(50-5) 15 * << * >> * Russian * English * Content * All Issues

A lightweight and high-accuracy algorithm
for strip steel surface defect detection
J.N. Wang1,2,3, X. Xie2,*, L. Wu1,3, W.Y. Bai1 , W.Y. Guo1

1 School of Software Engineering, Anyang Normal University, Anyang, 455000, Henan, China;
2 School of Information and Software Engineering, East China Jiaotong University, Nanchang, 330013, Jiangxi, China;
3 Engineering Research Center for Intelligent Industry of Oracle Bone Culture in Henan Province,
Anyang, 455000, Henan, China

  Full text (PDF)

DOI: 10.18287/COJ1808

Article ID: 1808

Language: English

Abstract:
Given the dual demands for lightweight design and high precision in detecting surface defects of strip steel within complex industrial environments, this paper develops an improved YOLOv8 algorithm based on the YOLOv8n architecture. To start with, a SlimNeck lightweight structure is integrated into the neck network to reconstruct the feature extraction pathway, allowing the improved YOLOv8 model to markedly reduce its complexity while preserving robust feature expression capabilities. Next, the DySample dynamic upsampling operator is utilized in the neck network instead of the conventional UpSample operator, aiming to boost the model’s sensitivity to minute defects on the strip steel surface. Finally, the InnerMPDIoU loss function is incorporated to optimize bounding box regression accuracy, thereby fortifying the model’s adaptability to the multi-scale features of strip steel surface defects. Experimental results on the strip steel surface defect dataset, an open-source resource from Northeastern University (NEU-DET), indicate that the improved YOLOv8 algorithm achieves substantial improvements in detection accuracy while retaining its lightweight edge. Compared to the YOLOv8n baseline, it reduces parameters by 6.64 %, lowers computational load by 9.88 %, and maintains a model weight of just 5.9 MB. At the same time, its precision (P) reaches 80.1 % and mean average precision (mAP) hits 78.6 %, with these values being 3 % and 1.2 % higher than those of the baseline respectively. Consequently, the improved YOLOv8 algorithm proposed in this paper successfully balances lightweight design with high-precision detection performance, effectively meeting the dual requirements for strip steel surface defect detection tasks.

Keywords:
strip steel surface defect detection, YOLOv8, lightweight model, high accuracy detection, SlimNeck structure, DySample module, InnerMPDIoU loss function.

Acknowledgements:
This research was supported by the National Natural Science Foundation of China (No. 62162026), the Natural Science Foundation of Jiangxi Province (No. 20242BAB26019).

Citation:
Wang JN, Xie X, Wu L, Bai WY, Guo WY. A lightweight and high-accuracy algorithm for strip steel surface defect detection. Computer Optics 2026; 50(5): 1808. doi: 10.18287/COJ1808.

References:

  1. Yang W, Yang J, Xu CY. Strip surface defect detection based on improved YOLOv5. Acta Metrologica Sinica 2024; 45(11): 1671-1680. DOI: 10.3969/1000-1158.2024.11.11.
  2. Dai LH, Li YS, Shi R. YOLOv8-SSDW: Strip surface defect detection algorithm based on YOLOv8. Journal of Chongqing Technology and Business University 2025; 42(4): 44-52. DOI: 10.16055/1672-058X.2025.0004.006.
  3. Li WG, Ye X, Zhao YT, Wang WB. Strip steel surface defect detection based on improved YOLOv3 algorithm. Acta Electronica Sinica 2020; 48(7): 1284-1292. DOI: 10.3969/0372-2112.2020.07.006.
  4. Zhou S, Zeng Y, Li S. Surface defect detection of rolled steel based on lightweight model. Applied Sciences 2022; 12(17): 8905. DOI: 10.3390/app12178905.
  5. Yu JB, Cheng X, Li QF. Surface defect detection of steel strips based on anchor-free network with channel attention and bidirectional feature fusion. IEEE Transactions on Instrumentation and Measurement 2022; 71: 5000710.
  6. Zou W, Ji C. Real-time detection method of surface defects of hot-rolled strip via improved YOLOv4-tiny model. Mechanical Science and Technology for Aerospace Engineering 2023; 42(6): 883-889. DOI: 10.13433/1003-8728.20230034.
  7. Cao YQ, Wu ML, Xu L. Steel surface defect detection based on improved YOLOv5 algorithm. Journal of Graphics 2023; 44(2): 335-345.
  8. Deng H, Zeng L. Surface defect detection of hot-rolled strip steel based on improved Faster R-CNN. Control Engineering of China 2024; 31(4): 752-759. DOI: 10.14107/20220095.
  9. Yang LS, Li MJ, Hu JW, Wang DX. Strip steel surface defect detection algorithm based on improved YOLOv7-tiny. Computer Engineering 2025; 51(1): 208-215. DOI: 10.19678/1000-3428.0068397.
  10. Reis D, Kupec J, Hong J, et al. Real-time flying object detection with YOLOv8. arXiv preprint arXiv:2305.09972; 2023. DOI: 10.48550/arXiv.2305.09972.
  11. Shan F, Li H, Sun H. Pavement distress identification method based on improved SimAM-YOLOv8. Journal of Jilin University 2024: 1-14.
  12. Xu D, Wang S, Wang Z. Improved YOLOv8 urban vehicle target detection algorithm. Computer Engineering and Applications 2024; 60(18): 136-146.
  13. Huang G, Liu Z, van der Maaten L, Weinberger KQ. Densely connected convolutional networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. IEEE; 2017: 4700-4708. DOI: 10.1109/CVPR.2017.243.
  14. Lee Y, Hwang J, Lee S, Bae Y, Park J. An energy- and GPU-computation-efficient backbone network for real-time object detection. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. IEEE; 2019: 752-760. DOI: 10.1109/CVPRW.2019.00103.
  15. Wang CY, Liao HYM, Wu YH, Chen PY, Hsieh JW, Yeh IH. CSPNet: A new backbone that can enhance learning capability of CNN. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. IEEE; 2020: 390-391. DOI: 10.1109/CVPRW50498.2020.00203.
  16. Liu W, Lu H, Fu H. Learning to upsample by learning to sample. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. IEEE; 2023: 6027-6037. DOI: 10.1109/ICCV51070.2023.00557.
  17. Rezatofighi H, Tsoi N, Gwak JY, Sadeghian A, Reid I, Savarese S. Generalized intersection over union: A metric and a loss for bounding box regression. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE; 2019: 658-666. DOI: 10.1109/CVPR.2019.00075.
  18. Zheng ZH, Wang P, Liu W, Li J, Ye RG, Ren DW. Distance-IoU loss: Faster and better learning for bounding box regression. In: Proceedings of the AAAI Conference on Artificial Intelligence. AAAI Press; 2020; 34(7): 12993-13000. DOI: 10.1609/aaai.v34i07.6999.
  19. Zheng ZH, Wang P, Ren DW, Liu W, Ye RG, Hu QH. Enhancing geometric factors in model learning and inference for object detection and instance segmentation. IEEE Transactions on Cybernetics 2022; 52(8): 8574-8586. DOI: 10.1109/TCYB.2021.3095305.
  20. Qi W, Chen H, Ye Y, Yu G. Indoor object recognition based on YOLOv5 with EIoU loss function. In: Proceedings of SPIE 12799, Fourth International Conference on Machine Learning and Computer Application. SPIE; 2023: 1279907. DOI: 10.1117/12.3005836.
  21. Ma SL, Xu Y. MPDIoU: A loss for efficient and accurate bounding box regression. arXiv preprint arXiv:2307.07662; 2023. DOI: 10.48550/arXiv.2307.07662.
  22. Hao Z. Inner-IoU: More effective intersection over union loss with auxiliary bounding box. arXiv preprint arXiv:2311.02877; 2023. DOI: 10.48550/arXiv.2311.02877.
  23. Chu Y, Yu X, Rong X. A lightweight strip steel surface defect detection network based on improved YOLOv8. Sensors 2024; 24(19): 6495. DOI: 10.3390/s24196495.
  24. Zhou X, Zhu J, Wang Y. PCB bare board defect detection based on improved YOLOv7 algorithm. Radio Engineering 2023; 53(12): 2791-2797.
  25. Liu SX, Zhang L. DFW-YOLO: A small insulator target defect detection algorithm based on improved YOLOv8s. Computer Optics 2025; 49(5): 835-843. DOI: 10.18287/2412-6179-CO-1600.
  26. Zhao Z, Zhao S. KAN-YOLOv8: A method for surface defect detection of strip steel. Control Engineering of China 2025: 1-8.

151, Molodogvardeiskaya str., Samara, 443001, Russia; E-mail: journal@computeroptics.ru; Tel: +7 (846) 242-41-24 (Executive secretary), +7 (846) 332-56-22 (Issuing editor), Fax: +7 (846) 332-56-20