Abstract:Anomaly Detection is a relevant problem that arises in numerous real-world applications, especially when dealing with images. However, there has been little research for this task in the Continual Learning setting. In this work, we introduce a novel approach called SCALE (SCALing is Enough) to perform Compressed Replay in a framework for Anomaly Detection in Continual Learning setting. The proposed technique scales and compresses the original images using a Super Resolution model which, to the best of our knowledge, is studied for the first time in the Continual Learning setting. SCALE can achieve a high level of compression while maintaining a high level of image reconstruction quality. In conjunction with other Anomaly Detection approaches, it can achieve optimal results. To validate the proposed approach, we use a real-world dataset of images with pixel-based anomalies, with the scope to provide a reliable benchmark for Anomaly Detection in the context of Continual Learning, serving as a foundation for further advancements in the field.

What problem does this paper attempt to address?

### Problems the paper attempts to solve This article explores the problem of pixel - level anomaly detection (PLAD) in the continual learning (CL) setting. Specifically, the paper aims to solve the following two main problems: 1. **Performance degradation caused by data distribution changes**: - In real - world applications, the distribution of input data may change over time, which will lead to a significant decline in the performance of existing models. For example, new products may be added in an industrial environment, and the model needs to be able to detect the defects of new products while still being able to correctly detect the defects of old products. 2. **Catastrophic forgetting in continual learning**: - Neural networks are prone to forget previously learned knowledge when learning new tasks, a phenomenon known as catastrophic forgetting (CF). To solve this problem, this article studies how to effectively adapt to new data in the continual learning setting while maintaining knowledge of old data. ### Solutions To address the above problems, the paper proposes the following methods: 1. **Replay method**: - Use the Replay method to prevent catastrophic forgetting. The Replay method maintains knowledge of old tasks by storing a portion of old data and combining this old data with new data when training new tasks. 2. **Compressed Replay**: - In response to the problem that the Replay method requires additional memory, the paper proposes a new compressed replay method called SCALE (SCALing is Enough). The SCALE method significantly reduces memory usage by performing high - magnification compression on stored images through super - resolution technology (SR). 3. **Experimental verification**: - Use the MVTec anomaly detection dataset (MVTec AD Dataset) for experiments to evaluate the performance of different methods in the continual learning setting. The MVTec dataset contains high - resolution images of various industrial products, and each object corresponds to a learning task. ### Main contributions 1. **Studied the pixel - level anomaly detection problem in the continual learning setting**. 2. **Implemented multiple anomaly detection techniques and adapted them to the continual learning setting**. 3. **Proposed a new compressed replay method SCALE**, which can maintain high - quality image reconstruction while significantly reducing memory consumption. Through these methods, the paper provides an effective solution for solving anomaly detection problems in the real world, especially in cases of data distribution changes and resource limitations.

Continual Learning Approaches for Anomaly Detection

Lifelong Continual Learning for Anomaly Detection: New Challenges, Perspectives, and Insights

Multi-Scale Video Anomaly Detection by Multi-Grained Spatio-Temporal Representation Learning

Augment to Detect Anomalies with Continuous Labelling

Context-aware Feature Reconstruction for Class-Incremental Anomaly Detection and Localization

Look At Me, No Replay! SurpriseNet: Anomaly Detection Inspired Class Incremental Learning

Anomalies cannot materialize or vanish out of thin air: A hierarchical multiple instance learning with position-scale awareness for video anomaly detection

Continuous image anomaly detection based on contrastive lifelong learning

Anomaly detection in surveillance videos: a thematic taxonomy of deep models, review and performance analysis

Learning Not to Reconstruct Anomalies

Autoencoder-based Anomaly Detection in Streaming Data with Incremental Learning and Concept Drift Adaptation

Deep Anomaly Detection by Residual Adaptation

Continual Learning for Class- and Domain-Incremental Semantic Segmentation

Continual Learning for Unsupervised Anomaly Detection in Continuous Auditing of Financial Accounting Data

Continual Novelty Detection

Continual Object Detection: A review of definitions, strategies, and challenges

Continual Learning of Unsupervised Monocular Depth from Videos

Decoupled appearance and motion learning for efficient anomaly detection in surveillance video

Deep evolving semi-supervised anomaly detection

A preliminary study on continual learning in computer vision using Kolmogorov-Arnold Networks

On The Relationship between Visual Anomaly-free and Anomalous Representations