This paper proposes YOLOv5-CJ for steel surface defect detection. C3_MSBlock enhances multi-scale feature extraction and enlarges the receptive field, while DyHead introduces scale-, spatial-, and task-aware attention to improve robustness in complex scenes. Soft non-maximum suppression (NMS) further improves recognition in overlapping regions. Compared with YOLOv5s, YOLOv5-CJ improves the mean average precision at intersection over union (IoU) of 0.5 (mAP0.5) and the mean average precision averaged over IoU threshold from 0.5 to 0.95 (mAP0.5: 0.95) by 1.9% and 7.2% on the Northeastern University steel surface defect (NEU-DET) dataset, and by 5.3% and 4.3% on the GC10 steel surface defect (GC10-DET) dataset, respectively, demonstrating its effectiveness for industrial defect detection.
| [1] |
Zhang H K, Li S Q, Miao Q Q, et al. . Surface defect detection of hot rolled steel based on multi-scale feature fusion and attention mechanism residual block. Scientific reports, 2024, 14: 7671 J]
|
| [2] |
Leng Y F, Liu Z, Xu B Y, et al. . Detection of steel surface defect based on improved Faster R-CNN. Mechanical science and technology for aerospace engineering, 2025, 44(1): 75-83[J]
|
| [3] |
Dario G L, Lidia S G, Rubén U, et al. . Benchmarking deep learning models for surface defect detection: a reproducible and statistically-rigorous approach. Journal of intelligent manufacturing, 2026, 37(7): 3001-3018 J]
|
| [4] |
Yang Z, Liu Y. A steel surface defect detection method based on improved RetinaNet. Scientific reports, 2025, 156045 J]
|
| [5] |
Liang L M, Chen K Q, Chen L J, et al. . Improving the lightweight FCM-YOLOv8n for steel surface defect detection. Opto-electron engineering, 2025, 52(2): 240280[J]
|
| [6] |
Wang H, Zhan H F. Li-YOLO Net: a lightweight steel defect detection framework with dynamic feature selection and task alignment. Signal, image and video processing, 2025, 1911257 J]
|
| [7] |
Li J, Xu Z J, Xu L. Vehicle and pedestrian detection method based on improved YOLOv4-tiny. Optoelectronics letters, 2023, 1910623-628 J]
|
| [8] |
Chen G Z, Liu S, Xu J. Memory-boosting RNN with dynamic graph for event-based action recognition. Optoelectronics letters, 2023, 19(10): 629-634 J]
|
| [9] |
Wang P F, Huang H M, Wang M Q. Complex road target detection algorithm based on improved YOLOv5. Computer engineering and applications, 2022, 58(17): 81-92[J]
|
| [10] |
Li H L, Liu M, Yin Y F, et al. . Steel surface defect detection based on multi-layer fusion networks. Scientific reports, 2025, 1510371 J]
|
| [11] |
Zhou C D, Lu Z Y, Lv Z L, et al. . Metal surface defect detection based on improved YOLOv5. Scientific reports, 2023, 1320803 J]
|
| [12] |
Lu J B, Zhu M R, Ma X Y, et al. . Steel strip surface defect detection method based on improved YOLOv5s. Biomimetics, 2024, 9(1): 28 J]
|
| [13] |
Hu K T, Ma X H, Sun X Y, et al. . MSPC-Net: strip surface defect detection algorithm fused with Res2Net and partial convolution. Computer engineering and applications, 2025, 615334-342[J]
|
| [14] |
Dai X, Chen Y, Ao B, et al. . Dynamic head: unifying object detection heads with attentions. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 19–25, 2021, on line. 2021, New York, IEEE7373-7382[C]
|
| [15] |
Bodla N, Singh B, Chellappa R, et al. . Soft-NMS-improving object detection with one line of code. Proceedings of the IEEE International Conference on Computer Vision, October 22–29, 2017, Venice, Italy. 2017, New York, IEEE55615569[C]
|
RIGHTS & PERMISSIONS
Tianjin University of Technology
PDF
