Abstract:Although mainstream unsupervised anomaly detection (AD) (including image-level classification and pixel-level segmentation)algorithms perform well in academic datasets, their performance is limited in practical application due to the ideal experimental setting of clean training data. Training with noisy data is an inevitable problem in real-world anomaly detection but is seldom discussed. This paper is the first to consider fully unsupervised industrial anomaly detection (i.e., unsupervised AD with noisy data). To solve this problem, we proposed memory-based unsupervised AD methods, SoftPatch and SoftPatch+, which efficiently denoise the data at the patch level. Noise discriminators are utilized to generate outlier scores for patch-level noise elimination before coreset construction. The scores are then stored in the memory bank to soften the anomaly detection boundary. Compared with existing methods, SoftPatch maintains a strong modeling ability of normal data and alleviates the overconfidence problem in coreset, and SoftPatch+ has more robust performance which is articularly useful in real-world industrial inspection scenarios with high levels of noise (from 10% to 40%). Comprehensive experiments conducted in diverse noise scenarios demonstrate that both SoftPatch and SoftPatch+ outperform the state-of-the-art AD methods on the MVTecAD, ViSA, and BTAD benchmarks. Furthermore, the performance of SoftPatch and SoftPatch+ is comparable to that of the noise-free methods in conventional unsupervised AD setting. The code of the proposed methods can be found at <a class="link-external link-https" href="https://github.com/TencentYoutuResearch/AnomalyDetection-SoftPatch" rel="external noopener nofollow">this https URL</a>.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is: in industrial scenarios, how to achieve fully unsupervised anomaly detection (AD) in the presence of noisy data, that is, not relying on clean training data for anomaly classification and segmentation. ### Problem Background Existing unsupervised anomaly detection methods perform well on academic datasets, but they usually assume that the training data is clean, which is not realistic in practical applications. The training data in actual industrial scenarios often contains noise (such as defective samples), which will lead to a decline in model performance. Therefore, it is of great significance to study how to perform effective anomaly detection on data containing noise. ### Paper Contributions 1. **Propose fully unsupervised industrial anomaly detection for the first time**: - Existing unsupervised AD methods rely too much on the cleanliness of training data, and their performance will decline significantly under noise interference. This paper focuses on fully unsupervised industrial anomaly detection for the first time, which is a more practical but rarely studied area. 2. **Propose a patch - level denoising strategy**: - In order to improve data utilization and reduce the impact of noise on the model, this paper proposes a patch - level denoising strategy. Specifically, the SoftPatch and SoftPatch+ methods identify and exclude abnormal patches by assigning an outlier factor to each patch. Compared with traditional image - level denoising, this method can better handle local anomalies. 3. **Introduce multiple discriminators to enhance robustness**: - Considering the limitations of a single discriminator under a high noise ratio, SoftPatch+ further introduces a multi - discriminator fusion method, thereby improving the model's robustness and performance, especially in cases where the noise ratio is high (from 10% to 40%). 4. **Establish benchmarks and verify effectiveness**: - This paper conducts extensive experiments in various noise scenarios. The results show that SoftPatch and SoftPatch+ outperform the existing state - of - the - art AD methods on the MVTecAD, VisA, and BTAD benchmarks, and their performance is comparable to that of methods in traditional noise - free environments. ### Method Overview - **Feature Extraction**: Use a pre - trained backbone network to extract image features. - **Core Set Construction**: Select normal patches through a denoising algorithm to construct a core set. - **Anomaly Detection**: In the test phase, calculate the anomaly score through nearest - neighbor search, and re - weight the anomaly score according to the soft weight to determine the final anomaly detection result. Through these improvements, the SoftPatch and SoftPatch+ methods proposed in this paper can perform anomaly detection more effectively in the presence of noise, providing a new solution for industrial quality inspection.

SoftPatch+: Fully Unsupervised Anomaly Classification and Segmentation

SoftPatch: Unsupervised Anomaly Detection with Noisy Data

Semi-supervised Noise-Resilient Anomaly Detection with Feature Autoencoder

AnomalySD: Few-Shot Multi-Class Anomaly Detection with Stable Diffusion Model

AnomalyNCD: Towards Novel Anomaly Class Discovery in Industrial Scenarios

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

A Diffusion-Based Framework for Multi-Class Anomaly Detection

Mixed Noise-Guided Mutual Constraint Framework for Unsupervised Anomaly Detection in Smart Industries

FUN-AD: Fully Unsupervised Learning for Anomaly Detection with Noisy Training Data

Patch SVDD: Patch-Level SVDD for Anomaly Detection and Segmentation

Patch-wise Auto-Encoder for Visual Anomaly Detection

Attention-Guided Perturbation for Unsupervised Image Anomaly Detection

Neural Architecture Search for Visual Anomaly Segmentation

AnomalySeg: Deep Learning-Based Fast Anomaly Segmentation Approach for Surface Defect Detection

Towards Efficient Pixel Labeling for Industrial Anomaly Detection and Localization

Unsupervised Defect Detection with Patch-Aware Mutual Reasoning Network in Image Data

MAPL: Memory Augmentation and Pseudo-Labeling for Semi-Supervised Anomaly Detection

A Comprehensive Library for Benchmarking Multi-class Visual Anomaly Detection

Generating and Reweighting Dense Contrastive Patterns for Unsupervised Anomaly Detection

AnomalousPatchCore: Exploring the Use of Anomalous Samples in Industrial Anomaly Detection