Abstract:In this paper, we propose SPACE, a novel anomaly detection methodology that integrates a Feature Encoder (FE) into the structure of the Student-Teacher method. The proposed method has two key elements: Spatial Consistency regularization Loss (SCL) and Feature converter Module (FM). SCL prevents overfitting in student models by avoiding excessive imitation of the teacher model. Simultaneously, it facilitates the expansion of normal data features by steering clear of abnormal areas generated through data augmentation. This dual functionality ensures a robust boundary between normal and abnormal data. The FM prevents the learning of ambiguous information from the FE. This protects the learned features and enables more effective detection of structural and logical anomalies. Through these elements, SPACE is available to minimize the influence of the FE while integrating various data <a class="link-external link-http" href="http://augmentations.In" rel="external noopener nofollow">this http URL</a> this study, we evaluated the proposed method on the MVTec LOCO, MVTec AD, and VisA datasets. Experimental results, through qualitative evaluation, demonstrate the superiority of detection and efficiency of each module compared to state-of-the-art methods.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve the Anomaly Detection (AD) problem in industrial applications, especially when it is impractical to obtain abnormal samples for training. Specifically, the paper proposes a new method - **SPACE (SPAtial - aware Consistency rEgularization)** to address the following challenges: 1. **Limitations of data augmentation**: In industrial AD datasets, the abnormal area usually accounts for only a small part of the entire image, while the rest is normal. Therefore, strong data augmentation will make the normal data closer to the abnormal data instead of maintaining its normal characteristics. This makes existing methods based on strong augmentation difficult to effectively distinguish between normal and abnormal data. 2. **Over - fitting of the student model**: In the Student - Teacher (S - T) framework, the student model may over - imitate the teacher model, leading to over - fitting. This will reduce the ability to detect real anomalies. 3. **Detection of logical anomalies**: In addition to structural anomalies, there are also logical anomalies in industrial applications, that is, a single object may be normal, but the overall image level is abnormal. For example, in the push - nail dataset, there should be one nail in one position, but if there are more than one nail or no nail, it is considered a logical anomaly. To overcome these challenges, SPACE introduces two key components: - **Spatial Consistency regularization Loss (SCL)**: SCL prevents over - fitting by avoiding excessive imitation of the teacher model by the student model, and expands the boundaries of normal data by selectively learning features, thereby ensuring a robust boundary between normal and abnormal data. - **Feature converter Module (FM)**: FM prevents learning of ambiguous information from the feature encoder (FE), protects the learned features, and improves the effectiveness of structural and logical anomaly detection. Through these improvements, SPACE can more effectively detect structural and logical anomalies in industrial applications while using weakly - augmented and strongly - augmented data for training. ### Summary The paper proposes a new anomaly detection method SPACE, which aims to solve the challenges faced by anomaly detection in industrial applications, especially the limitations of data augmentation and the over - fitting problem of the student model. By introducing SCL and FM, SPACE can more effectively detect structural and logical anomalies and has achieved better performance than existing methods on multiple benchmark datasets.

SPACE: SPAtial-aware Consistency rEgularization for anomaly detection in Industrial applications

An Unsupervised Framework for Anomaly Detection in a Water Treatment System

Segmenting Remote Sensing Anomalies at Instance-level Via Anomaly Map Guided Adaptation

Structural Teacher-Student Normality Learning for Multi-Class Anomaly Detection and Localization

Masked feature regeneration based asymmetric student–teacher network for anomaly detection

A space-embedding strategy for anomaly detection in multivariate time series

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

Memory-Augmented Spatial-Temporal Consistency Network for Video Anomaly Detection.

MSTAD: A masked subspace-like transformer for multi-class anomaly detection

Spatiotemporal consistency-enhanced network for video anomaly detection

Enhancing Anomaly Detection in Pedestrian Walkways using Improved Sparrow Search Algorithm with Parallel Features Fusion Model

Abnormal events detection using spatio-temporal saliency descriptor and fuzzy representation analysis

A Non-Parametric Subspace Analysis Approach with Application to Anomaly Detection Ensembles

Multi-Class Anomaly Detection based on Regularized Discriminative Coupled hypersphere-based Feature Adaptation

HCL-MTSAD: Hierarchical Contrastive Consistency Learning for Accurate Detection of Industrial Multivariate Time Series Anomalies

Configurable Spatial-Temporal Hierarchical Analysis for Flexible Video Anomaly Detection

A novel and robust data anomaly detection framework using LAL-AdaBoost for structural health monitoring

Convolutional Recurrent Reconstructive Network for Spatiotemporal Anomaly Detection in Solder Paste Inspection

Multimodal data consistency anomaly detection based on sparse subspace clustering

Dual-path Frequency Discriminators for Few-shot Anomaly Detection

Self-supervised Feature Adaptation for 3D Industrial Anomaly Detection