EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

An Improved Lightweight YOLOv8 Model for Frozen Dumplings to Surface Defect Detection

Baoyong Guo, Du Xiong, Yuping Wang

IEEE Access · 2026 · 14 · 79987-80000

DOI: 10.1109/access.2026.3696389 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Baoyong Guo und 2 weiteren Autor:innen. Die Arbeit nennt 41 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Scientific and Technological Project of Henan Province, Zhengzhou University of Science and Technology Natural Science Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Surface defect detection in industrial scenarios still faces many challenges, such as small defect size, weak texture, high similarity, and the need for real-time performance. To address these issues, this paper proposes a lightweight defect detection method based on the improved YOLOv8 framework, called D-YOLOv8. Firstly, a lightweight heterogeneous structure is introduced into the backbone network, replacing some standard convolutions with GSConv, reducing the number of parameters while maintaining feature representation capabilities. Secondly, in order to enhance the perception of minute defects, a four-scale feature fusion structure, BiFPN, was constructed. A high-resolution feature layer, P2, is introduced to achieve efficient feature fusion. Finally, an attention mechanism module CA and a dual IoU threshold decoupling mechanism are integrated into the detection head to optimize localization and classification tasks, improving robustness in high-similarity scenarios. Experimental results on a self-built frozen dumpling defect dataset and a publicly available PCB dataset show that the proposed method achieves 93.8% mAP@0.5, outperforming several mainstream detection models, while maintaining high inference speed and low model complexity. The proposed method provides an efficient and practical solution for real-time industrial defect detection, especially suitable for small targets and samples with high similarity.

Förderung

  • Scientific and Technological Project of Henan Province
  • Zhengzhou University of Science and Technology Natural Science Foundation

Zitieren

APA

Guo, B., Xiong, D., & Wang, Y. (2026). An Improved Lightweight YOLOv8 Model for Frozen Dumplings to Surface Defect Detection. IEEE Access, 14, 79987-80000. https://doi.org/10.1109/access.2026.3696389

Vancouver

Guo B, Xiong D, Wang Y. An Improved Lightweight YOLOv8 Model for Frozen Dumplings to Surface Defect Detection. IEEE Access. 2026;14:79987-80000. doi:10.1109/access.2026.3696389

BibTeX
@article{Guo2026,
  author = {Guo, Baoyong and Xiong, Du and Wang, Yuping},
  title = {An Improved Lightweight YOLOv8 Model for Frozen Dumplings to Surface Defect Detection},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {79987--80000},
  doi = {10.1109/access.2026.3696389}
}
Kostenlos starten →