Unsupervised Image Anomaly Detection Based on Constrained BidiRectional Distillation

LI Bo; LI Ze-chao; XING Peng; TANG Jin-hui

doi:10.12263/DZXB.20240733

您当前的位置：

首页 >

文章列表页 >

Unsupervised Image Anomaly Detection Based on Constrained BidiRectional Distillation

PAPERS | 更新时间：2025-12-08

- Unsupervised Image Anomaly Detection Based on Constrained BidiRectional Distillation
- ACTA ELECTRONICA SINICA Vol. 53, Issue 3, Pages: 895-909(2025)
- 作者机构：
  
  南京理工大学计算机科学与工程学院，江苏南京 210094
- 作者简介：
- 基金信息：
  
  National Natural Science Foundation of China(U20B2064;U21B2043)
- DOI：10.12263/DZXB.20240733
  CLC： TP37;TP391.4
- Received：07 August 2024，
  
  Revised：2024-10-22，
  
  Published：25 March 2025
- 稿件说明：
移动端阅览
李波, 李泽超, 邢鹏, 等. 基于双向约束蒸馏的无监督图像异常检测[J]. 电子学报, 2025, 53(03): 895-909.

LI Bo, LI Ze-chao, XING Peng, et al. Unsupervised Image Anomaly Detection Based on Constrained BidiRectional Distillation[J]. Acta Electronica Sinica, 2025, 53(03): 895-909.
李波, 李泽超, 邢鹏, 等. 基于双向约束蒸馏的无监督图像异常检测[J]. 电子学报, 2025, 53(03): 895-909. DOI：10.12263/DZXB.20240733

LI Bo, LI Ze-chao, XING Peng, et al. Unsupervised Image Anomaly Detection Based on Constrained BidiRectional Distillation[J]. Acta Electronica Sinica, 2025, 53(03): 895-909. DOI：10.12263/DZXB.20240733

摘要

异常检测是一项重要的计算机视觉任务，它的目标是检测异常样本同时定位异常区域.近期，主流的无监督异常检测方案通常基于蒸馏方法和重构方法.然而，它们仍存在相似的局限.在基于蒸馏方法的异常检测中，学生网络通常能学习到教师网络相似的表征能力，无法针对某些异常区域产生与教师网络有明显差异的表征.在重构模型中，编码-解码结构容易学习到简单的复原捷径，导致复原图像与输入相似，无法有效地检

测异常.为了解决上述挑战，本文提出基于双向约束蒸馏的无监督图像异常检测方法

https://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=83539255&type=

2.20133328

https://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=83539256&type=

2.70933342

-Net，它通过双向蒸馏模块和多级过滤模块缓解了上述局限.具体地，在教师学生网络中，本文首先提出蒸馏适应域特征而非原始域特征，它通过双向蒸馏分支保证了正常适应域特征的高效对齐.然后，本文提出多级过滤模块，通过查询和压缩的方式过滤异常特征，进一步增强学习正常语义特征分布的能力，提升异常检测性能.最后，本文在两个基准异常检测数据集MVTec和VisA上进行了大量实验，结果表明所提方法在异常检测和定位任务上取得了先进的性能.

Abstract

Anomaly detection has been widely studied and applied to various visual scenes. Recently

the mainstream unsupervised anomaly detection schemes are usually based on distillation methods and reconstruction methods. However

they still have some limitations. In distillation model

the student network can usually learn the strong representation ability of the teacher network

thus can not represent differently for the abnormal regions. In reconstruction model

the encoder-decoder model can easily learn a restoration shortcut and recover features indiscriminately. To address the above challenges

we propose

https://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=83539255&type=

2.20133328

https://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=83539256&type=

2.70933342

-Net

which integrates the advantages of above two methods and

alleviates limitations through the bidirectional distillation module and the multistage filtration mechanism. Specifically

in the teacher-student network

this paper first proposes distilling adaptive domain features instead of original domain features

which ensures efficient alignment of normal adaptive domain features through bidirectional distillation branches. Then

we propose a multilevel filtering module to filter abnormal features through query and compression to further enhance the ability to learn normal semantic feature distribution and improve the anomaly detection performance. Finally

a large number of experiments are carried out on two benchmark anomaly detection datasets

MVTec and VisA. The results show that the proposed method achieves advanced performance in anomaly detection and location tasks.

关键词

Keywords

references

GU Z H , LIU L , CHEN X , et al . Remembering normality: Memory-guided knowledge distillation for unsupervised anomaly detection [C ] // 2023 IEEE/CVF International Conference on Computer Vision (ICCV) . Piscataway : IEEE , 2023 : 16355 - 16363 .

BERGMANN P , FAUSER M , SATTLEGGER D , et al . MVTec AD: A comprehensive real-world dataset for unsupervised anomaly detection [C ] // 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Piscataway : IEEE , 2019 : 9584 - 9592 .

ZOU Y , JEONG J , PEMULA L , et al . SPot-the-difference self-supervised pre-training for anomaly detection and segmentation [M ] // Computer Vision - ECCV 2022 . Cham : Springer Nature Switzerland , 2022 : 392 - 408 .

严莉 , 张凯 , 徐浩 , 等 . 基于图注意力机制和Transformer的异常检测 [J ] . 电子学报 , 2022 , 50 ( 4 ): 900 - 908 .

YAN L , ZHANG K , XU H , et al . Abnormal detection based on graph attention mechanisms and transformer [J ] . Acta Electronica Sinica , 2022 , 50 ( 4 ): 900 - 908 . (in Chinese)

PARK H , NOH J , HAM B . Learning memory-guided normality for anomaly detection [C ] // 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Piscataway : IEEE , 2020 : 14372 - 14381 .

ZHAO Y R , DENG B , SHEN C , et al . Spatio-temporal AutoEncoder for video anomaly detection [C ] // Proceedings of the 25th ACM International Conference on Multimedia . New York : ACM , 2017 : 1933 - 1941 .

XIANG T G , ZHANG Y X , LU Y Y , et al . SQUID: Deep feature in-painting for unsupervised anomaly detection [C ] // 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Piscataway : IEEE , 2023 : 23890 - 23901 .

尚文利 , 石贺 , 赵剑明 , 等 . 基于SAE-LSTM的工艺数据异常检测方法 [J ] . 电子学报 , 2021 , 49 ( 8 ): 1561 - 1568 .

SHANG W L , SHI H , ZHAO J M , et al . An anomaly detection method of process data based on SAE-LSTM [J ] . Acta Electronica Sinica , 2021 , 49 ( 8 ): 1561 - 1568 . (in Chinese)

GONG D , LIU L Q , LE V , et al . Memorizing normality to detect anomaly: Memory-augmented deep autoencoder for unsupervised anomaly detection [C ] // 2019 IEEE/CVF International Conference on Computer Vision (ICCV) . Piscataway : IEEE , 2019 : 1705 - 1714 .

BERGMANN P , FAUSER M , SATTLEGGER D , et al . Uninformed students: Student-teacher anomaly detection with discriminative latent embeddings [C ] // 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Piscataway : IEEE , 2020 : 4182 - 4191 .

HOU J L , ZHANG Y Y , ZHONG Q Y , et al . Divide-and-assemble: Learning Block-wise memory for unsupervised anomaly detection [C ] // 2021 IEEE/CVF International Conference on Computer Vision (ICCV) . Piscataway : IEEE , 2021 : 8771 - 8780 .

SCHLEGL T , SEEBÖCK P , WALDSTEIN S M , et al . Unsupervised Anomaly Detection with Generative Adversarial Networks to Guide Marker Discovery [M ] // Information Processing in Medical Imaging . Cham : Springer International Publishing , 2017 : 146 - 157 .

AKCAY S , ATAPOUR-ABARGHOUEI A , BRECKON T P . GANomaly: Semi-Supervised Anomaly Detection via Adversarial Training [M ] // Computer Vision - ACCV 2018 . Cham : Springer International Publishing , 2019 : 622 - 637 .

AKÇAY S , ATAPOUR-ABARGHOUEI A , BRECKON T P . Skip-GANomaly: Skip connected and adversarially trained encoder-decoder anomaly detection [C ] // 2019 International Joint Conference on Neural Networks (IJCNN) . Piscataway : IEEE , 2019 : 1 - 8 .

ZONG B , SONG Q , MIN M R , et al . Deep autoencoding Gaussian mixture model for unsupervised anomaly detection [C ] // International Conference on Learning Representations . Scottsdale : ICLR , 2018 : 1 - 19 .

XING P , LI Z C . Visual anomaly detection via partition memory bank module and error estimation [J ] . IEEE Transactions on Circuits and Systems for Video Technology , 2023 , 33 ( 8 ): 3596 - 3607 .

YAN X D , ZHANG H D , XU X M , et al . Learning semantic context from normal samples for unsupervised anomaly detection [J ] . Proceedings of the AAAI Conference on Artificial Intelligence , 2021 , 35 ( 4 ): 3110 - 3118 .

DEHAENE D , ELINE P . Anomaly localization by modeling perceptual features [EB/OL ] . ( 2020-08-12 )[ 2024-08-07 ] . https://arxiv.org/abs/2008.05369v1 https://arxiv.org/abs/2008.05369v1 .

LI Z , LI N , JIANG K , et al . Superpixel masking and inpainting for self-supervised anomaly detection [C ] // British Machine Vision Conference . London : BMVC , 2020 : 219 - 232 .

LIU J Q , XIE G Y , WANG J B , et al . Deep industrial image anomaly detection: A survey [J ] . Machine Intelligence Research , 2024 , 21 ( 1 ): 104 - 135 .

WANG G , HAN S , DING E , et al . Student-teacher feature pyramid matching for anomaly detection [C ] // British Machine Vision Conference . London : BMVC , 2021 : 1589 - 1605 .

YAMADA S , KAMIYA S , HOTTA K . Reconstructed student-teacher and discriminative networks for anomaly detection [C ] // 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . Piscataway : IEEE , 2022 : 2725 - 2732 .

邢鹏 , 蒋鑫 , 潘永华 , 等 . 基于特征约束蒸馏学习的视觉异常检测 [J ] . 软件学报 , 2023 , 34 ( 9 ): 4378 - 4391 .

XING P , JIANG X , PAN Y H , et al . Feature constrained restricted distillation learning for visual anomaly detection [J ] . Journal of Software , 2023 , 34 ( 9 ): 4378 - 4391 . (in Chinese)

SALEHI M , SADJADI N , BASELIZADEH S , et al . Multiresolution knowledge distillation for anomaly detection [C ] // Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . Piscataway : IEEE , 2021 : 14902 - 14912 .

DENG H Q , LI X Y . Anomaly detection via reverse distillation from one-class embedding [C ] // 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Piscataway : IEEE , 2022 : 9727 - 9736 .

MOYA M M , HUSH D R . Network constraints and multi-objective optimization for one-class classification [J ] . Neural Networks , 1996 , 9 ( 3 ): 463 - 474 .

SCHÖLKOPF B , PLATT J C , SHAWE-TAYLOR J , et al . Estimating the support of a high-dimensional distribution [J ] . Neural Computation , 2001 , 13 ( 7 ): 1443 - 1471 .

RUFF L , VANDERMEULEN R , GOERNITZ N , et al . Deep one-class classification [C ] // International Conference on Machine Learning . Stockholm : ICML , 2018 : 4393 - 4402 .

YI J H , YOON S . Patch SVDD: Patch-Level SVDD for anomaly detection and segmentation [M ] // Computer Vision - ACCV 2020 . Cham : Springer International Publishing , 2021 : 375 - 390 .

YAO X C , LI R Q , ZHANG J , et al . Explicit boundary guided semi-push-pull contrastive learning for supervised anomaly detection [C ] // 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Piscataway : IEEE , 2023 : 24490 - 24499 .

KINGMA D P , WELLING M . Auto-encoding variational Bayes [J ] . 2nd International Conference on Learning Representations ICLR 2014 - Conference Track Proceedings. Scottsdale: ICLR , 2014 : 33 .

GOODFELLOW I , POUGET-Abadie J , MIRZA M , et al . Generative adversarial nets [C ] // Proceedings of the 27th International Conference on Neural Information Processing Systems . New York : ACM , 2014 : 2672 - 2680 .

ZAVRTANIK V , KRISTAN M , SKOČAJ D . Reconstruction by inpainting for visual anomaly detection [J ] . Pattern Recognition , 2021 , 112 : 107706 .

ZHANG X , LI S Y , LI X , et al . DeSTSeg: Segmentation guided denoising student-teacher for anomaly detection [C ] // 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Piscataway : IEEE , 2023 : 3914 - 3923 .

COHEN N , HOSHEN Y . Sub-image anomaly detection with deep pyramid correspondences [EB/OL ] . ( 2021-02-03 )[ 2024-08-07 ] . https://arxiv.org/abs/2005.02357v3 https://arxiv.org/abs/2005.02357v3 .

COVER T , HART P . Nearest neighbor pattern classification [J ] . IEEE Transactions on Information Theory , 1967 , 13 ( 1 ): 21 - 27 .

DEFARD T , SETKOV A , LOESCH A , et al . PaDiM: A patch distribution modeling framework for anomaly detection and localization [M ] // Pattern Recognition ICPR International Workshops and Challenges . Cham : Springer International Publishing , 2021 : 475 - 489 .

ROTH K , PEMULA L , ZEPEDA J , et al . Towards total recall in industrial anomaly detection [C ] // 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Piscataway : IEEE , 2022 : 14298 - 14308 .

LIU Z K , ZHOU Y M , XU Y S , et al . SimpleNet: A simple network for image anomaly detection and localization [C ] // 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Piscataway : IEEE , 2023 : 20402 - 20411 .

RUDOLPH M , WEHRBEIN T , ROSENHAHN B , et al . Fully convolutional cross-scale-flows for image-based defect detection [C ] // 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) . Piscataway : IEEE , 2022 : 1829 - 1838 .

YU J W , ZHENG Y , WANG X , et al . FastFlow: Unsupervised anomaly detection and localization via 2D normalizing flows [EB/OL ] . ( 2021-11-16 )[ 2024-08-07 ] . https://doi.org/10.48550/arXiv.2111.07677 https://doi.org/10.48550/arXiv.2111.07677 .

TIEN T D , NGUYEN A T , TRAN N H , et al . Revisiting reverse distillation for anomaly detection [C ] // 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Piscataway : IEEE , 2023 : 24511 - 24520 .

XING P , TANG H , TANG J H , et al . ADPS: Asymmetric distillation postsegmentation for image anomaly detection [J ] . IEEE Transactions on Neural Networks and Learning Systems , 2025 , 36 ( 4 ): 7051 - 7064 .

ZAGORUYKO S , KOMODAKIS N . Wide residual networks [EB/OL ] . ( 2017-06-14 )[ 2024-08-07 ] . https://arxiv.org/abs/1605.07146v4 https://arxiv.org/abs/1605.07146v4 .

KINGMA D P , BA J . Adam: A method for stochastic optimization [EB/OL ] . ( 2017-1-30 )[ 2024-08-07 ] . https://doi.org/10.48550/arXiv.1412.6980 https://doi.org/10.48550/arXiv.1412.6980 .

ZAVRTANIK V , KRISTAN M , SKOČAJ D . DRÆM-A discriminatively trained reconstruction embedding for surface anomaly detection [C ] // 2021 IEEE/CVF International Conference on Computer Vision (ICCV) . Piscataway : IEEE , 2021 : 8310 - 8319 .

ZHOU Q , PANG G , TIAN Y , et al . Anomalyclip: Objectagnostic prompt learning for zero-shot anomaly detection [C ] // International Conference on Learning Representations . Scottsdale : ICLR , 2024 : 3565 - 3583 .

LI X R , HUANG Z M , XUE F , et al . MuSc: Zero-shot industrial anomaly classification and segmentation with mutual scoring of the unlabeled images [EB/OL ] . ( 2024-01-30 )[ 2024-08-07 ] . https://arxiv.org/abs/2401.16753v1 https://arxiv.org/abs/2401.16753v1 .

HU T , ZHANG J N , YI R , et al . AnomalyDiffusion: Few-shot anomaly image generation with diffusion model [J ] . Proceedings of the AAAI Conference on Artificial Intelligence , 2024 , 38 ( 8 ): 8526 - 8534 .

HE H Y , ZHANG J N , CHEN H X , et al . A diffusion-based framework for multi-class anomaly detection [C ] // Proceedings of the Thirty-Eighth AAAI Conference on Artificial Intelligence and Thirty-Sixth Conference on Innovative Applications of Artificial Intelligence and Fourteenth Symposium on Educational Advances in Artificial Intelligence . New York : ACM , 2024 : 8472 - 8480 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Variable Horizon Multi-Directional Scanning Method for Time Series Anomaly Detection

A Vehicle Trajectory Anomaly Detection Method Based on Noise Label Re-Weighting

A Survey of Network Attack Investigation Based on Provenance Graph

Improved Adaptive Model Pools for Online Anomaly Detection Algorithms

Related Author

TANG Jin-hui

XING Peng

LI Ze-chao

LI Bo

WEI Song-jie

GUAN Yong-yuan

HUANG Yu-zhe

BI Jing-ping

Related Institution

School of Computer Science and Engineering, School of Cyber Science and Engineering, Nanjing University of Science and Technology

Institute of Computing Technology, Chinese Academy of Sciences

University of Chinese Academy of Sciences

NSFOCUS Technologies Group Co., Ltd

⁰