Continual-MEGA: A Large-scale Benchmark for Generalizable Continual Anomaly Detection

2025-06-01

AI Summary

This work addresses the combined challenge of continual learning and anomaly detection in realistic scenarios, where models must adapt to sequential tasks while detecting unseen anomalies without labeled examples. Method: We introduce Continual-MEGA—the first benchmark for continual zero-shot anomaly detection—along with a novel multi-source dataset, ContinualAD. We formally define and evaluate how continual adaptation enhances zero-shot generalization, proposing the “continual adaptation-enhanced zero-shot generalization” paradigm, and design a unified baseline algorithm balancing few-shot robustness and cross-task generalization. Results: Experiments reveal significant bottlenecks of existing methods in pixel-level defect localization. Our approach comprehensively outperforms state-of-the-art methods. ContinualAD substantially improves strong models’ generalization to unseen anomaly classes. The benchmark establishes a standardized evaluation protocol and scalable infrastructure for continual anomaly detection.

Technology Category

• Machine Learning: Life-Long and Continual Learning
• Computer Vision: Adversarial Attacks & Robustness
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

In this paper, we introduce a new benchmark for continual learning in anomaly detection, aimed at better reflecting real-world deployment scenarios. Our benchmark, Continual-MEGA, includes a large and diverse dataset that significantly expands existing evaluation settings by combining carefully curated existing datasets with our newly proposed dataset, ContinualAD. In addition to standard continual learning with expanded quantity, we propose a novel scenario that measures zero-shot generalization to unseen classes, those not observed during continual adaptation. This setting poses a new problem setting that continual adaptation also enhances zero-shot performance. We also present a unified baseline algorithm that improves robustness in few-shot detection and maintains strong generalization. Through extensive evaluations, we report three key findings: (1) existing methods show substantial room for improvement, particularly in pixel-level defect localization; (2) our proposed method consistently outperforms prior approaches; and (3) the newly introduced ContinualAD dataset enhances the performance of strong anomaly detection models. We release the benchmark and code in https://github.com/Continual-Mega/Continual-Mega.

Problem

• Introduces a benchmark for continual anomaly detection in real-world scenarios
• Proposes zero-shot generalization to unseen classes during continual adaptation
• Develops a unified baseline algorithm for robust few-shot detection

Innovation

• Large diverse dataset for continual learning
• Zero-shot generalization to unseen classes
• Unified baseline algorithm for robustness