EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Regularized Latent Adaptive Framework for Unsupervised Industrial Anomaly Detection via Multi-Scale Generative–Discriminative Learning

Leqi Chi, Tao Ma iD, Yuhang Lang, Xinran Lv, Xingfan Li, Xiaoguang Li

Sensors · 01.07.2026 · 26 (13) · 4151

DOI: 10.3390/s26134151 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 01.07.2026. Verfasst von Leqi Chi und 5 weiteren Autor:innen. Beteiligte Einrichtung: School of Electronic Information Engineering, Changchun University, Changchun 130000, China. Die Arbeit nennt 46 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Jilin Provincial Development and Reform Commission.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Industrial visual inspection relies heavily on unsupervised anomaly detection due to the scarcity of annotated defect samples. However, existing methods struggle to balance global structural consistency and local defect sensitivity, leading to limited accuracy in practical scenarios. To address this challenge, we propose a unified generative–discriminative framework that combines regularized latent space encoding with multi-scale discriminator-guided supervision. Specifically, a dynamic compactness regularization strategy constrains latent representations of normal samples into a compact manifold to suppress anomaly reconstruction, while a multi-scale discriminator provides hierarchical perceptual constraints to enhance fine-grained anomaly localization across different spatial resolutions. Here we show that the proposed method achieves 98.6% image-level and 98.4% pixel-level AUROC on the MVTec AD benchmark, outperforming state-of-the-art approaches. This framework provides a stable and effective solution for real-world industrial quality inspection.

Schlagwörter

Stichwörter der Autor:innen

attention mechanismfeature fusiongenerative adversarial networksindustrial anomaly detectionlatent space distributionunsupervised learning

Autor:innen und Institutionen

  • School of Electronic Information Engineering, Changchun University, Changchun 130000, China.
  • School of Economics and Management, Harbin University of Science and Technology, Harbin 150000, China.
  • School of Mechanical and Power Engineering, Harbin University of Science and Technology, Harbin 150000, China.
  • School of Electrical and Information Engineering, Changchun Guanghua University, Changchun 130000, China.

ORCID-Profile: Tao Ma

Förderung

  • Jilin Provincial Development and Reform Commission

Zitieren

APA

Chi, L., Ma, T., Lang, Y., Lv, X., Li, X., & Li, X. (2026). Regularized Latent Adaptive Framework for Unsupervised Industrial Anomaly Detection via Multi-Scale Generative–Discriminative Learning. Sensors, 26(13), 4151. https://doi.org/10.3390/s26134151

Vancouver

Chi L, Ma T, Lang Y, Lv X, Li X, Li X. Regularized Latent Adaptive Framework for Unsupervised Industrial Anomaly Detection via Multi-Scale Generative–Discriminative Learning. Sensors. 2026;26(13):4151. doi:10.3390/s26134151

BibTeX
@article{Chi2026,
  author = {Chi, Leqi and Ma, Tao and Lang, Yuhang and Lv, Xinran and Li, Xingfan and Li, Xiaoguang},
  title = {Regularized Latent Adaptive Framework for Unsupervised Industrial Anomaly Detection via Multi-Scale Generative–Discriminative Learning},
  journal = {Sensors},
  year = {2026},
  volume = {26},
  number = {13},
  pages = {4151},
  doi = {10.3390/s26134151}
}
Kostenlos starten →