MVTec Anomaly Detection

MVTec anomaly detection focuses on identifying deviations from normal patterns in images and 3D point clouds, typically in industrial settings where labeled anomaly data is scarce. Current research emphasizes unsupervised learning approaches, employing autoencoders, diffusion models, and student-teacher networks to learn representations of normal data and detect deviations as reconstruction errors or discrepancies between student and teacher network outputs. These methods are evaluated on the MVTec dataset and its 3D counterpart, with a growing focus on robust evaluation methodologies and data augmentation techniques to improve performance and generalization. The field's advancements have significant implications for automated quality control and predictive maintenance in manufacturing and other industries.

Papers

May 15, 2024

A Hierarchically Feature Reconstructed Autoencoder for Unsupervised Anomaly Detection
Honghui Chen, Pingping Chen, Huan Mao, Mengxi Jiang
Anomaly Detection Unsupervised Anomaly Detection Anomaly Localization Anomaly Map Anomalous Image MVTec Anomaly Detection

February 29, 2024

A Novel Approach to Industrial Defect Generation through Blended Latent Diffusion Model with Online Adaptation
Hanxi Li, Zhengxun Zhang, Hao Chen, Lin Wu, Bo Li, Deyin Liu, Mingwen Wang
Novel Approach Industrial Anomaly Detection Online Adaptation Synthetic Sample Stochastic Defect MVTec Anomaly Detection

August 29, 2023

A Comprehensive Augmentation Framework for Anomaly Detection
Jiang Lin, Yaping Yan
Anomaly Detection Data Augmentation Data Augmentation Method Anomaly Detection Model MVTec Anomaly Detection

August 24, 2023

Multivariate Time Series Anomaly Detection: Fancy Algorithms and Flawed Evaluation Methodology
Mohamed El Amine Sehili, Zonghua Zhang
Anomaly Detection Practical Algorithm Multivariate Time Series Principal Component Analysis PCa Detection Multivariate Time Series Anomaly Detection MVTec Anomaly Detection

January 23, 2023

Adapting the Hypersphere Loss Function from Anomaly Detection to Anomaly Segmentation
Joao P. C. Bertoldo, Santiago Velasco-Forero, Jesus Angulo, Etienne Decencière
Anomaly Detection Anomaly Localization Anomaly Segmentation MVTec Anomaly Detection Anomaly Aware Semantic Segmentation

October 14, 2022

Asymmetric Student-Teacher Networks for Industrial Anomaly Detection
Marco Rudolph, Tom Wehrbein, Bodo Rosenhahn, Bastian Wandt
Anomaly Detection Industrial Anomaly Detection Information Asymmetry Teacher Network Industrial Defect MVTec Anomaly Detection

April 21, 2022

Feature anomaly detection system (FADS) for intelligent manufacturing
Anthony Garland, Kevin Potter, Matt Smith
Anomaly Detection Smart Manufacturing MVTec Anomaly Detection Deep Convolutional Activation Feature

February 25, 2022

Data refinement for fully unsupervised visual inspection using pre-trained networks
Antoine Cordier, Benjamin Missaoui, Pierre Gutierrez
Anomaly Detection Unsupervised Setting Outlier Detection Pre Trained Network Defect Detection System MVTec Anomaly Detection Data Refinement

February 23, 2022

Anomaly Detection in 3D Point Clouds using Deep Geometric Descriptors
Paul Bergmann, David Sattlegger
Anomaly Detection 3D Point Cloud Unsupervised Detection Anomaly Detection Framework Point Cloud Anomaly Detection MVTec Anomaly Detection

MVTec Anomaly Detection

Papers

A Hierarchically Feature Reconstructed Autoencoder for Unsupervised Anomaly Detection

A Novel Approach to Industrial Defect Generation through Blended Latent Diffusion Model with Online Adaptation

A Comprehensive Augmentation Framework for Anomaly Detection

Multivariate Time Series Anomaly Detection: Fancy Algorithms and Flawed Evaluation Methodology

Adapting the Hypersphere Loss Function from Anomaly Detection to Anomaly Segmentation

Asymmetric Student-Teacher Networks for Industrial Anomaly Detection

Feature anomaly detection system (FADS) for intelligent manufacturing

Data refinement for fully unsupervised visual inspection using pre-trained networks

Anomaly Detection in 3D Point Clouds using Deep Geometric Descriptors