AdapTS: Lightweight Teacher-Student Approach for Multi-Class and Continual Visual Anomaly Detection

2026-03-18

AI Summary

Existing visual anomaly detection methods are predominantly confined to single-category settings, limiting their applicability in industrial scenarios that demand multi-category support and continual learning. This work proposes AdapTS—a lightweight, unified teacher-student framework that, for the first time, extends the teacher-student paradigm to continual learning. AdapTS employs a shared frozen backbone with trainable adapters, integrating a segmentation-guided loss, synthetic Perlin noise augmentation, and a prototype-based task identification mechanism to dynamically select adapters for efficient edge deployment. Evaluated on MVTec AD and VisA, AdapTS achieves performance comparable to state-of-the-art methods while drastically reducing memory overhead: AdapTS-S requires only 8 MB of additional memory, representing reductions of 13×, 48×, and 149× compared to STFPM, RD4AD, and DeSTSeg, respectively.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Multiagent Systems: Adversarial Agents

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Visual Anomaly Detection (VAD) is crucial for industrial inspection, yet most existing methods are limited to single-category scenarios, failing to address the multi-class and continual learning demands of real-world environments. While Teacher-Student (TS) architectures are efficient, they remain unexplored for the Continual Setting. To bridge this gap, we propose AdapTS, a unified TS framework designed for multi-class and continual settings, optimized for edge deployment. AdapTS eliminates the need for two different architectures by utilizing a single shared frozen backbone and injecting lightweight trainable adapters into the student pathway. Training is enhanced via a segmentation-guided objective and synthetic Perlin noise, while a prototype-based task identification mechanism dynamically selects adapters at inference with 99\% accuracy. Experiments on MVTec AD and VisA demonstrate that AdapTS matches the performance of existing TS methods across multi-class and continual learning scenarios, while drastically reducing memory overhead. Our lightest variant, AdapTS-S, requires only 8 MB of additional memory, 13x less than STFPM (95 MB), 48x less than RD4AD (360 MB), and 149x less than DeSTSeg (1120 MB), making it a highly scalable solution for edge deployment in complex industrial environments.

Problem

• Visual Anomaly Detection
• Multi-Class
• Continual Learning
• Edge Deployment
• Teacher-Student Architecture

Innovation

• Teacher-Student architecture
• Continual Learning
• Lightweight Adapters
• Visual Anomaly Detection
• Edge Deployment