Details
| Original language | English |
|---|---|
| Title of host publication | 2023 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) |
| Pages | 2591-2601 |
| Number of pages | 11 |
| ISBN (electronic) | 9781665493468 |
| Publication status | Published - 2023 |
Abstract
Industrial defect detection is commonly addressed with anomaly detection (AD) methods where no or only incomplete data of potentially occurring defects is available. This work discovers previously unknown problems of student-teacher approaches for AD and proposes a solution, where two neural networks are trained to produce the same output for the defect-free training examples. The core assumption of student-teacher networks is that the distance between the outputs of both networks is larger for anomalies since they are absent in training. However, previous methods suffer from the similarity of student and teacher architecture, such that the distance is undesirably small for anomalies. For this reason, we propose asymmetric student-teacher networks (AST). We train a normalizing flow for density estimation as a teacher and a conventional feed-forward network as a student to trigger large distances for anomalies: The bijectivity of the normalizing flow enforces a divergence of teacher outputs for anomalies compared to normal data. Outside the training distribution the student cannot imitate this divergence due to its fundamentally different architecture. Our AST network compensates for wrongly estimated likelihoods by a normalizing flow, which was alternatively used for anomaly detection in previous work. We show that our method produces state-of-the-art results on the two currently most relevant defect detection datasets MVTec AD and MVTec 3D-AD regarding image-level anomaly detection on RGB and 3D data.
Keywords
- Algorithms: Image recognition and understanding (object detection, categorization, segmentation), and algorithms (including transfer, low-shot, semi-, self-, and un-supervised learning), formulations, Machine learning architectures
ASJC Scopus subject areas
- Computer Science(all)
- Artificial Intelligence
- Computer Science(all)
- Computer Vision and Pattern Recognition
- Computer Science(all)
- Computer Science Applications
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
2023 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). 2023. p. 2591-2601.
Research output: Chapter in book/report/conference proceeding › Conference contribution › Research
}
TY - GEN
T1 - Asymmetric Student-Teacher Networks for Industrial Anomaly Detection
AU - Rudolph, Marco
AU - Wehrbein, Tom
AU - Rosenhahn, Bodo
AU - Wandt, Bastian
N1 - Funding Information: This work was supported by the Fed- eral Ministry of Education and Research (BMBF), Germany under the project LeibnizKILabor (grant no. 01DD20003), the Center for Digital Innovations (ZDIN) and the Deutsche Forschungsge- meinschaft (DFG) under Germany’s Excellence Strategy within the Cluster of Excellence PhoenixD (EXC 2122).
PY - 2023
Y1 - 2023
N2 - Industrial defect detection is commonly addressed with anomaly detection (AD) methods where no or only incomplete data of potentially occurring defects is available. This work discovers previously unknown problems of student-teacher approaches for AD and proposes a solution, where two neural networks are trained to produce the same output for the defect-free training examples. The core assumption of student-teacher networks is that the distance between the outputs of both networks is larger for anomalies since they are absent in training. However, previous methods suffer from the similarity of student and teacher architecture, such that the distance is undesirably small for anomalies. For this reason, we propose asymmetric student-teacher networks (AST). We train a normalizing flow for density estimation as a teacher and a conventional feed-forward network as a student to trigger large distances for anomalies: The bijectivity of the normalizing flow enforces a divergence of teacher outputs for anomalies compared to normal data. Outside the training distribution the student cannot imitate this divergence due to its fundamentally different architecture. Our AST network compensates for wrongly estimated likelihoods by a normalizing flow, which was alternatively used for anomaly detection in previous work. We show that our method produces state-of-the-art results on the two currently most relevant defect detection datasets MVTec AD and MVTec 3D-AD regarding image-level anomaly detection on RGB and 3D data.
AB - Industrial defect detection is commonly addressed with anomaly detection (AD) methods where no or only incomplete data of potentially occurring defects is available. This work discovers previously unknown problems of student-teacher approaches for AD and proposes a solution, where two neural networks are trained to produce the same output for the defect-free training examples. The core assumption of student-teacher networks is that the distance between the outputs of both networks is larger for anomalies since they are absent in training. However, previous methods suffer from the similarity of student and teacher architecture, such that the distance is undesirably small for anomalies. For this reason, we propose asymmetric student-teacher networks (AST). We train a normalizing flow for density estimation as a teacher and a conventional feed-forward network as a student to trigger large distances for anomalies: The bijectivity of the normalizing flow enforces a divergence of teacher outputs for anomalies compared to normal data. Outside the training distribution the student cannot imitate this divergence due to its fundamentally different architecture. Our AST network compensates for wrongly estimated likelihoods by a normalizing flow, which was alternatively used for anomaly detection in previous work. We show that our method produces state-of-the-art results on the two currently most relevant defect detection datasets MVTec AD and MVTec 3D-AD regarding image-level anomaly detection on RGB and 3D data.
KW - Algorithms: Image recognition and understanding (object detection, categorization, segmentation)
KW - and algorithms (including transfer, low-shot, semi-, self-, and un-supervised learning)
KW - formulations
KW - Machine learning architectures
UR - http://www.scopus.com/inward/record.url?scp=85149018196&partnerID=8YFLogxK
U2 - 10.48550/arXiv.2210.07829
DO - 10.48550/arXiv.2210.07829
M3 - Conference contribution
SP - 2591
EP - 2601
BT - 2023 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)
ER -