomalyNCD, a multi-class anomaly classification framework compatible with existing anomaly detection methods. This framework learns anomaly-specific features and classifies anomalies in a self-supervised manner. Initially, a technique called Main Element Binarization (MEBin) is first designed, which segments primary anomaly regions into masks to alleviate the impact of incorrect detections on learning. Subsequently, we employ mask-guided contrastive representation learning to improve feature discrimination, which focuses network attention on isolated anomalous regions and reduces the confusion of erroneous inputs through re-corrected pseudo labels. Finally, to enable flexible classification at both region and image levels during inference, we develop a region merging strategy that determines the overall image category based on the classified anomaly regions. Our method outperforms the state-of-the-art works on the MVTec AD and MTD datasets. Compared with the current methods, AnomalyNCD combined with zero-shot anomaly detection method achieves a 10.8% $F_1$ gain, 8.8% NMI gain, and 9.5% ARI gain on MVTec AD, 12.8% $F_1$ gain, 5.7% NMI gain, and 10.8% ARI gain on MTD. The source code is available at <a class="link-external link-https" href="https://github.com/HUST-SLOW/AnomalyNCD" rel="external noopener nofollow">this https URL</a>.

Hard-normal Example-aware Template Mutual Matching for Industrial Anomaly Detection

Incremental Template Neighborhood Matching for 3D Anomaly Detection

MTHM: Self-Supervised Multitask Anomaly Detection with Hard Example Mining

Learning Discrimination from Contaminated Data: Multi-Instance Learning for Unsupervised Anomaly Detection

Template-guided Hierarchical Feature Restoration for Anomaly Detection

AnomalyNCD: Towards Novel Anomaly Class Discovery in Industrial Scenarios

A Two-Stage Anomaly Detection Framework: Towards Low Omission Rate in Industrial Vision Applications.

Pull & Push: Leveraging Differential Knowledge Distillation for Efficient Unsupervised Anomaly Detection and Localization

Enhancing Anomaly Detection Models for Industrial Applications through SVM-Based False Positive Classification

DMMGNet: A Discrimination Mapping and Memory Bank Mean Guidance-based Network for High-performance Few-shot Industrial Anomaly Detection

Anomaly Multi-classification in Industrial Scenarios: Transferring Few-shot Learning to a New Task

Fast Template Matching in Multi-modal Image under Pixel Distribution Mapping

Scale-Adaptive NN-Based Similarity for Robust Template Matching

Prior Normality Prompt Transformer for Multi-class Industrial Image Anomaly Detection

Soft matching network with application to defect inspection

PTMNet: Pixel-Text Matching Network for Zero-Shot Anomaly Detection

Self-Supervised Learning for Industrial Image Anomaly Detection by Simulating Anomalous Samples

Reconstruction-based Multi-Normal Prototypes Learning for Weakly Supervised Anomaly Detection

Ckt: cross-image knowledge transfer for texture anomaly detection

Advancing Pre-trained Teacher: Towards Robust Feature Discrepancy for Anomaly Detection

Divide and Conquer: High-Resolution Industrial Anomaly Detection via Memory Efficient Tiled Ensemble