ODNet: A High Real-Time Network Using Orthogonal Decomposition for Few-Shot Strip Steel Surface Defect Classification
<p>Episodes’ training process. The testing phase presents a few-shot classification task.</p> "> Figure 2
<p>ODNet architecture, where colored boxes represent orthogonal features.</p> "> Figure 3
<p>The pipeline of the feature extractor. Layer1, Layer2, Layer3, and Layer4 have the same structure.</p> "> Figure 4
<p>Orthogonal decomposition process.</p> "> Figure 5
<p>Data flow of ODNet in case of the 5-way, 5-shot. The black dotted line indicates the training stage, and the red dotted line represents the concrete steps of the orthogonal decomposition operation.</p> "> Figure 6
<p>Examples for each class in the FSC-20. The classes in the blue box are cold-rolled defects, and the classes in the red box are hot-rolled defects.</p> "> Figure 7
<p>The double <span class="html-italic">Y</span>-axis histogram–line chart of the model with real time(s) and accuracy(%). The histogram represents the time it takes for the model to run an episode corresponding to the left <span class="html-italic">Y</span>-axis. The blue and red dots in the line chart represent the intra-domain and cross-domain accuracy of the model corresponding to the right <span class="html-italic">Y</span>-axis, respectively. (<b>a</b>), The real time and accuracy of the model in the case of 5-way, 1-shot. (<b>b</b>), The real time and accuracy of the model in the case of 5-way 5-shot.</p> "> Figure 8
<p>Comparison showing the effect of <span class="html-italic">N</span> and <span class="html-italic">K</span> for the ODNet. The blue line and red line represent the 3-way and 5-way, respectively. The <span class="html-italic">X</span>-axis indicates values of the <span class="html-italic">K</span>-shot. The <span class="html-italic">Y</span>-axis indicates the test accuracy.</p> "> Figure 9
<p>ODNet’s accuracy and real-time statistics.</p> ">
Abstract
:1. Introduction
- A high real-time network for few-shot strip steel surface defect classification is proposed.
- ODNet employs orthogonal decomposition to derive orthogonal features, thereby minimizing the impact of redundant information on the model. The inclusion of a skip connection ensures that the valuable correlation information remains intact, especially after orthogonal decomposition.
- The features extracted by the model with orthogonality also adhere more closely to the orthogonality requirement of the Euclidean distance on input, thereby enhancing the classifier performance.
- Compared to alternative methods, ODNet exhibits superior real-time performance, precision, and generalization, aligning more closely with the specific demands of industrial production.
2. Methodology
2.1. Problem Definition
2.2. ODNet
2.2.1. Feature Extractor
2.2.2. Orthogonal Decomposition
Algorithm 1: Define sample x using ResNet18 to obtain X. Two fully connected layers: FC_1 (▪) and FC_2 (▪). is the feature obtained from x after undergoing the operations of orthogonal decomposition and skip connection. |
|
2.2.3. Classifier
Algorithm 2: For an episode, is the number of all classes including the support set and query set, is the number of samples in each class of the support set, is the number of samples of each class in the query set, is the set of K-th samples in the support set, is the query samples collection, and J is the loss function. denotes the feature extractor, and d denotes the Euclidean distance. |
|
3. Experiment
3.1. Dataset and Implementation
3.2. Precision
3.2.1. Intra-Domain Results
3.2.2. Cross-Domain Results
3.2.3. Real-Time Results
3.3. Ablation
3.3.1. Module Results
3.3.2. Classifier Results
3.3.3. N and K Results
- With the increase in the number of samples, the model’s performance tends to saturate;
- The increase in the number of classes increases the classification challenge for the model. However, as the sample number increases, this difficulty becomes negligible.
4. Discussion
5. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Cumbajin, E.; Rodrigues, N.; Costa, P.; Miragaia, R.; Frazão, L.; Costa, N.; Fernández-Caballero, A.; Carneiro, J.; Buruberri, L.H.; Pereira, A. A systematic review on deep learning with CNNs applied to surface defect detection. J. Imaging 2023, 9, 193. [Google Scholar] [CrossRef] [PubMed]
- Ma, J.; Zhang, T.; Yang, C.; Cao, Y.; Xie, L.; Tian, H.; Li, X. Review of Wafer Surface Defect Detection Methods. Electronics 2023, 12, 1787. [Google Scholar] [CrossRef]
- Liu, Y.; Zhang, C.; Dong, X. A survey of real-time surface defect inspection methods based on deep learning. Artif. Intell. Rev. 2023, 56, 12131–12170. [Google Scholar] [CrossRef]
- Tang, B.; Chen, L.; Sun, W.; Lin, Z.K. Review of surface defect detection of steel products based on machine vision. IET Image Processing 2023, 17, 303–322. [Google Scholar] [CrossRef]
- Shen, K.; Zhou, X.; Liu, Z. MINet: Multiscale Interactive Network for Real-Time Salient Object Detection of Strip Steel Surface Defects. IEEE Trans. Ind. Inform. 2024, 20, 7842–7852. [Google Scholar] [CrossRef]
- Dong, Y.; Xie, C.; Xu, L.; Cai, H.; Shen, W.; Tang, H. Generative and Contrastive Combined Support Sample Synthesis Model for Few/Zero-Shot Surface Defect Recognition. IEEE Trans. Instrum. Meas. 2023, 73, 5010515. [Google Scholar] [CrossRef]
- Hu, K.; Chen, Z.; Kang, H.; Tang, Y. 3D vision technologies for a self-developed structural external crack damage recognition robot. Autom. Constr. 2024, 159, 105262. [Google Scholar] [CrossRef]
- Lei, B.; Yi, P.; Xiang, J. A new defect classification approach based on the fusion matrix of multi-eigenvalue. IEEE Sensors J. 2020, 21, 3398–3407. [Google Scholar] [CrossRef]
- Liu, Y.; Yuan, Y.; Liu, J. Deep learning model for imbalanced multi-label surface defect classification. Meas. Sci. Technol. 2021, 33, 035601. [Google Scholar] [CrossRef]
- Sampath, V.; Maurtua, I.; Martín, J.J.A.; Rivera, A.; Molina, J.; Gutierrez, A. Attention guided multi-task learning for surface defect identification. IEEE Trans. Ind. Inform. 2023, 99, 1–9. [Google Scholar]
- Feng, X.; Gao, X.; Luo, L. A ResNet50-based method for classifying surface defects in hot-rolled strip steel. Mathematics 2021, 9, 2359. [Google Scholar] [CrossRef]
- Zhang, H.; Sun, Q.; Xu, K. A Self-Supervised Model Based on CutPaste-Mix for Ductile Cast Iron Pipe Surface Defect Classification. Sensors 2023, 23, 8243. [Google Scholar] [CrossRef] [PubMed]
- Chen, G.; Yu, H.; Jiang, L.; Shang, H. Few-Shot Learning on 3D Surface Defect Detection with PM Networks. In Proceedings of the 2021 3rd International Conference on Artificial Intelligence and Advanced Manufacture, Manchester, UK, 23–25 October 2021; pp. 104–108. [Google Scholar]
- Ma, Z.; Li, Y.; Huang, M.; Deng, N. Online visual end-to-end detection monitoring on surface defect of aluminum strip under the industrial few-shot condition. J. Manuf. Syst. 2023, 70, 31–47. [Google Scholar] [CrossRef]
- Li, J.; Su, Z.; Geng, J.; Yin, Y. Real-time detection of steel strip surface defects based on improved yolo detection network. IFAC-PapersOnLine 2018, 51, 76–81. [Google Scholar] [CrossRef]
- Wu, S.; Zhao, S.; Zhang, Q.; Chen, L.; Wu, C. Steel Surface defect classification based on small sample learning. Appl. Sci. 2021, 11, 11459. [Google Scholar] [CrossRef]
- Wan, S.; Guan, S.; Tang, Y. Advancing bridge structural health monitoring: Insights into knowledge-driven and data-driven approaches. J. Data Sci. Intell. Syst. 2023. [Google Scholar] [CrossRef]
- Guo, R.; Chen, Q.; Liu, H.; Wang, W. Adversarial Robustness Enhancement for Deep Learning-Based Soft Sensors: An Adversarial Training Strategy Using Historical Gradients and Domain Adaptation. Sensors 2024, 24, 3909. [Google Scholar] [CrossRef] [PubMed]
- Li, Z.; Gao, L.; Gao, Y.; Li, X.; Li, H. Zero-shot surface defect recognition with class knowledge graph. Adv. Eng. Inform. 2022, 54, 101813. [Google Scholar] [CrossRef]
- Ma, S.; Song, K.; Niu, M.; Tian, H.; Wang, Y.; Yan, Y. Shape-Consistent One-Shot Unsupervised Domain Adaptation for Rail Surface Defect Segmentation. IEEE Trans. Ind. Inform. 2023, 19, 9667–9679. [Google Scholar] [CrossRef]
- Yu, R.; Guo, B.; Yang, K. Selective prototype network for few-shot metal surface defect segmentation. IEEE Trans. Instrum. Meas. 2022, 71, 5020010. [Google Scholar] [CrossRef]
- Liu, Z.; Guo, Z.; Li, C.; Gao, C.; Huang, N. Few-shot Steel Surface Defect Detection Based on Meta Learning. In Proceedings of the 2021 10th International Conference on Computing and Pattern Recognition, Shanghai, China, 15–17 October 2021; pp. 113–119. [Google Scholar]
- Liu, Z.; Song, Y.; Tang, R.; Duan, G.; Tan, J. Few-shot defect recognition of metal surfaces via attention-embedding and self-supervised learning. J. Intell. Manuf. 2023, 34, 3507–3521. [Google Scholar] [CrossRef]
- Goodfellow, I.; Pouget-Abadie, J.; Mirza, M.; Xu, B.; Warde-Farley, D.; Ozair, S.; Courville, A.; Bengio, Y. Generative adversarial networks. Commun. ACM 2020, 63, 139–144. [Google Scholar] [CrossRef]
- Yun, S.; Han, D.; Oh, S.J.; Chun, S.; Choe, J.; Yoo, Y. CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features. arXiv 2019, arXiv:1905.04899. Available online: http://arxiv.org/abs/1905.04899 (accessed on 30 June 2024).
- Liu, X.; Teng, W.; Liu, Y. A Model-Agnostic Meta-Baseline Method for Few-Shot Fault Diagnosis of Wind Turbines. Sensors 2022, 22, 3288. [Google Scholar] [CrossRef] [PubMed]
- Gomes, J.C.; Borges, L.d.A.B.; Borges, D.L. A Multi-Layer Feature Fusion Method for Few-Shot Image Classification. Sensors 2023, 23, 6880. [Google Scholar] [CrossRef] [PubMed]
- Gong, Y.; Wang, X.; Zhou, C.; Ge, M.; Liu, C.; Zhang, X. Human-machine knowledge hybrid augmentation method for surface defect detection based few-data learning. J. Intell. Manuf. 2024, 1–20. [Google Scholar] [CrossRef]
- Zhang, K.; Xiao, Y.; Wang, J.; Du, M.; Guo, X.; Zhou, R.; Shi, C.; Zhao, Z. DP-GAN: A Transmission Line Bolt Defects Generation Network Based on Dual Discriminator Architecture and Pseudo-Enhancement Strategy. IEEE Trans. Power Deliv. 2024, 39, 1622–1633. [Google Scholar] [CrossRef]
- Duan, G.; Song, Y.; Liu, Z.; Ling, S.; Tan, J. Cross-domain few-shot defect recognition for metal surfaces. Meas. Sci. Technol. 2023, 34, 015202. [Google Scholar] [CrossRef]
- Zhang, H.; Liu, H.; Liang, L.; Ma, W.; Liu, D. BiLSTM-TANet: An adaptive diverse scenes model with context embeddings for few-shot learning. Appl. Intell. 2024, 54, 5097–5116. [Google Scholar] [CrossRef]
- Deshpande, A.M.; Minai, A.A.; Kumar, M. One-shot recognition of manufacturing defects in steel surfaces. Procedia Manuf. 2020, 48, 1064–1071. [Google Scholar] [CrossRef]
- Finn, C.; Abbeel, P.; Levine, S. Model-agnostic meta-learning for fast adaptation of deep networks. In Proceedings of the International Conference on Machine Learning, PMLR, Sydney, NSW, Australia, 6–11 August 2017; pp. 1126–1135. [Google Scholar]
- Pang, S.; Zhang, L.; Yuan, Y.; Zhao, W.; Wang, S.; Wang, S. Adaptive-MAML: Few-shot metal surface defects diagnosis based on model-agnostic meta-learning. Measurement 2023, 223, 113612. [Google Scholar] [CrossRef]
- Pang, S.; Zhao, W.; Wang, S.; Zhang, L.; Wang, S. Permute-MAML: Exploring industrial surface defect detection algorithms for few-shot learning. Complex Intell. Syst. 2024, 10, 1473–1482. [Google Scholar] [CrossRef]
- Guo, R.; Liu, H.; Xie, G.; Zhang, Y.; Liu, D. A self-interpretable soft sensor based on deep learning and multiple attention mechanism: From data selection to sensor modeling. IEEE Trans. Ind. Inform. 2023, 19, 6859–6871. [Google Scholar] [CrossRef]
- Bao, Y.; Song, K.; Liu, J.; Wang, Y.; Yan, Y.; Yu, H.; Li, X. Triplet-graph reasoning network for few-shot metal generic surface defect segmentation. IEEE Trans. Instrum. Meas. 2021, 70, 5011111. [Google Scholar] [CrossRef]
- Shan, D.; Zhang, Y.; Coleman, S.; Kerr, D.; Liu, S.; Hu, Z. Unseen-material few-shot defect segmentation with optimal bilateral feature transport network. IEEE Trans. Ind. Inform. 2022, 19, 8072–8082. [Google Scholar] [CrossRef]
- Zhang, J.; Li, S.; Yan, Y.; Ni, Z.; Ni, H. Surface Defect Classification of Steel Strip with Few Samples Based on Dual-Stream Neural Network. Steel Res. Int. 2022, 93, 2100554. [Google Scholar] [CrossRef]
- Wang, W.; Wu, Z.; Lu, K.; Long, H.; Li, D.; Zhang, J.; Chen, P.; Wang, B. Surface Defects Classification of Hot Rolled Strip Based on Few-shot Learning. ISIJ Int. 2022, 62, 1222–1226. [Google Scholar] [CrossRef]
- Yu, J.; Liu, K.; Qin, L.; Li, Q.; Zhao, F.; Wang, Q.; Liu, H.; Li, B.; Wang, J.; Li, K. DMnet: A New Few-Shot Framework for Wind Turbine Surface Defect Detection. Machines 2022, 10, 487. [Google Scholar] [CrossRef]
- Zhang, P.; Zheng, P.; Guo, X.; Chen, E. Few-shot defect classification via feature aggregation based on graph neural network. J. Vis. Commun. Image Represent. 2024, 101, 104172. [Google Scholar] [CrossRef]
- Zhao, J.; Qian, X.; Zhang, Y.; Shan, D.; Liu, X.; Coleman, S.; Kerr, D. A knowledge distillation-based multi-scale relation-prototypical network for cross-domain few-shot defect classification. J. Intell. Manuf. 2024, 35, 841–857. [Google Scholar] [CrossRef]
- Feng, H.; Song, K.; Cui, W.; Zhang, Y.; Yan, Y. Cross Position Aggregation Network for Few-Shot Strip Steel Surface Defect Segmentation. IEEE Trans. Instrum. Meas. 2023, 72, 5007410. [Google Scholar] [CrossRef]
- Gao, P.; Wang, J.; Xia, M.; Qin, Z.; Zhang, J. Dual-Metric Neural Network with Attention Guidance for Surface Defect Few-Shot Detection in Smart Manufacturing. J. Manuf. Sci.-Eng.-Trans. Asme 2023, 145, 121010. [Google Scholar] [CrossRef]
- Kang, G.W.; Liu, H.B. Surface defects inspection of cold rolled strips based on neural network. In Proceedings of the 2005 International Conference on Machine Learning and Cybernetics, Guangzhou, China, 18–21 August 2005. [Google Scholar]
- Sun, Q.; Cai, J.; Sun, Z. Detection of surface defects on steel strips based on singular value decomposition of digital image. Math. Probl. Eng. 2016, 2016, 5797654. [Google Scholar] [CrossRef]
- He, K.; Zhang, X.; Ren, S.; Sun, J. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 27–30 June 2016; pp. 770–778. [Google Scholar]
- Zhao, W.; Song, K.; Wang, Y.; Liang, S.; Yan, Y. FaNet: Feature-aware network for few shot classification of strip steel surface defects. Measurement 2023, 208, 112446. [Google Scholar] [CrossRef]
- Song, K.; Yan, Y. A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defects. Appl. Surf. Sci. 2013, 285, 858–864. [Google Scholar] [CrossRef]
- Feng, X.; Gao, X.; Luo, L. X-SDD: A new benchmark for hot rolled steel strip surface defects detection. Symmetry 2021, 13, 706. [Google Scholar] [CrossRef]
- Lv, X.; Duan, F.; Jiang, J.j.; Fu, X.; Gan, L. Deep metallic surface defect detection: The new benchmark and detection network. Sensors 2020, 20, 1562. [Google Scholar] [CrossRef]
- Ziko, I.; Dolz, J.; Granger, E.; Ayed, I.B. Laplacian regularized few-shot learning. In Proceedings of the International Conference on Machine Learning, PMLR, Virtual, 13–18 July 2020; pp. 11660–11670. [Google Scholar]
- Wang, Y.; Xu, C.; Liu, C.; Zhang, L.; Fu, Y. Instance credibility inference for few-shot learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, WA, USA, 13–19 June 2020; pp. 12836–12845. [Google Scholar]
- Zhang, C.; Cai, Y.; Lin, G.; Shen, C. Deepemd: Few-shot image classification with differentiable earth mover’s distance and structured classifiers. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, WA, USA, 13–19 June 2020; pp. 12203–12213. [Google Scholar]
- Snell, J.; Swersky, K.; Zemel, R. Prototypical networks for few-shot learning. Adv. Neural Inf. Process. Syst. 2017, 30. [Google Scholar]
- Boudiaf, M.; Masud, Z.; Rony, J.; Dolz, J.; Piantanida, P.; Ayed, I. Transductive information maximization for few-shot learning. arXiv 2020, arXiv:2008.11297. [Google Scholar]
- Chen, W.Y.; Liu, Y.C.; Kira, Z.; Wang, Y.C.F.; Huang, J.B. A closer look at few-shot classification. arXiv 2019, arXiv:1904.04232. [Google Scholar]
- Xiao, W.; Song, K.; Liu, J.; Yan, Y. Graph embedding and optimal transport for few-shot classification of metal surface defect. IEEE Trans. Instrum. Meas. 2022, 71, 5010310. [Google Scholar] [CrossRef]
Feature Extractor | Stage | Detail |
---|---|---|
ResNet18 | Pre-processing | Conv2d (, stride 2, pad 1, BatchNorm, RelU) |
Each block | Conv2d (, stride 2, pad 1) | |
BatchNorm | ||
ReLU | ||
Conv2d (, stride 2, pad 1) | ||
BatchNorm | ||
Sum with Input | ||
ReLU | ||
Post-processing | AvgPool, Flatten | |
Orthogonal decomposition layer | Orthogonal decomposition | Fully connected layer () |
Skip connection | Fully connected layer () | |
Sum with Input |
Training Set (50%) [1] | Validation Set (25%) [1] | Testing Set (25%) [1] |
---|---|---|
Crescent gap | Welding line | One inclusion |
Oil spot | Water spot | Waist folding |
Rolled pit | Silk spot | Crazing |
Crease | Rolled in scale | Patches |
Punching hole | Iron sheet ash | Red iron |
Scratches | - | - |
Pitted surface | - | - |
Two inclusion | - | - |
Oxide scale | - | - |
Slag inclusions | - | - |
Training Epochs | Query Sample [1] | Regularizer Constant | Learning Rate | Decay Rate | Decay Episodes | Test Episodes [2] | |
---|---|---|---|---|---|---|---|
Train | Test | ||||||
150 | 5 | 15 | 0.5 | 0.5 | 2000 | 1000 |
Method | 5-Way | |
---|---|---|
1-Shot | 5-Shot | |
LaplacianShot [53] | 79.86 ± 0.11 | 87.83 ± 0.08 |
ICI [54] | 63.50 ± 0.66 | 72.86 ± 0.51 |
DeepEMD [55] | 62.62 ± 0.67 | 71.10 ± 0.45 |
Prototypical Nets [56] | 43.31 ± 0.34 | 80.29 ± 0.31 |
TIM [57] | 71.72 ± 0.13 | 81.27 ± 0.09 |
Baseline [58] | 67.72 ± 0.13 | 81.97 ± 0.10 |
GTNet [59] | 76.76 ± 0.19 | 85.56 ± 0.08 |
ODNet (proposed) | 80.45 ± 0.47 | 93.41 ± 0.26 |
Method | Cold-Rolled → Hot-Rolled | |
---|---|---|
5-Way, 1-Shot | 5-Way, 5-Shot | |
LaplacianShot [53] | 59.90 ± 0.19 | 71.05 ± 0.13 |
ICI [54] | 49.64 ± 0.32 | 71.62 ± 0.18 |
Prototypical Networks [56] | 49.59 ± 0.37 | 75.13 ± 0.56 |
DeepEMD [55] | 66.98 ± 0.56 | 80.80 ± 0.45 |
TIM [57] | 70.11 ± 0.17 | 86.05 ± 0.08 |
Baseline [58] | 67.12 ± 0.13 | 81.97 ± 0.10 |
GTNet [59] | 77.61 ± 0.21 | 87.95 ± 0.08 |
OdNet (proposed) | 80.32 ± 0.22 | 86.52 ± 0.31 |
Method | Time(s) | |
---|---|---|
1-Shot | 5-Shot | |
LaplacianShot [53] | 0.3581 | 2.5784 |
ICI [54] | 1.1528 | 1.5622 |
DeepEMD [55] | 11.8745 | 12.6195 |
TIM [57] | 2.7006 | 5.6421 |
Baseline [58] | 4.1243 | 4.3781 |
GTNet [59] | 5.3617 | 11.5875 |
ODNet (ours) | 1.0358 | 2.3467 |
ResNet18 | Orthogonal Decomposition | Skip Connection | 5-Way | |
---|---|---|---|---|
1-Shot | 5-Shot | |||
✓ | 65.27 ± 0.23 | 81.13 ± 0.34 | ||
✓ | ✓ | 74.08 ± 0.31 | 87.65 ± 0.42 | |
✓ | ✓ | ✓ | 80.45 ± 0.47 | 93.41 ± 0.26 |
Backbone | 5-Way | |
---|---|---|
1-Shot | 5-Shot | |
ResNet12 | 77.76 ± 0.62 | 89.98 ± 0.29 |
ResNet18 | 80.45 ± 0.47 | 93.41 ± 0.26 |
ResNet34 | 80.02 ± 0.24 | 94.35 ± 0.31 |
Classifier | 5-Way | |
---|---|---|
1-Shot | 5-Shot | |
Euclidean | 80.45 ± 0.47 | 93.41 ± 0.26 |
Cosine | 48.61 ± 0.42 | 61.14 ± 0.39 |
N-Way | K-Shot | Accuracy |
---|---|---|
3 | 1 | 83.43 ± 0.25 |
5 | 93.87 ± 0.31 | |
10 | 95.70 ± 0.42 | |
15 | 95.38 ± 0.43 | |
5 | 1 | 80.45 ± 0.47 |
5 | 93.41 ± 0.26 | |
10 | 95.37 ± 0.29 | |
15 | 95.24 ± 0.36 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Zhang, H.; Liu, H.; Guo, R.; Liang, L.; Liu, Q.; Ma, W. ODNet: A High Real-Time Network Using Orthogonal Decomposition for Few-Shot Strip Steel Surface Defect Classification. Sensors 2024, 24, 4630. https://doi.org/10.3390/s24144630
Zhang H, Liu H, Guo R, Liang L, Liu Q, Ma W. ODNet: A High Real-Time Network Using Orthogonal Decomposition for Few-Shot Strip Steel Surface Defect Classification. Sensors. 2024; 24(14):4630. https://doi.org/10.3390/s24144630
Chicago/Turabian StyleZhang, He, Han Liu, Runyuan Guo, Lili Liang, Qing Liu, and Wenlu Ma. 2024. "ODNet: A High Real-Time Network Using Orthogonal Decomposition for Few-Shot Strip Steel Surface Defect Classification" Sensors 24, no. 14: 4630. https://doi.org/10.3390/s24144630