YANG, Wenjia ;ZHOU, Youhang ;MENG, Gaolei ;LI, Yuze ;GONG, Tianyu . Improving the Efficiency of Steel Plate Surface Defect Classification by Reducing the Labelling Cost Using Deep Active Learning. Strojniški vestnik - Journal of Mechanical Engineering, [S.l.], v. 70, n.11-12, p. 554-568, october 2024. ISSN 0039-2480. Available at: <https://www.sv-jme.eu/article/improving-the-efficiency-of-steel-plate-surface-defect-classification-by-reducing-the-labeling-cost-using-deep-active-learning-method/>. Date accessed: 23 dec. 2024. doi:http://dx.doi.org/10.5545/sv-jme.2023.900.
Yang, W., Zhou, Y., Meng, G., Li, Y., & Gong, T. (2024). Improving the Efficiency of Steel Plate Surface Defect Classification by Reducing the Labelling Cost Using Deep Active Learning. Strojniški vestnik - Journal of Mechanical Engineering, 70(11-12), 554-568. doi:http://dx.doi.org/10.5545/sv-jme.2023.900
@article{sv-jmesv-jme.2023.900,
author = {Wenjia Yang and Youhang Zhou and Gaolei Meng and Yuze Li and Tianyu Gong},
title = {Improving the Efficiency of Steel Plate Surface Defect Classification by Reducing the Labelling Cost Using Deep Active Learning},
journal = {Strojniški vestnik - Journal of Mechanical Engineering},
volume = {70},
number = {11-12},
year = {2024},
keywords = {surface defect classification; multiscale convolutional neural networks; active learning; global pooling; },
abstract = {Efficient surface defects classification is one of the research hotpots in steel plate defect recognition. Compared with traditional methods, deep learning methods have been effective in improving classification accuracy and efficiency, but require a large amount of labeled data, resulting in limited improvement of detection efficiency. To reduce the labeling effort under the premise of satisfying the classification accuracy, a deep active learning method is proposed for steel plate surface defects classification. Firstly, a lightweight convolutional neural network is designed, which speeds up the training process and enhances the model regularization. Secondly, a novel uncertainty-based sampling strategy, which calculates Kullback-Leibler (KL) divergence between two kinds of distributions, is used as an uncertainty measure to select new samples for labeling. Finally, the performance of the proposed method is validated using the steel surface defects dataset from Northeastern University (NEU-CLS) and the milling steel surface defects dataset from a local laboratory. The proposed global pooling-based classifier with global average pooling (GAPC) network model combined with the Kullback-Leibler divergence sampling (KLS) strategy has the best performance in the classification of steel plate surface defects. This method achieves 97 % classification accuracy with 44 % labeled data on the NEU-CLS dataset and 92.3 % classification accuracy with 50 % labeled data on the milling steel surface defects dataset. The experimental results show that the proposed method can achieve steel surface defects classification accuracy of not less than 92 % with no more than 50 % of the dataset to be labeled, which indicates that this method has potential application in surface defect classification of industrial products.},
issn = {0039-2480}, pages = {554-568}, doi = {10.5545/sv-jme.2023.900},
url = {https://www.sv-jme.eu/article/improving-the-efficiency-of-steel-plate-surface-defect-classification-by-reducing-the-labeling-cost-using-deep-active-learning-method/}
}
Yang, W.,Zhou, Y.,Meng, G.,Li, Y.,Gong, T. 2024 October 70. Improving the Efficiency of Steel Plate Surface Defect Classification by Reducing the Labelling Cost Using Deep Active Learning. Strojniški vestnik - Journal of Mechanical Engineering. [Online] 70:11-12
%A Yang, Wenjia %A Zhou, Youhang %A Meng, Gaolei %A Li, Yuze %A Gong, Tianyu %D 2024 %T Improving the Efficiency of Steel Plate Surface Defect Classification by Reducing the Labelling Cost Using Deep Active Learning %B 2024 %9 surface defect classification; multiscale convolutional neural networks; active learning; global pooling; %! Improving the Efficiency of Steel Plate Surface Defect Classification by Reducing the Labelling Cost Using Deep Active Learning %K surface defect classification; multiscale convolutional neural networks; active learning; global pooling; %X Efficient surface defects classification is one of the research hotpots in steel plate defect recognition. Compared with traditional methods, deep learning methods have been effective in improving classification accuracy and efficiency, but require a large amount of labeled data, resulting in limited improvement of detection efficiency. To reduce the labeling effort under the premise of satisfying the classification accuracy, a deep active learning method is proposed for steel plate surface defects classification. Firstly, a lightweight convolutional neural network is designed, which speeds up the training process and enhances the model regularization. Secondly, a novel uncertainty-based sampling strategy, which calculates Kullback-Leibler (KL) divergence between two kinds of distributions, is used as an uncertainty measure to select new samples for labeling. Finally, the performance of the proposed method is validated using the steel surface defects dataset from Northeastern University (NEU-CLS) and the milling steel surface defects dataset from a local laboratory. The proposed global pooling-based classifier with global average pooling (GAPC) network model combined with the Kullback-Leibler divergence sampling (KLS) strategy has the best performance in the classification of steel plate surface defects. This method achieves 97 % classification accuracy with 44 % labeled data on the NEU-CLS dataset and 92.3 % classification accuracy with 50 % labeled data on the milling steel surface defects dataset. The experimental results show that the proposed method can achieve steel surface defects classification accuracy of not less than 92 % with no more than 50 % of the dataset to be labeled, which indicates that this method has potential application in surface defect classification of industrial products. %U https://www.sv-jme.eu/article/improving-the-efficiency-of-steel-plate-surface-defect-classification-by-reducing-the-labeling-cost-using-deep-active-learning-method/ %0 Journal Article %R 10.5545/sv-jme.2023.900 %& 554 %P 15 %J Strojniški vestnik - Journal of Mechanical Engineering %V 70 %N 11-12 %@ 0039-2480 %8 2024-10-08 %7 2024-10-08
Yang, Wenjia, Youhang Zhou, Gaolei Meng, Yuze Li, & Tianyu Gong. "Improving the Efficiency of Steel Plate Surface Defect Classification by Reducing the Labelling Cost Using Deep Active Learning." Strojniški vestnik - Journal of Mechanical Engineering [Online], 70.11-12 (2024): 554-568. Web. 23 Dec. 2024
TY - JOUR AU - Yang, Wenjia AU - Zhou, Youhang AU - Meng, Gaolei AU - Li, Yuze AU - Gong, Tianyu PY - 2024 TI - Improving the Efficiency of Steel Plate Surface Defect Classification by Reducing the Labelling Cost Using Deep Active Learning JF - Strojniški vestnik - Journal of Mechanical Engineering DO - 10.5545/sv-jme.2023.900 KW - surface defect classification; multiscale convolutional neural networks; active learning; global pooling; N2 - Efficient surface defects classification is one of the research hotpots in steel plate defect recognition. Compared with traditional methods, deep learning methods have been effective in improving classification accuracy and efficiency, but require a large amount of labeled data, resulting in limited improvement of detection efficiency. To reduce the labeling effort under the premise of satisfying the classification accuracy, a deep active learning method is proposed for steel plate surface defects classification. Firstly, a lightweight convolutional neural network is designed, which speeds up the training process and enhances the model regularization. Secondly, a novel uncertainty-based sampling strategy, which calculates Kullback-Leibler (KL) divergence between two kinds of distributions, is used as an uncertainty measure to select new samples for labeling. Finally, the performance of the proposed method is validated using the steel surface defects dataset from Northeastern University (NEU-CLS) and the milling steel surface defects dataset from a local laboratory. The proposed global pooling-based classifier with global average pooling (GAPC) network model combined with the Kullback-Leibler divergence sampling (KLS) strategy has the best performance in the classification of steel plate surface defects. This method achieves 97 % classification accuracy with 44 % labeled data on the NEU-CLS dataset and 92.3 % classification accuracy with 50 % labeled data on the milling steel surface defects dataset. The experimental results show that the proposed method can achieve steel surface defects classification accuracy of not less than 92 % with no more than 50 % of the dataset to be labeled, which indicates that this method has potential application in surface defect classification of industrial products. UR - https://www.sv-jme.eu/article/improving-the-efficiency-of-steel-plate-surface-defect-classification-by-reducing-the-labeling-cost-using-deep-active-learning-method/
@article{{sv-jme}{sv-jme.2023.900},
author = {Yang, W., Zhou, Y., Meng, G., Li, Y., Gong, T.},
title = {Improving the Efficiency of Steel Plate Surface Defect Classification by Reducing the Labelling Cost Using Deep Active Learning},
journal = {Strojniški vestnik - Journal of Mechanical Engineering},
volume = {70},
number = {11-12},
year = {2024},
doi = {10.5545/sv-jme.2023.900},
url = {https://www.sv-jme.eu/article/improving-the-efficiency-of-steel-plate-surface-defect-classification-by-reducing-the-labeling-cost-using-deep-active-learning-method/}
}
TY - JOUR AU - Yang, Wenjia AU - Zhou, Youhang AU - Meng, Gaolei AU - Li, Yuze AU - Gong, Tianyu PY - 2024/10/08 TI - Improving the Efficiency of Steel Plate Surface Defect Classification by Reducing the Labelling Cost Using Deep Active Learning JF - Strojniški vestnik - Journal of Mechanical Engineering; Vol 70, No 11-12 (2024): Strojniški vestnik - Journal of Mechanical Engineering DO - 10.5545/sv-jme.2023.900 KW - surface defect classification, multiscale convolutional neural networks, active learning, global pooling, N2 - Efficient surface defects classification is one of the research hotpots in steel plate defect recognition. Compared with traditional methods, deep learning methods have been effective in improving classification accuracy and efficiency, but require a large amount of labeled data, resulting in limited improvement of detection efficiency. To reduce the labeling effort under the premise of satisfying the classification accuracy, a deep active learning method is proposed for steel plate surface defects classification. Firstly, a lightweight convolutional neural network is designed, which speeds up the training process and enhances the model regularization. Secondly, a novel uncertainty-based sampling strategy, which calculates Kullback-Leibler (KL) divergence between two kinds of distributions, is used as an uncertainty measure to select new samples for labeling. Finally, the performance of the proposed method is validated using the steel surface defects dataset from Northeastern University (NEU-CLS) and the milling steel surface defects dataset from a local laboratory. The proposed global pooling-based classifier with global average pooling (GAPC) network model combined with the Kullback-Leibler divergence sampling (KLS) strategy has the best performance in the classification of steel plate surface defects. This method achieves 97 % classification accuracy with 44 % labeled data on the NEU-CLS dataset and 92.3 % classification accuracy with 50 % labeled data on the milling steel surface defects dataset. The experimental results show that the proposed method can achieve steel surface defects classification accuracy of not less than 92 % with no more than 50 % of the dataset to be labeled, which indicates that this method has potential application in surface defect classification of industrial products. UR - https://www.sv-jme.eu/article/improving-the-efficiency-of-steel-plate-surface-defect-classification-by-reducing-the-labeling-cost-using-deep-active-learning-method/
Yang, Wenjia, Zhou, Youhang, Meng, Gaolei, Li, Yuze, AND Gong, Tianyu. "Improving the Efficiency of Steel Plate Surface Defect Classification by Reducing the Labelling Cost Using Deep Active Learning" Strojniški vestnik - Journal of Mechanical Engineering [Online], Volume 70 Number 11-12 (08 October 2024)
Strojniški vestnik - Journal of Mechanical Engineering 70(2024)11-12, 554-568
© The Authors 2024. CC BY 4.0 Int.
Efficient surface defects classification is one of the research hotpots in steel plate defect recognition. Compared with traditional methods, deep learning methods have been effective in improving classification accuracy and efficiency, but require a large amount of labeled data, resulting in limited improvement of detection efficiency. To reduce the labeling effort under the premise of satisfying the classification accuracy, a deep active learning method is proposed for steel plate surface defects classification. Firstly, a lightweight convolutional neural network is designed, which speeds up the training process and enhances the model regularization. Secondly, a novel uncertainty-based sampling strategy, which calculates Kullback-Leibler (KL) divergence between two kinds of distributions, is used as an uncertainty measure to select new samples for labeling. Finally, the performance of the proposed method is validated using the steel surface defects dataset from Northeastern University (NEU-CLS) and the milling steel surface defects dataset from a local laboratory. The proposed global pooling-based classifier with global average pooling (GAPC) network model combined with the Kullback-Leibler divergence sampling (KLS) strategy has the best performance in the classification of steel plate surface defects. This method achieves 97 % classification accuracy with 44 % labeled data on the NEU-CLS dataset and 92.3 % classification accuracy with 50 % labeled data on the milling steel surface defects dataset. The experimental results show that the proposed method can achieve steel surface defects classification accuracy of not less than 92 % with no more than 50 % of the dataset to be labeled, which indicates that this method has potential application in surface defect classification of industrial products.