[论文解读] SLSG: Industrial Image Anomaly Detection by Learning Better Feature Embeddings and One-Class Classification

Minghui Yang, Jing Liu+2 more|arXiv (Cornell University)|Apr 30, 2023

Anomaly Detection Techniques and Applications被引用 8

一句话总结

SLSG 提出了一种自监督和自注意图方法用于工业图像异常检测，利用伪先验异常样本和基于图的推理来改进单类分类。

ABSTRACT

Industrial image anomaly detection under the setting of one-class classification has significant practical value. However, most existing models struggle to extract separable feature representations when performing feature embedding and struggle to build compact descriptions of normal features when performing one-class classification. One direct consequence of this is that most models perform poorly in detecting logical anomalies which violate contextual relationships. Focusing on more effective and comprehensive anomaly detection, we propose a network based on self-supervised learning and self-attentive graph convolution (SLSG) for anomaly detection. SLSG uses a generative pre-training network to assist the encoder in learning the embedding of normal patterns and the reasoning of position relationships. Subsequently, SLSG introduces the pseudo-prior knowledge of anomaly through simulated abnormal samples. By comparing the simulated anomalies, SLSG can better summarize the normal features and narrow down the hypersphere used for one-class classification. In addition, with the construction of a more general graph structure, SLSG comprehensively models the dense and sparse relationships among elements in the image, which further strengthens the detection of logical anomalies. Extensive experiments on benchmark datasets show that SLSG achieves superior anomaly detection performance, demonstrating the effectiveness of our method.

研究动机与目标

• 在单类分类下动机化工业异常检测及对可分特征嵌入的需求。
• 学习鲁棒的特征嵌入，捕捉正常模式及其位置关系。
• 通过模拟异常样本引入伪先验的异常知识。
• 用通用图结构建模密集与稀疏元素关系，以更好地检测逻辑异常。

提出的方法

• 使用生成式预训练网络帮助编码器学习正常模式嵌入及位置推理。
• 融入自注意图卷积以捕捉图像元素之间的关系。
• 通过模拟的异常样本引入伪先验的异常知识，以完善对正常特征的描述。
• 通过比较模拟的异常来收窄单类分类的超球体，以更好地概括正常特征。
• 构建通用图结构以建模密集和稀疏元素关系，从而改进对逻辑异常的检测。

实验结果

研究问题

• RQ1 我们如何学习出更可分、鲁棒的工业异常检测特征嵌入？
• RQ2 来自模拟异常样本的伪先验知识是否能改进单类分类？
• RQ3 自注意图构建是否更好地捕捉关系并检测工业图像中的逻辑异常？

主要发现

• SLSG 在基准数据集上实现了优于其他方法的异常检测性能。
• 该方法有效地概括正常特征并收紧单类超球体。
• 基于图的推理通过建模密集和稀疏关系来提升对逻辑异常的检测。
• 自监督预训练有助于同时学习嵌入和位置关系。
• 模拟异常有助于提供伪先验知识，指导单类学习。