Rethinking Continual Anomaly Detection on the Edge: Benchmarking Under Realistic Industrial Conditions

2026-05-22

AI Summary

This study addresses critical limitations in existing continual anomaly detection methods—namely, inadequate evaluation protocols, lack of systematic comparisons, and poor suitability for edge deployment—rendering them ill-equipped for dynamically evolving industrial production conditions. To bridge this gap, the work establishes the first unified benchmark tailored to real-world industrial edge environments, encompassing structural and logical anomalies, continuous distribution shifts, cross-method comparisons, and hardware efficiency analysis. Building upon this foundation, the authors propose DINOSaur, a training-free approach that leverages a frozen DINOv2 backbone, a spatially indexed core-set memory mechanism, and a neighborhood-constrained scoring strategy to achieve zero forgetting and rapid online adaptation. Experiments demonstrate that DINOSaur consistently outperforms state-of-the-art methods across all five evaluation protocols, achieves inference latency under 100 ms, and adapts to new tasks within 30 seconds on a Jetson Orin Nano device.

Abstract

Continual anomaly detection (CAD) addresses the need for industrial inspection systems to adapt to evolving production conditions, yet existing methods share three critical gaps: unrealistic evaluation, no systematic comparison, and no consideration of edge deployment constraints. We introduce a unified benchmark combining discrete-task evaluation on structural and logical anomalies, a novel continuous drift protocol, the first head-to-head comparison of all published CAD methods, and computational efficiency profiling on edge hardware. Our results reveal that existing CAD methods do not consistently outperform traditional approaches with simple experience replay. Thus motivated, we propose DINOSaur, a training-free method combining a frozen DINOv3 backbone with spatially-indexed coreset memory and neighborhood-restricted anomaly scoring. DINOSaur achieves zero forgetting by construction, outperforms all evaluated methods across all five protocols, and runs at sub-100\,ms inference on an NVIDIA Jetson Orin Nano, with on-device adaptation to new tasks in under 30 seconds.

Problem

• continual anomaly detection
• edge computing
• industrial inspection
• realistic evaluation
• deployment constraints

Innovation

• Continual Anomaly Detection
• Edge Deployment
• Training-Free Method
• DINOv3 Backbone
• Coreset Memory