A Survey on Unsupervised Industrial Anomaly Detection Algorithms

Abstract

In line with the development of Industry 4.0, more and more attention is attracted to the field of surface defect detection. Improving efficiency as well as saving labor costs has steadily become a matter of great concern in industry field, where deep learning-based algorithms performs better than traditional vision inspection methods in recent years. While existing deep learning-based algorithms are biased towards supervised learning, which not only necessitates a huge amount of labeled data and a significant amount of labor, but it is also inefficient and has certain limitations. In contrast, recent research shows that unsupervised learning has great potential in tackling above disadvantages for visual anomaly detection. In this survey, we summarize current challenges and provide a thorough overview of recently proposed unsupervised algorithms for visual anomaly detection covering five categories, whose innovation points and frameworks are described in detail. Meanwhile, information on publicly available datasets containing surface image samples are provided. By comparing different classes of methods, the advantages and disadvantages of anomaly detection algorithms are summarized. It is expected to assist both the research community and industry in developing a broader and cross-domain perspective.

Deep Industrial Image Anomaly Detection: A Survey

“…Secondly, we analyze the performance of IAD using the most comprehensive image level and pixel level metrics. Nevertheless, Cui et al [3] and Tao et al [2] only employ image level metrics, neglecting the anomalies localization performance of IAD.…”

ViV-Ano: Anomaly Detection and Localization Combining Vision Transformer and Variational Autoencoder in the Manufacturing Process

“…Existing material anomaly detection, identification, and classification could be performed manually by professional technicians, but this is neither efficient nor accurate. Equipment capable of detecting abnormalities in warehousing processes and process lines can be expected to improve quality and reduce costs compared to manual inspections [11]. In the factory, when there is a large difference between the training dataset and the test dataset, a robot can be stopped so that the robot itself, the manipulated object, and the environment are not damaged.…”

Faulty Detection System Based on SPC and Machine Learning Techniques

“…Since application domains have their specificity depending on the data generated or exploited, not all methods of anomaly detection are suitable for all application domains. Yajie Cui made a review covering several application domains [1]. Gupta and al reviewed the temporal anomaly detection methods that are applicable in several different domains [2].…”

“…The Intrusion detection consists of the analysis of a target usually a network or host to detect anomalous behavior [3]. It is in effect fraudulent attempts to access a resource by violating the security set up for the target in question [1]. Fraud detection allows the identification of suspicious activities conducted by an individual usually under a false identity (impersonation) [4].…”

“…statistical methods build a model with a confidence interval from the existing data. New data that do not fit this model will be considered abnormal [1].…”