Steel Surface Defect Detection Using Deep Learning: A Comparative Study of MobileNetV3Large and VGG19+InceptionV3 Ensemble
Keywords:
Surface defect detection, deep learning, transfer learning, MobileNetV3Large, VGG19, InceptionV3, ensemble learning, CLAHE, LIME, NEU dataset.Abstract
Automated surface defect detection in steel manufacturing is a critical quality-control challenge that directly impacts production efficiency and product reliability. This paper presents a comprehensive deep learning-based approach for classifying six types of steel surface defects—crazing, inclusion, patches, pitted surface, rolled-in scale, and scratches—using the NEU Surface Defect Database (1,440 images). We evaluate two architectures: (1) MobileNetV3Large with transfer learning, and (2) a dual-input ensemble of VGG19 and InceptionV3. An advanced preprocessing pipeline (CLAHE, Gaussian denoising, unsharp masking, normalization) and extensive data augmentation are applied. The ensemble achieves 99.07% accuracy, 99.10% precision, 99.07% recall, and 99.07% F1-score, outperforming MobileNetV3Large (89.81% accuracy) by ~9.3 percentage points. LIME explainability analysis validates the model's focus on semantically meaningful defect regions, making the framework suitable for industrial deployment.
References
Y. Shi, K. Cui, Z. Qi, F. Meng, and Z. Chen, "Automatic road crack detection using random structured forests," IEEE Trans. Intell. Transp. Syst., vol. 17, no. 12, pp. 3434–3445, 2018.
K. Song and Y. Yan, "A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defects," Appl. Surf. Sci., vol. 285, pp. 858–864, 2020.
Q. Luo, X. Fang, L. Liu, C. Yang, and Y. Sun, "Automated visual defect detection for flat steel surface: A survey," IEEE Trans. Instrum. Meas., vol. 69, no. 3, pp. 626–644, 2021.
Y.-J. Cha, W. Choi, and O. Büyüköztürk, "Deep learning-based crack damage detection using CNNs," Comput.-Aided Civ. Infrastruct. Eng., vol. 32, no. 5, pp. 361–378, 2019.
M. Tan and Q. Le, "EfficientNet: Rethinking model scaling for CNNs," in Proc. ICML, 2019, pp. 6105–6114.
D. Tabernik, S. Šela, J. Skvarč, and D. Skočaj, "Segmentation-based deep-learning approach for surface-defect detection," J. Intell. Manuf., vol. 31, no. 3, pp. 759–776, 2020.
K. Dikshit, "NEU Surface Defect Database," Kaggle, 2020. [Online]. Available: https://www.kaggle.com/datasets/kaustubhdikshit/neu-surface-defect-database
M. T. Ribeiro, S. Singh, and C. Guestrin, "'Why should I trust you?': Explaining the predictions of any classifier," in Proc. ACM SIGKDD, 2016, pp. 1135–1144.
K. Simonyan and A. Zisserman, "Very deep convolutional networks for large-scale image recognition," in Proc. ICLR, 2015.
C. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens, and Z. Wojna, "Rethinking the inception architecture for computer vision," in Proc. IEEE CVPR, 2016, pp. 2818–2826.
A. G. Howard et al., "Searching for MobileNetV3," in Proc. IEEE ICCV, 2019, pp. 1314–1324.
I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning. Cambridge, MA: MIT Press, 2016.
Downloads
How to Cite
Issue
Section
License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
