Page 65 - Read Online

Hu et al. J. Mater. Inf. 2025, 5, 44 https://dx.doi.org/10.20517/jmi.2025.21 Page 15 of 17

In this study, we present a self-supervised learning method for detecting surface defects in steel materials.
Our method achieves high accuracy in target detection on the NEU-DET steel defects dataset, without
reliance on a large amount of labeled data, with mAP and mAP_50 values of 0.3850 and 0.7680, respectively,
which demonstrates the effectiveness of the method. The above discussion highlights the value of self-
supervised learning for detecting surface defects in steels. While further improvements are possible in the
mAP_50 scores obtained in this study, our results demonstrate the potential of these methods as potential
alternatives or supplements to traditional supervised learning methods. Future research efforts ought to
concentrate on refining the pre-training process, exploring the integration of various datasets, and
developing sophisticated data enhancement techniques to enhance the capabilities of self-supervised models
in this domain. This study not only advances the state of the art in steel surface defect detection but also
provides guidance for constructing robust image analysis models with minimal reliance on labeled data
across related fields.

DECLARATIONS
Authors’ contributions
Made substantial contributions to conception and design of this review, writing and editing: Hu, S.; Zhang,
Y.; Xu, W.
Made substantial contributions to collation of literature, figures preparation, and writing: Hu, S.; Ma, X.;
Zhang, Y.; Xu, W.
Performed data analysis, discussion and writing review: Hu, S.; Ma, X.; Zhang, Y.; Xu, W.
Provided administrative, technical, and material support: Zhang, Y.; Xu, W.

Availability of data and materials
The original contributions presented in this study are included in the article/Supplementary Materials.
Further inquiries can be directed to the corresponding author(s).

Financial support and sponsorship
The research was supported by the National Key Research and Development Program of China (No.
2022YFB3707500) and the National Natural Science Foundation of China (No. 52304392).

Conflicts of interest
All authors declared that there are no conflicts of interest.

Ethical approval and consent to participate
Not applicable.

Consent for publication
Not applicable.

Copyright
© The Author(s) 2025.

REFERENCES
1. Ren, B.; Wang, C.; Zhang, Y.; Wei, X.; Xu, W. Industrial big data analysis strategy based on automatic data classification and
interpretable knowledge graph. J. Mater. Inf. 2025, 5, 2. DOI
2. Zhou, C.; Lu, Z.; Lv, Z.; et al. Metal surface defect detection based on improved YOLOv5. Sci. Rep. 2023, 13, 20803. DOI
3. Park, J. K.; Kwon, B. K.; Park, J. H.; Kang, D. J. Machine learning-based imaging system for surface defect inspection. Int. J. Precis.
Eng. Manuf. Green. Tech. 2016, 3, 303-10. DOI
4. Han, S.; Wang, C.; Zhang, Y.; Xu, W.; Di, H. Employing deep learning in non-parametric inverse visualization of elastic–plastic