Prototypical Residual Networks for Anomaly Detection and Localization

被引:52
作者
Zhang, Hui [1 ,2 ]
Wu, Zuxuan [1 ,2 ]
Wang, Zheng [3 ]
Chen, Zhineng [1 ,2 ]
Jiang, Yu-Gang [1 ,2 ]
机构
[1] Fudan Univ, Shanghai Key Lab Intell Info Proc, Sch CS, Shanghai, Peoples R China
[2] Shanghai Collaborat Innovat Ctr Intelligent Visua, Shanghai, Peoples R China
[3] Zhejiang Univ Technol, Sch Comp Sci, Hangzhou, Peoples R China
来源
2023 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR) | 2023年
关键词
D O I
10.1109/CVPR52729.2023.01562
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Anomaly detection and localization are widely used in industrial manufacturing for its efficiency and effectiveness. Anomalies are rare and hard to collect and supervised models easily over-fit to these seen anomalies with a handful of abnormal samples, producing unsatisfactory performance. On the other hand, anomalies are typically subtle, hard to discern, and of various appearance, making it difficult to detect anomalies and let alone locate anomalous regions. To address these issues, we propose a framework called Prototypical Residual Network (PRN), which learns feature residuals of varying scales and sizes between anomalous and normal patterns to accurately reconstruct the segmentation maps of anomalous regions. PRN mainly consists of two parts: multi-scale prototypes that explicitly represent the residual features of anomalies to normal patterns; a multisize self-attention mechanism that enables variable-sized anomalous feature learning. Besides, we present a variety of anomaly generation strategies that consider both seen and unseen appearance variance to enlarge and diversify anomalies. Extensive experiments on the challenging and widely used MVTec AD benchmark show that PRN outperforms current state-of-the-art unsupervised and supervised methods. We further report SOTA results on three additional datasets to demonstrate the effectiveness and generalizability of PRN.
引用
收藏
页码:16281 / 16291
页数:11
相关论文
共 70 条
[1]  
Akcay S., 2022, arXiv
[2]   Skip-GANomaly: Skip Connected and Adversarially Trained Encoder-Decoder Anomaly Detection [J].
Akcay, Samet ;
Atapour-Abarghouei, Amir ;
Breckon, Toby P. .
2019 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2019,
[3]   GANomaly: Semi-supervised Anomaly Detection via Adversarial Training [J].
Akcay, Samet ;
Atapour-Abarghouei, Amir ;
Breckon, Toby P. .
COMPUTER VISION - ACCV 2018, PT III, 2019, 11363 :622-637
[4]  
Bergmann P, 2019, Arxiv, DOI [arXiv:1807.02011, DOI 10.48550/ARXIV.1807.02011]
[5]   Uninformed Students: Student-Teacher Anomaly Detection with Discriminative Latent Embeddings [J].
Bergmann, Paul ;
Fauser, Michael ;
Sattlegger, David ;
Steger, Carsten .
2020 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2020, :4182-4191
[6]   MVTec AD - A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection [J].
Bergmann, Paul ;
Fauser, Michael ;
Sattlegger, David ;
Steger, Carsten .
2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2019), 2019, :9584-9592
[7]   Mixed supervision for surface-defect detection: From weakly to fully supervised learning [J].
Bozic, Jakob ;
Tabernik, Domen ;
Skocaj, Danijel .
COMPUTERS IN INDUSTRY, 2021, 129
[8]   Describing Textures in the Wild [J].
Cimpoi, Mircea ;
Maji, Subhransu ;
Kokkinos, Iasonas ;
Mohamed, Sammy ;
Vedaldi, Andrea .
2014 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2014, :3606-3613
[9]  
Cohen N, 2021, Arxiv, DOI arXiv:2005.02357
[10]  
Defard Thomas, 2021, Pattern Recognition. ICPR International Workshops and Challenges. Proceedings. Lecture Notes in Computer Science (LNCS 12664), P475, DOI 10.1007/978-3-030-68799-1_35