YOLO12 and YOLO13 for Steel Surface Defect Detection: Experimental Evaluation

Adem Dilbaz

doi:10.58190/imiens.2026.175

Authors

Adem Dilbaz Sahin Tanker Limited Company, Asagipinarbasi Organized Industrial Zone, 521 Street 7, 42250, Konya, Türkiye https://orcid.org/0000-0002-3135-7032

DOI:

https://doi.org/10.58190/imiens.2026.175

Keywords:

deep learning, metal surface defect, object detection, yolo12, yolo13

Abstract

Intelligent inspection and defect detection of steel surfaces are essential for ensuring product quality, minimizing production losses, and supporting quality control in modern manufacturing systems. In particular, the identification of defects on steel plates is an important step, as surface imperfections can significantly affect the mechanical performance and commercial value of the final product. This study presents a comparative evaluation of YOLO12 and YOLO13 architectures for steel surface defect detection using the NEU-DET dataset under different validation ratios and input image resolutions. Four models (YOLO12-S, YOLO12-L, YOLO13-S, and YOLO13-L) were evaluated under three experimental settings using standard object detection metrics and three different random seeds. Experimental results showed that YOLO13-S consistently demonstrated superior overall detection performance, reaching a best single-run mAP@50 value of 0.753 with seed 42 and a precision value of 0.848 with seed 44. In addition, it achieved the highest single-run mAP@50–95 value of 0.449, with seed 43. In contrast, YOLO12-L achieved the highest single-run recall (0.749), with seed 42, indicating its stronger capability for detecting positive defect instances. Overall, the results demonstrate that YOLO13-S provides the best balance between detection accuracy and computational efficiency, while YOLO12-L offers superior recall performance under the evaluated conditions.

Downloads

Download data is not yet available.

References

[1] S. Wang, X. Xia, L. Ye, and B. Yang, "Automatic detection and classification of steel surface defects using deep convolutional neural networks," Metals, vol. 11, no. 3, pp. 388, 2021, doi: 10.3390/met11030388.

[2] I. Konovalenko, P. Maruschak, J. Brezinová, J. Viňáš, and J. Brezina, "Steel surface defect classification using deep residual neural network," Metals, vol. 10, pp. 846, 2020, doi: 10.3390/met10070846.

[3] K. Kripalani, "Experimental work on cast defect detection by nanoindentation machine using NiTinol wire sensors using augmented technique optimized by PP YOLOv3 algorithm comparative analysis procedure," Intelligent Methods in Engineering Sciences, vol. 2, no. 3, pp. 75–83, 2023.

[4] Y. Lin, Y. Zhang, D. Yang, W. Guan, D. Hou, Q. Wang, and W. Xing, "Steel surface defect detection based on YOLOv12," Engineering Letters, vol. 33, no. 12, pp. 4806–4817, 2025.

[5] Z. Fei, L. Liu, R. Zhao, C. Yang, Y. Li, and X. Xiang, "EDS-YOLO: An enhanced steel surface defect detection method based on improved YOLO11," Measurement Science and Technology, vol. 37, no. 13, p. 136204, 2026, doi: 10.1088/1361-6501/ae5271.

[6] T. Liu, "Enhanced Zero-Shot YOLOv10 for multi-class tiny-object detection of steel surface defects," in 2024 6th International Conference on Robotics and Computer Vision (ICRCV), 2024, pp. 44–52.

[7] W. Liu, S. Chen, S. Wang, W. Zhong, J. Wang, and H. Liao, "YOLOv13-EnhCrack for low-contrast steel surface crack detection," in International Conference on Machine Learning and Artificial Intelligence Applications (MLAIA 2025), SPIE, vol. 14134, Art. no. 1413402, 2026.

[8] I.A. Ozkan and A. Dilbaz, “Simulation-based spot welding inspection on automotive chassis using YOLO-powered image processing,” International Journal of Automotive Engineering and Technologies, vol. 14, no. 3, pp. 170–180, 2025, doi: 10.18245/ijaet.1729908.

[9] X. Ren, Z. Liu, H. Cheng, and L. Chen, "DDR-YOLO: An efficient detection model based on YOLO11s for steel surface defect," IEEE Access, vol. 13, pp. 147202–147217, 2025, doi: 10.1109/ACCESS.2025.3599978.

[10] X. Y. Lu and M. X. Qu, "Steel surface defect detection based on improved YOLOv8," Engineering Research Express, vol. 7, no. 1, 2025, doi: 10.1088/2631-8695/ad9b9d.

[11] H. Chen and J. Tang, "Performance comparison and analysis of YOLOv8/v9/v10 in steel plate surface defect detection," Science and Technology of Engineering, Chemistry and Environmental Protection, vol. 1, no. 1, 2025, doi: 10.61173/6bnb5p03.

[12] F. Wu and Y. Zhang, "A lightweight YOLOv11-based model for steel surface defect detection," Engineering Letters, vol. 34, no. 7, p. 2728, 2026.

[13] Q. Qin, A. S. M. Khairuddin, N. Idros, H. Liu, L. Fan, Z. Yang, J. Li, and C. Zhu, “Hierarchical depth aware YOLO for efficient metal surface defect detection,” Scientific Reports, vol. 16, Article 16248, 2026, doi: 10.1038/s41598-026-46074-z.

[14] J. Tang, Q. Zhang, J. Jiang, Y. Pan, and X. Ding, "CPFS-YOLO: A steel surface defect detection method based on cross-stage and multi-scale feature fusion," Measurement Science and Technology, vol. 36, no. 11, p. 115402, 2025, doi: 10.1088/1361-6501/ae1854.

[15] S. Wang, L. Zhang, and G. Yang, "Defect identification method for steel surfaces based on improved YOLOv5," Journal of Southeast University (English Edition), vol. 40, no. 1, 2024.

[16] W. Zhao, F. Chen, H. Huang, D. Li, and W. Cheng, "A new steel defect detection algorithm based on deep learning," Computational Intelligence and Neuroscience, vol. 2021, Art. no. 5592878, 2021, doi: 10.1155/2021/5592878.

[17] Y. Tian, Q. Ye, and D. Doermann, "YOLOv12: Attention-Centric Real-Time Object Detectors," in Advances in Neural Information Processing Systems, vol. 38, pp. 78433–78457, 2025, doi: 10.52202/085713-2627.

[18] M. Yurdakul and A. Çakmak, "Multi-scale performance benchmarking of YOLO models for effervescent tablet defect detection," Preprints.org, 2026, doi: 10.20944/preprints202604 .1475.v1.

[19] D. Saif, H. Askr, A. M. Sarhan, and A. E. Hassanien, "Enhanced YOLO12 with spatial pyramid pooling for real-time cotton insect detection," Scientific Reports, 2026.

[20] X. Zhou and S. Agaian, "A review of YOLO-based object detection for e-waste sorting and recycling," in Proc. SPIE, vol. 14027, Jun. 2026, Art. no. 140270O.

[21] Ö. Akar, H. Selim, and O. Er, "Hyperparameter optimization for enhancing personal protective equipment (PPE) detection performance using YOLO-based deep learning," Intelligent Data Analysis, pp. 1–15, 2026, doi: 10.1177/1088467X26144573

[22] L. Ye, X. Chen, J. Lei, and X. Chen, "RFDNet-YOLOv13: A lightweight YOLOv13n-based model with receptive-field and dynamic head for robust tobacco leaf disease detection," Smart Agricultural Technology, p. 102069, 2026.

[23] N. Köklü and S. A. Sulak, "Classification of environmental attitudes with artificial intelligence algorithms," Intelligent Methods in Engineering Sciences, vol. 3, no. 2, pp. 54–62, 2024.

[24] L. T. Ramos and A. D. Sappa, "A decade of You Only Look Once (YOLO) for object detection: A review," IEEE Access, vol. 13, pp. 192747–192794, 2025, doi: 10.1109/ACCESS.2025.3597504.

[25] C. M. Liu and J. C. Juang, "Systematic evaluation of YOLOv8 variants for UAV-based object detection," Applied Sciences, vol. 16, no. 7, p. 3559, 2026.