FR-PatchCore: An Industrial Anomaly Detection Method for Improving Generalization
In recent years, a multitude of self-supervised anomaly detection algorithms have been proposed. Among them, PatchCore has emerged as one of the state-of-the-art methods on the widely used MVTec AD benchmark due to its efficient detection capabilities and cost-saving advantages in terms of labeled d...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2024-02-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/24/5/1368 |
| _version_ | 1827319400888270848 |
|---|---|
| author | Zhiqian Jiang Yu Zhang Yong Wang Jinlong Li Xiaorong Gao |
| author_facet | Zhiqian Jiang Yu Zhang Yong Wang Jinlong Li Xiaorong Gao |
| author_sort | Zhiqian Jiang |
| collection | DOAJ |
| description | In recent years, a multitude of self-supervised anomaly detection algorithms have been proposed. Among them, PatchCore has emerged as one of the state-of-the-art methods on the widely used MVTec AD benchmark due to its efficient detection capabilities and cost-saving advantages in terms of labeled data. However, we have identified that the PatchCore similarity principal approach faces significant limitations in accurately locating anomalies when there are positional relationships between similar samples, such as rotation, flipping, or misaligned pixels. In real-world industrial scenarios, it is common for samples of the same class to be found in different positions. To address this challenge comprehensively, we introduce Feature-Level Registration PatchCore (FR-PatchCore), which serves as an extension of the PatchCore method. FR-PatchCore constructs a feature matrix that is extracted into the memory bank and continually updated using the optimal negative cosine similarity loss. Extensive evaluations conducted on the MVTec AD benchmark demonstrate that FR-PatchCore achieves an impressive image-level anomaly detection AUROC score of up to 98.81%. Additionally, we propose a novel method for computing the mask threshold that enables the model to scientifically determine the optimal threshold and accurately partition anomalous masks. Our results highlight not only the high generalizability but also substantial potential for industrial anomaly detection offered by FR-PatchCore. |
| first_indexed | 2024-04-25T00:20:09Z |
| format | Article |
| id | doaj.art-7efb70bf3d8a446abd46636fb5d7903a |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-04-25T00:20:09Z |
| publishDate | 2024-02-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-7efb70bf3d8a446abd46636fb5d7903a2024-03-12T16:54:32ZengMDPI AGSensors1424-82202024-02-01245136810.3390/s24051368FR-PatchCore: An Industrial Anomaly Detection Method for Improving GeneralizationZhiqian Jiang0Yu Zhang1Yong Wang2Jinlong Li3Xiaorong Gao4School of Physical Science and Technology, Southwest Jiaotong University, Chengdu 611756, ChinaSchool of Physical Science and Technology, Southwest Jiaotong University, Chengdu 611756, ChinaSchool of Physical Science and Technology, Southwest Jiaotong University, Chengdu 611756, ChinaSchool of Physical Science and Technology, Southwest Jiaotong University, Chengdu 611756, ChinaSchool of Physical Science and Technology, Southwest Jiaotong University, Chengdu 611756, ChinaIn recent years, a multitude of self-supervised anomaly detection algorithms have been proposed. Among them, PatchCore has emerged as one of the state-of-the-art methods on the widely used MVTec AD benchmark due to its efficient detection capabilities and cost-saving advantages in terms of labeled data. However, we have identified that the PatchCore similarity principal approach faces significant limitations in accurately locating anomalies when there are positional relationships between similar samples, such as rotation, flipping, or misaligned pixels. In real-world industrial scenarios, it is common for samples of the same class to be found in different positions. To address this challenge comprehensively, we introduce Feature-Level Registration PatchCore (FR-PatchCore), which serves as an extension of the PatchCore method. FR-PatchCore constructs a feature matrix that is extracted into the memory bank and continually updated using the optimal negative cosine similarity loss. Extensive evaluations conducted on the MVTec AD benchmark demonstrate that FR-PatchCore achieves an impressive image-level anomaly detection AUROC score of up to 98.81%. Additionally, we propose a novel method for computing the mask threshold that enables the model to scientifically determine the optimal threshold and accurately partition anomalous masks. Our results highlight not only the high generalizability but also substantial potential for industrial anomaly detection offered by FR-PatchCore.https://www.mdpi.com/1424-8220/24/5/1368image anomaly detectionself-supervised learningfeature processing |
| spellingShingle | Zhiqian Jiang Yu Zhang Yong Wang Jinlong Li Xiaorong Gao FR-PatchCore: An Industrial Anomaly Detection Method for Improving Generalization Sensors image anomaly detection self-supervised learning feature processing |
| title | FR-PatchCore: An Industrial Anomaly Detection Method for Improving Generalization |
| title_full | FR-PatchCore: An Industrial Anomaly Detection Method for Improving Generalization |
| title_fullStr | FR-PatchCore: An Industrial Anomaly Detection Method for Improving Generalization |
| title_full_unstemmed | FR-PatchCore: An Industrial Anomaly Detection Method for Improving Generalization |
| title_short | FR-PatchCore: An Industrial Anomaly Detection Method for Improving Generalization |
| title_sort | fr patchcore an industrial anomaly detection method for improving generalization |
| topic | image anomaly detection self-supervised learning feature processing |
| url | https://www.mdpi.com/1424-8220/24/5/1368 |
| work_keys_str_mv | AT zhiqianjiang frpatchcoreanindustrialanomalydetectionmethodforimprovinggeneralization AT yuzhang frpatchcoreanindustrialanomalydetectionmethodforimprovinggeneralization AT yongwang frpatchcoreanindustrialanomalydetectionmethodforimprovinggeneralization AT jinlongli frpatchcoreanindustrialanomalydetectionmethodforimprovinggeneralization AT xiaoronggao frpatchcoreanindustrialanomalydetectionmethodforimprovinggeneralization |