EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Discrepancy-Guided Complementary Fusion for Unsupervised Multimodal Anomaly Detection

Taehui Lee iD, Seyoung Jeong iD, Sang Jun Lee iD

Sensors · 12.06.2026 · 26 (12) · 3757

DOI: 10.3390/s26123757 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 12.06.2026. Verfasst von Taehui Lee und 2 weiteren Autor:innen. Beteiligte Einrichtung: Division of Electronic Engineering, Jeonbuk National University, 567 Baekje-daero, Deokjin-gu, Jeonju 54896, Republic of Korea. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 24 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Institute of Information & Communications Technology Planning & Evaluation, Institute of Information & Communications Technology Planning & Evaluation.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In industrial inspection, subtle defects often appear as local variations in appearance or geometry, making reliable anomaly detection challenging. A single sensing modality can miss important defect cues, while multimodal inspection combines appearance and geometric information to represent industrial objects more comprehensively. Many existing multimodal anomaly detection methods adopt early fusion strategies that integrate features at an early stage of the network. Such early integration can dilute modality-specific anomaly responses and cause anomaly smoothing, leading to degraded detection and localization performance. To address these challenges, we propose a reconstruction-based unsupervised multimodal anomaly detection framework integrating Discrepancy-Guided Complementary Fusion (DGCF) and Noise to Feature (N2F). Specifically, DGCF reduces anomaly smoothing by exploiting cross-modal discrepancies to extract complementary information, rather than directly summing or concatenating features from different modalities. Furthermore, N2F injects Gaussian noise into the feature space to regularize feature reconstruction and encourage the decoder to learn robust normal representations. Experimental results on the MVTec 3D-AD and Eyecandies datasets demonstrate the effectiveness of the proposed method. The proposed method achieves 97.3% I-AUROC, 99.6% P-AUROC, and 97.6% AUPRO on MVTec 3D-AD, and 94.8% I-AUROC, 98.6% P-AUROC, and 93.4% AUPRO on Eyecandies.

Schlagwörter

Stichwörter der Autor:innen

feature extractionfeature-level fusionindustrial anomaly detectionmulti-sensor fusionmultimodal anomaly detectionunsupervised learning

Autor:innen und Institutionen

  • Division of Electronic Engineering, Jeonbuk National University, 567 Baekje-daero, Deokjin-gu, Jeonju 54896, Republic of Korea.

ORCID-Profile: Taehui Lee, Seyoung Jeong, Sang Jun Lee

Förderung

  • Institute of Information & Communications Technology Planning & Evaluation
  • Institute of Information & Communications Technology Planning & Evaluation

Zitieren

APA

Lee, T., Jeong, S., & Lee, S. J. (2026). Discrepancy-Guided Complementary Fusion for Unsupervised Multimodal Anomaly Detection. Sensors, 26(12), 3757. https://doi.org/10.3390/s26123757

Vancouver

Lee T, Jeong S, Lee SJ. Discrepancy-Guided Complementary Fusion for Unsupervised Multimodal Anomaly Detection. Sensors. 2026;26(12):3757. doi:10.3390/s26123757

BibTeX
@article{Lee2026,
  author = {Lee, Taehui and Jeong, Seyoung and Lee, Sang Jun},
  title = {Discrepancy-Guided Complementary Fusion for Unsupervised Multimodal Anomaly Detection},
  journal = {Sensors},
  year = {2026},
  volume = {26},
  number = {12},
  pages = {3757},
  doi = {10.3390/s26123757}
}
Kostenlos starten →