C3D-AD: Toward Continual 3D Anomaly Detection via Kernel Attention with Learnable Advisor

2025-08-02

AI Summary

Existing 3D anomaly detection methods for industrial point cloud defect inspection suffer from reliance on category-specific training and poor adaptability to newly emerging anomaly classes. To address this, we propose a continual learning framework for multi-category point clouds. Our key contributions are: (1) a Kernel Attention layer with random features (KAL) for generic local feature extraction; (2) a learnable consultant-guided Kernel Attention Adapter (KAA) enabling robust feature reconstruction under parameter perturbations; and (3) an RPP module integrating representation replay and consistency optimization to balance retention of prior knowledge with incremental learning of novel anomaly classes. Evaluated on Real3D-AD, Anomaly-ShapeNet, and MulSen-AD benchmarks, our method achieves average AUROC scores of 66.4%, 83.1%, and 63.4%, respectively—demonstrating significant improvements in generalization and continual learning capability for dynamically introduced anomaly categories.

Technology Category

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

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

3D Anomaly Detection (AD) has shown great potential in detecting anomalies or defects of high-precision industrial products. However, existing methods are typically trained in a class-specific manner and also lack the capability of learning from emerging classes. In this study, we proposed a continual learning framework named Continual 3D Anomaly Detection (C3D-AD), which can not only learn generalized representations for multi-class point clouds but also handle new classes emerging over time.Specifically, in the feature extraction module, to extract generalized local features from diverse product types of different tasks efficiently, Kernel Attention with random feature Layer (KAL) is introduced, which normalizes the feature space. Then, to reconstruct data correctly and continually, an efficient Kernel Attention with learnable Advisor (KAA) mechanism is proposed, which learns the information from new categories while discarding redundant old information within both the encoder and decoder. Finally, to keep the representation consistency over tasks, a Reconstruction with Parameter Perturbation (RPP) module is proposed by designing a representation rehearsal loss function, which ensures that the model remembers previous category information and returns category-adaptive representation.Extensive experiments on three public datasets demonstrate the effectiveness of the proposed method, achieving an average performance of 66.4%, 83.1%, and 63.4% AUROC on Real3D-AD, Anomaly-ShapeNet, and MulSen-AD, respectively.

Problem

• Detects anomalies in 3D industrial products efficiently
• Learns from new classes without forgetting old ones
• Ensures consistent performance across multiple tasks

Innovation

• Kernel Attention Layer for generalized local features
• Kernel Attention with learnable Advisor mechanism
• Reconstruction with Parameter Perturbation module