High-performance RT-DETR model for industrial defect detection with semantic guidance and hierarchical attention
PLOS One · 12.05.2026 · 21 (5) · e0348807
DOI: 10.1371/journal.pone.0348807 ↗Kurz gefasst
Fachartikel in der Zeitschrift PLOS One, online erschienen am 12.05.2026. Verfasst von Jin Huang und 7 weiteren Autor:innen. Beteiligte Einrichtung: School of Electronics and IoT Engineering; ChongQing Industry Polytechnic University, Chongqing, China. Hauptthemen laut MEDLINE: Semantics, Models, Theoretical, Industry. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 56 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Science and Technology Research Program of Chongqing Municipal Education Commission, Chongqing Vocational Education Key Research Topics, Key Project of the 2025 Special Program for Teaching Reform Research under the Chongqing Municipal Education Science Planning Program.
Abstract (Originalsprache)
With the advancement of intelligent manufacturing and Industry 4.0, surface defect detection plays a critical role in ensuring product quality and production safety. To address the limitations of existing detection models in handling small sample sizes, complex textures, and multi-scale defects, this paper proposes a high-performance industrial defect detection model based on the RT-DETR framework, incorporating semantic guidance and hierarchical attention mechanisms. Specifically, a Semantic-Guided Query Enhancement Module is designed to strengthen the contextual awareness of query vectors through multi-source semantic paths and a residual feedback structure. Additionally, a Hierarchical Attention Fusion Structure is constructed to build interactive graphs among multi-scale features, achieving cross-scale semantic alignment and structural consistency modeling. Experiments conducted on three benchmark industrial defect datasets—NEU-DET, DAGM2007, and PCB-DET—demonstrate the effectiveness of the proposed method, achieving mAP@0.5 scores of 78.9%, 84.7%, and 87.4%, respectively, outperforming the best baseline models by 1.2% to 3.0%. For the more stringent mAP@0.5:0.95 metric, the method achieves 44.3%, 48.1%, and 52.3%, significantly surpassing mainstream models such as YOLOv8, and BMA-YOLO. Furthermore, Grad-CAM visualizations validate the model’s superior focus capability and boundary-fitting accuracy in regions with complex textures and sparse targets. Overall, the proposed architecture enhances semantic perception, scale robustness, and generalization performance in industrial defect detection while maintaining real-time efficiency.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- School of Electronics and IoT Engineering; ChongQing Industry Polytechnic University, Chongqing, China.
ORCID-Profile: Jin Huang
Förderung
- Science and Technology Research Program of Chongqing Municipal Education Commission
- Chongqing Vocational Education Key Research Topics
- Key Project of the 2025 Special Program for Teaching Reform Research under the Chongqing Municipal Education Science Planning Program
Zitieren
APA
Huang, J., Gu, B., Tang, Y., Xu, B., Li, P., Ye, J., Xu, J., & Zhang, T. (2026). High-performance RT-DETR model for industrial defect detection with semantic guidance and hierarchical attention. PLOS One, 21(5), e0348807. https://doi.org/10.1371/journal.pone.0348807
Vancouver
Huang J, Gu B, Tang Y, Xu B, Li P, Ye J, et al. High-performance RT-DETR model for industrial defect detection with semantic guidance and hierarchical attention. PLOS One. 2026;21(5):e0348807. doi:10.1371/journal.pone.0348807
BibTeX
@article{Huang2026,
author = {Huang, Jin and Gu, Bin and Tang, Yunlong and Xu, Bingyue and Li, Puxuan and Ye, Jiayu and Xu, Jingwen and Zhang, Taihua},
title = {High-performance RT-DETR model for industrial defect detection with semantic guidance and hierarchical attention},
journal = {PLOS One},
year = {2026},
volume = {21},
number = {5},
pages = {e0348807},
doi = {10.1371/journal.pone.0348807}
}
EUVIMED