X-SDD: A New Benchmark for Hot Rolled Steel Strip Surface Defects Detection
Open accessThe DOI matches its Crossref record; 46 of the 46 references it cites were found in a registry or an index; 11 of the 13 sources its claims cite were checked for support.
Abstract
It is important to accurately classify the defects in hot rolled steel strip since the detection of defects in hot rolled steel strip is closely related to the quality of the final product. The lack of actual hot-rolled strip defect data sets currently limits further research on the classification of hot-rolled strip defects to some extent. In real production, the convolutional neural network (CNN)-based algorithm has some difficulties, for example, the algorithm is not particularly accurate in classifying some uncommon defects. Therefore, further research is needed on how to apply deep learning to the actual detection of defects on the surface of hot rolled steel strip. In this paper, we proposed a hot rolled steel strip defect dataset called Xsteel surface defect dataset (X-SDD) which contains seven typical types of hot rolled strip defects with a total of 1360 defect images. Compared with the six defect types of the commonly used NEU surface defect database (NEU-CLS), our proposed X-SDD contains more types. Then, we adopt the newly proposed RepVGG algorithm and combine it with the spatial attention (SA) mechanism to verify the effect on the X-SDD. Finally, we apply multiple algorithms to test on our proposed X-SDD to provide the corresponding benchmarks. The test results show that our algorithm achieves an accuracy of 95.10% on the testset, which exceeds other comparable algorithms by a large margin. Meanwhile, our algorithm achieves the best results in Macro-Precision, Ma…
Overview
CoChat Summary
TL;DR
The paper introduces X-SDD, a hot-rolled steel strip surface defect dataset with seven defect types and 1,360 images, and benchmarks deep learning classifiers on it. It reports that RepVGG combined with spatial attention achieves 95.10% accuracy and the best Macro-Precision, Macro-Recall, and Macro-F1 among compared models.
Key findings
- X-SDD contains 1,360 defect images at 128 × 128 pixels in 3-channel JPG, covering seven defect types: 238 slag inclusions, 397 red iron sheet, 122 iron sheet ash, 134 surface scratches, 63 oxide scale of plate system, 203 finishing roll printing, and 203 oxide scale of temperature system. It has one more category than NEU-CLS (six types) and is imbalanced, with the largest class more than 6 times the smallest.
- RepVGG+SA achieves 95.10% accuracy, 93.92% Macro-Recall, 95.16% Macro-Precision, and 93.25% Macro-F1 on the 408-image test set; single RepVGG_B3g4 achieves 91.67% accuracy. Across compared algorithms, accuracy ranges 87.01–95.10%, Macro-Recall 82.04–93.92%, Macro-Precision 82.04–95.16%, and Macro-F1 81.58–93.25%.
- Per-class accuracy for RepVGG+SA: finishing roll printing 100%, oxide scale of plate system 78.95% (lowest), red iron sheet 94.12%, scratches 97.50%, inclusion 98.36%, iron sheet ash 83.78%, and oxide scale of temperature system 98.36%.
- Model complexity: RepVGG+SA has 83.825M parameters and 17.892 G MACs, compared with EspNet-v2 at 0.627M parameters and 0.090 G MACs.
Methods
- Dataset was collected from the hot rolled steel strip field; 70% was used as trainset (952 images) and 30% as testset (408 images). Images were resized to 224 × 224 pixels; training used 100 epochs, mini-batch 10, learning rate 0.0001, and Adam optimization on a single NVIDIA RTX2080S GPU, Intel Core i7-9700 CPU, 16GB RAM, Windows 10, and PyTorch.
- The proposed model is RepVGG_B3g4 combined with a spatial attention mechanism (RepVGG+SA), using RepVGG structural re-parameterization and a spatial attention module. It is compared against EspNet-v2, GhostNet, ShuffleNet, SqueezeNet, Xception, VGG16, ResNet50/101/152, RepVGG_B1g2, and RepVGG_B3g4.
- Evaluation metrics are Accuracy, Macro-Recall, Macro-Precision, and Macro-F1; macro metrics are obtained by treating the multiclass problem as multiple binary classification problems and averaging per-class Recall, Precision, and F1.
Limitations
- Stated: RepVGG+SA performs poorly on oxide scale of plate system, with only 78.95% accuracy; the authors attribute this to the relatively small sample size for that category and its morphological similarity to slag inclusion, suggesting more samples, cascade structure, or artificial prior knowledge as possible improvements.
- Stated: RepVGG+SA is computationally costly, with 83.825M parameters and 17.892 G MACs, much larger than EspNet-v2 (0.627M, 0.090 G); future work aims to reduce computational complexity.
- Evident from the text: X-SDD is imbalanced (largest class more than 6 times the smallest) and relatively small (1,360 images from a single hot-rolled strip source); no cross-dataset training or evaluation on NEU-CLS is reported, only dataset comparison.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures









Verification
Claims the paper makes
7 Valid · 5 Not valid · 2 Not checked · 54 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“Existing hot rolled strip lines are usually equipped with surface defect detection systems that can detect defects on the strip surface [6]; however, unfortunately, the system is less accurate in classifying defects.”
§ X-SDD: A New Benchmark for Hot Rolled Steel Strip Surface Defects Detection > 1. Introduction
What the sources say
- Supportedcites [6] 7. Luo, Q.; Fang, X.; Sun, Y.; Liu, L.; Ai, J.; Yang, C.; Simpson, O. Surface Defect Classification for Hot-Rolled Steel Strips by Selectively Dominant Local Binary Patterns. IEEE Access 2019, 7, 23488-23499. [CrossRef] source
The Introduction reports that defect detection has made substantial progress but says surface defect classification has “much improvement space” and describes high-accuracy classification as a continuing obstacle. This supports the citing sentence’s broad characterization of existing systems’ classification performance. (LLM, conf=0.88)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“Compared with the six defect types of the commonly used NEU surface defect database (NEU-CLS), our proposed X-SDD contains more types.”
- Also stated in: § 3.2. The Comparison between Xsteel Surface Defect Dataset and NEU Surface Defect Database · plain
“The NEU-CLS contains six types of defects, while the X-SDD we proposed contains seven types of defects.”
§ X-SDD: A New Benchmark for Hot Rolled Steel Strip Surface Defects Detection > Abstract
What the sources say
- Could not verifycites [31] 32. Song, K.; Yan, Y. Micro Surface defect detection method for silicon steel strip based on saliency convex active contour model. Math. Probl. Eng. 2013, 2013, 429094. [CrossRef] sourceExperimental — accuracy not yet measured
Sections 4.2 and 4.3 discuss spot defects and steel-pit defects, but do not give a defect-type count for NEU-CLS or X-SDD. Searches for “NEU-CLS six defect types” and “X-SDD seven defect types” found no passage in our copy addressing the comparison. Page numbers were not available. (LLM, conf=0.76)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 3.2. The Comparison between Xsteel Surface Defect Dataset and NEU Surface Defect Database · plain
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“In this paper, the model achieves results second only to our proposed RegVGG+SA model in both Accuary and Macro-Precision metrics.”
- Also stated in: § 6. Discussion and Conclusions · plain
“The comparison results show that ResNet50, used in the literature [41], achieves results on Accuracy and Macro-Precision that are second only to our proposed RepVGG+SA algorithm.”
§ X-SDD: A New Benchmark for Hot Rolled Steel Strip Surface Defects Detection > 5.2. Experimental Results
What the sources say
- Partly supportedcites [41] 42. Konovalenko, I.; Maruschak, P.; Brezinová, J.; Vi ňáš, J.; Brezina, J. Steel Surface Defect Classification Using Deep Residual Neural Network. Metals 2020, 10, 846. [CrossRef] sourceExperimental — accuracy not yet measured
The Methodology and Conclusions (pages unavailable) describe a ResNet50 classifier for steel-surface defects and report its classification accuracy; the Training section also identifies precision as a measured metric. The paper does not report results on X-SDD or compare ResNet50 with RepVGG+SA, so it does not support the claimed second-place ranking on Accuracy and Macro-Precision. (LLM, conf=0.85)
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 6. Discussion and Conclusions · plain
Every source the paper cites for this claim was checked and supports it.
In the paper
“The excellent performance of ResNet50 in strip classification indicates that the residual network has some advantages in the classification of strip defects.”
§ X-SDD: A New Benchmark for Hot Rolled Steel Strip Surface Defects Detection > 6. Discussion and Conclusions
What the sources say
- Supportedcites [41] 42. Konovalenko, I.; Maruschak, P.; Brezinová, J.; Vi ňáš, J.; Brezina, J. Steel Surface Defect Classification Using Deep Residual Neural Network. Metals 2020, 10, 846. [CrossRef] source
The Conclusions report that the ResNet50 classifier achieved 0.9691 average accuracy on surface defects in rolled metal and performed better on test data than the other ResNet depths examined. The Methodology also describes how residual connections enable deeper networks and better performance, supporting the stated inference about advantages for defect classification. (LLM, conf=0.90)
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Our test results demonstrate the effectiveness of the ResNet50 model used in [41]”
§ X-SDD: A New Benchmark for Hot Rolled Steel Strip Surface Defects Detection > 5.2. Experimental Results
What the sources say
- Supportedcites [41] 42. Konovalenko, I.; Maruschak, P.; Brezinová, J.; Vi ňáš, J.; Brezina, J. Steel Surface Defect Classification Using Deep Residual Neural Network. Metals 2020, 10, 846. [CrossRef] source
The citation identifies the ResNet50 model used in the source, rather than attributing the citing authors’ test results to it. The source’s Methodology section describes its ResNet50 classifier, and its Conclusions report results for that model. (LLM, conf=0.93)
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
36 more claims
Create a free account to see all 68 claims and every source verdict.
- Probably not a substantial claim. 13
- It may state two separate claims as one. 15
- The quote only partly states this claim. 3
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 10×
“In addition, ResNet50 performs better on X-SDD than the deeper ResNet101 and 152, indicating that the deeper the network level is not better when the amount of data is not particularly large.”
BackgroundInfluentialVaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404
Very Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 1×
“…defect classification networks tend to use off-the-shelf deep learning network structures and their various variants, including AlexNet [23], VGGNet [24], GoogleNet [25], ResNet [26], DenseNet [27], SENet [28], Shuf-fleNet [29] and MobileNet [30], etc. Compared with traditional algorithms such as…”
MethodImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155
Densely Connected Convolutional Networks
Huang, G., Liu, Z., Van Der Maaten, L., & Weinberger, K. Q. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.243 · Cited by 48,566 · cited 2×
“In general, existing strip surface defect classification networks tend to use off-the-shelf deep learning network structures and their various variants, including AlexNet [23], VGGNet [24], GoogleNet [25], ResNet [26], DenseNet [27], SENet [28], Shuf-fleNet [29] and MobileNet [30], etc. Compared with traditional algorit…”
BackgroundGoing deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562 · cited 2×
“In general, existing strip surface defect classification networks tend to use off-the-shelf deep learning network structures and their various variants, including AlexNet [23], VGGNet [24], GoogleNet [25], ResNet [26], DenseNet [27], SENet [28], Shuf-fleNet [29] and MobileNet [30], etc. Compared with traditional algorit…”
BackgroundAn Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al. (2020) · arXiv · doi:10.48550/arxiv.2010.11929 · Cited by 47,472
Squeeze-and-Excitation Networks
Hu, J., Shen, L., & Sun, G. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00745 · Cited by 35,667 · cited 2×
“In general, existing strip surface defect classification networks tend to use off-the-shelf deep learning network structures and their various variants, including AlexNet [23], VGGNet [24], GoogleNet [25], ResNet [26], DenseNet [27], SENet [28], Shuf-fleNet [29] and MobileNet [30], etc. Compared with traditional algorit…”
BackgroundCBAM: Convolutional Block Attention Module
Woo, S., Park, J., Lee, J.-Y., & Kweon, I. S. (2018) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-01234-2_1 · Cited by 25,853 · cited 3×
“According to [40], max pooling and average pooling are used in channel dimension to get two different feature descriptions FS avg ∈ R1×H×W and FS max ∈ R1×H×W .”
BackgroundMethodInfluentialMobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., Andreetto, M., & Adam, H. (2017) · arXiv · doi:10.48550/arxiv.1704.04861 · Cited by 19,015
Research works printed but not linked: 5
- 1. Aldunin, A. Development of method for calculation of structure parameters of hot-rolled steel strip for sheet stamping. J. Chem. Technol. Metall. 2017, 52, 737-740.
- 12. Severstal is Mastering the Production of Video Inspection Systems for Rolled Surfaces. Available online: https://metallurgprom. org/en/press-releases/4952-severstal-osvaivaet-izgotovlenie-sistem-videoinspekcii-poverhnosti-prokata.html (accessed on 18 January 2021).
- 15. Zaghdoudi, R.; Seridi, H.; Boudiaf, A. Multiple classifier combination for steel surface inspection. In Proceedings of the 2nd Conference on Informatics and Applied Mathematics, IAM 2019, Guelma, Algeria, 24-25 April 2019.
- 17. Amid, E.; Aghdam, S.R.; Amindavar, H. Enhanced performance for support vector machines as multi-class classifiers in steel surface defects detection. World Acad. Sci. Eng. Technol. 2012, 6, 1096-1100.
- 38. Zhang, Z. Analysis the Causes of Scale on Hot Rolled Strips & Its Prevention Measures; Xinjiang Bayi Iron & Steel Stock Co., Ltd.: Ürümqi, China, 2009; pp. 41-44.
Cited by
Items 101–154 of 154
Sekaran, S. S., & Subaji, M. (2024) · Journal of Advances in Information Technology · doi:10.12720/jait.15.12.1355-1365 · Cited by 2
“The hot rolled steel strip field provided the dataset of surface flaws of hot rolled steel strip [16].”
DD-SSD: Deep Detector for Strip Steel Defects
Abualighah, S. M., Al-Naimi, A. F., & Duwairi, R. (2022) · International Conference on Information and Communication Systems (ICICS) · doi:10.1109/icics55353.2022.9811123 · Cited by 2
“X-DD Database The used database is X-DD [8], which is the most recent one for strip surface defect classification.”
MethodResultInfluentialAn adaptive weighting multimodal fusion classification system for steel plate surface defect
Tu, F., Wei, M. H., Liu, J., & Liao, L. L. (2023) · Journal of Intelligent & Fuzzy Systems · doi:10.3233/jifs-230170 · Cited by 3
“(2021)[5]proposedaRepVGG algorithmandcombineditwithspatialattention(SA) mechanismtoverifytheimpactonX-SDDdataset.”
Wu, Y., Dong, J., & Chen, J. (2024) · International Journal on Semantic Web and Information Systems · doi:10.4018/ijswis.339000 · Cited by 2
Deep Learning Classification System based on User Guidance
Wen, J., Song, Q., & Fang, J. (2021) · International Conference on Computer Information Science and Artificial Intelligence (CISAI) · doi:10.1109/cisai54367.2021.00092 · Cited by 1
Transfer Learning-Based Ensemble Model for Hot-rolled Steel Surface Defects Classification
Ibrahim, A. A. M. S., & Tapamo, J. (2025) · Journal of Artificial Intelligence and Technology · doi:10.37965/jait.2025.0683 · Cited by 2
Cui, J., Li, J., Mei, Z., Ni, Y., Chen, W., & Li, Z. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3593591 · Cited by 1
“Pitted surface Inclusion Rolled-in scale 2) X-SDD: The Xsteel Surface Defect Dataset (X-SDD) [35] contains seven types of hot-rolled steel strip defect images, including inclusion, red iron sheet, iron sheet ash, scratches, oxide scale of plate system, finishing roll printing, and oxide scale of…”
de la Rosa, F. L., Gómez-Sirvent, J. L., Sánchez-Reolid, R., Morales, R., & Fernández‐Caballero, A. (2026) · Sensors · doi:10.3390/s26020612 · Cited by 2
“In terms of future work, one possible research direction would be to apply the same methodology to the Xsteel Surface Defect Dataset [51], which is a similar database to the one used in the present study and offers new challenges.”
Simon, P., & Uma, V. (2022) · Lecture notes in networks and systems · doi:10.1007/978-3-030-96302-6_28 · Cited by 1
Environment Adaptive Deep Learning Classification System Based on One-shot Guidance
Jin, G., Pei, C., Zhao, N., et al. (2022) · Computers, materials & continua/Computers, materials & continua (Print) · doi:10.32604/cmc.2022.027307 · Cited by 1
“RepVGG (Re-parameterization Visual Geometry Group) is a classification model, which is improved on the basis of the existing models [23].”
InfluentialThe Classification of Multi‐Domain Samples Based on the Cooperation of Multiple Models
Song, Q., Xu, J., Ma, L., et al. (2022) · Complexity · doi:10.1155/2022/5578043 · Cited by 1
“We trained the deep learning models (VoVNet-57 [20], ResNeSt50 [21], RepVGG [22], DenseNet [23], VGG16 [24], and ResNet [25]) on some real datasets by the reported default settings of these models.”
MethodXinyu, 索. S., Jian, 刘. L., Jie, 张. Z., et al. (2023) · Laser & Optoelectronics Progress · doi:10.3788/lop223342 · Cited by 1
Zeng, L., Zheng, F., Chang, X., Chen, J. A., & Wang, S. (2024) · Measurement Science and Technology · doi:10.1088/1361-6501/ad956b · Cited by 1
“Gradient-adaptive sample augmentation is a method that combines various sample expansion techniques according to specific rules to obtain 8 gradients with expansion multiples of [2, 3, 4, 5, 9, 17, 37,59].”
InfluentialLin, L., Wen, A., Wang, C., et al. (2025) · Measurement Science and Technology · doi:10.1088/1361-6501/ae01cd · Cited by 1
“The X-SDD [30] hot-rolled strip SDD dataset from Northeastern University, the MT Defect Dataset published by the Institute of Automation, Chinese Academy of Sciences [31] and the Tianchi Aluminum Dataset [32] were used to verify the validity of our proposed method.”
Jin, G., Jiao, Y., Wang, J., Ma, M., & Song, Q. (2022) · Journal of Experimental & Theoretical Artificial Intelligence · doi:10.1080/0952813x.2022.2092558 · Cited by 1
“RepVGG is a classification network, which is improved on the basis of VGG network (Feng et al., 2021).”
BackgroundMethodInfluentialGuided-Crop Image Augmentation for Small Defect Classification
Orti, J., Moreno-Noguer, F., & Puig, V. (2022) · Proceedings - International Conference on Pattern Recognition/Proceedings/International Conference on Pattern Recognition · doi:10.1109/icpr56361.2022.9956623 · Cited by 1
“The solution, as proven in [23], [26], will have to be a little more sophisticated.”
BackgroundSteel Surface Defect Classification Based on Legendre Multiwavelet Transform and SVMPreprint
zheng, X., Liu, W., & Huang, Y. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4469855 · Cited by 2
Deep Learning Classification System Based on User Guidance
温, 俊. (2022) · Computer Science and Application · doi:10.12677/csa.2022.123072
Banf, M., & Steinhagen, G. (2022) · Proceedings of the World Congress on Electrical Engineering and Computer Systems and Science · doi:10.11159/mhci22.104 · Cited by 1
“Visual inspection of steel surface defects on the Xsteel benchmark dataset As a second experiment, we resort to another, very recent, benchmark dataset for steel surface defects, the Xsteel surface defect dataset [38].”
BackgroundMethodInfluentialModel-Assisted Labeling Deployment — Enablers for AI and IOT Paving Modern Iron- and Steelmaking
Husaković, A., Abbas, A., Melcher, A., Tushev, A., & Johnce, S. T. (2023) · Iron & Steel Technology Conference and Exposition (AISTech) · doi:10.33313/387/234 · Cited by 1
Steel Surface Defect Classification Based on Legendre Multiwavelet Transform and SVMPreprint
zheng, X., Liu, W., & Huang, Y. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4524941
Surface Defect Recognition of Steel Strip Based on Pluse Convolutional Neural Network
吴, 昊. (2023) · Modeling and Simulation · doi:10.12677/mos.2023.126474
Multi-scale Texture Network for Industrial Surface Defect Detection
Wang, L., Huang, Y., & Kong, F. (2024) · Communications in computer and information science · doi:10.1007/978-981-97-0827-7_16
Borno, N. A., Ghosh, A., & Mazumder, D. D. (2023) · International Conference on Industrial Engineering and Operations Management (IEOM) · doi:10.46254/ba06.20230101
“(Feng et al., 2021) adopted the RepVGG model to detect surface defects in hot-rolled steel.”
MethodResidual Flow Group Attention Network For Leather Defect Classification
Lei, S., Wu, C., Xu, H., et al. (2024) · doi:10.1145/3661725.3661740
“Feng et al. [4] presented a novel steel surface defect dataset, X-SDD, and employed RepVGG and spatial attention mechanism to achieve efficient and accurate defect classification.”
MethodInfluentialMulti-scale Fusion Attention Network for Industrial Surface Defect Classification
Wu, C., Lei, S., Xu, H., et al. (2024) · International Conference on Computer Supported Cooperative Work in Design (CSCWD) · doi:10.1109/cscwd61410.2024.10580645
“Feng et al. [9] presented a novel steel surface defect dataset, X-SDD, and employed RepVGG and spatial attention mechanism to achieve efficient and accurate defect classification.”
MethodInfluentialInfluence of Heat Treatment on the Key Service Properties of Built-Up Welding Layer on Pinch Rollers
Wang, Y., Wang, Y., Wang, T., et al. (2024) · Journal of Failure Analysis and Prevention · doi:10.1007/s11668-024-02022-5
Effective and Lightweight Surface Defect Detection via Enhanced Deep Learning Models
Kayanzari, K. P., Li, H.-Z., & Zhang, J. (2024) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata62323.2024.10825832
Steel Surface Defect Recognition Based on Evolutionary Network Architecture Search
Lu, T., & Wu, G. (2024) · China Automation Congress (CAC) · doi:10.1109/cac63892.2024.10864969
A study on Surface Flaw Detection using Deep Learning Approach
Shaikh, E. T. B., & Kokate, M. V. (2024) · Asia-Pacific Artificial Intelligence and Big Data Forum (AIBDF) · doi:10.1145/3718491.3718633
Zeng, L., Hu, J., Cai, S., Qin, M., & Wang, S. (2025) · steel research international · doi:10.1002/srin.202500002 · Cited by 2
A Hybrid Inception V3-SVM Based Framework for Steel Surface Defect Classification
Lin, W., Meng, Y., & Yao, S. (2025) · International Conference on Advanced Algorithms and Control Engineering (ICAACE) · doi:10.1109/icaace65325.2025.11019218
“On the X-SDD dataset, XL. Feng et al. [14] improved the RepVGG model, achieving an accuracy of 95.10%, while WH. Zeng et al. [15] evaluated the performance of the MobileNet model and obtained an accuracy of 96.47%.”
InfluentialZheng, C., Liu, H., & Zhang, H. (2025) · IEEE Data Driven Control and Learning Systems Conference (DDCLS) · doi:10.1109/ddcls66240.2025.11065548
“…FSC-20 dataset (Few-shot classification contains 20 types of strip steel surface defects)[10] was used in this project, which was constructed by Northeastern University in 2023 from the current collection of three world-recognised industrial defects datasets: NEU-CLS[13], X-SDD[14] and GC10-DET[15].”
Jiang, Y., Dai, J., & Jiang, Z. (2025) · Symmetry · doi:10.3390/sym17081313
ACR Net: Adaptive Contrast Reconstruction U-Net for Hot-Rolled Steel Defect Detection
Zhang, J., Wei, J., Li, J., Wang, P., Wang, S., Li, Z., & Liu, Y. (2025) · International Conference on Computer Information Science and Application Technology (CISAT) · doi:10.1109/cisat66811.2025.11181776
“The dataset is obtained by performing pixel-level annotation on the data samples in [14] using the open-source tool LabelMe.”
Yu, X., Sun, H., Peng, J., et al. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5668933
Refining and adaption: multi-modal learning in wear debris analysis
Chen, Y., Lei, F., Zhen, X., Li, X. S., & Zhang, L. (2025) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn64981.2025.11227388 · Cited by 1
“FSC-20 dataset : In order to compare state-of-the-art performance, we also adopt an industrial dataset named FSC-20, which contains 20 types of strip steel surface defect [23] from NEU-CLS [24], X-SDD [25], and GC10-DET [26].”
The influence of slab surface temperature on the descalability of fayalite in a hot strip mill
Alberts, E., & Siyasiya, C. W. (2025) · MATEC Web of Conferences · doi:10.1051/matecconf/202541703012
Contour Analysis System of Thick Plates in Hot Rolling via Machine Vision
Guan, Z., & Ding, J. (2025) · International Conference on Industrial Artificial Intelligence (IAI) · doi:10.1109/iai68403.2025.11276922
“Depending on the plate’s thickness and chemical composition, appropriate cooling strategies are selected to tailor the microstructure and enhance mechanical properties, thereby ensuring product quality [15] .”
Multi-scale Visual Feature Learning Framework for Metal Surface Defect Detection
Fu, B., Yang, X., Li, C., Zhang, Y., Lin, L., & Zhou, W. (2025) · IEEE International Conference on Unmanned Systems (ICUS) · doi:10.1109/icus66297.2025.11294451
“Feng et al. [12] constructed the X-SDD dataset and applied RepVGG [13] with spatial attention mechanisms for bounding box localization, but failed to provide detailed information on defect areas or morphological characteristics.”
Huang, Y., Chen, Z., Zhou, D., & Pan, E. (2026) · Computers in Industry · doi:10.1016/j.compind.2026.104461 · Cited by 2
Data augmentation strategy through key category prioritization for imbalanced image classification
Lin, L., Shi, L., Wang, C., et al. (2026) · Measurement Science and Technology · doi:10.1088/1361-6501/ae4e00
“The X steel surface defect dataset (X-SDD) [33] has a total of seven categories, and the categories are unbalanced.”
Hierarchical vision Mamba with adaptive multi-scale fusion for steel surface defect classification
Yu, X., Sun, H., Peng, J., Zeng, X., He, Y., Chen, L., & Chen, B. (2026) · Journal of King Saud University - Computer and Information Sciences · doi:10.1007/s44443-026-00623-8 · Cited by 3
“First, the inherent multi-scale nature of defects, ranging from microscopic scratches to macroscopic patches, poses significant difficul-ties for developing unified detection frameworks (Gao et al. 2021).”
InfluentialSurface Defect Detection in Hot-Rolled Steel Sheets Using Convolutional Neural Networks
Strgar, O., Selak, L., Požrl, T., & Podržaj, P. (2026) · International Conference on Artificial Intelligence, Computer, Data Sciences and Applications (ACDSA) · doi:10.1109/acdsa67686.2026.11468032
“It is widely used in automotive, appliance manufacturing, bridge construction, and electric motors, making it essential in both industrial and daily applications [3].”
Applications of Deep Learning to Metal Surface Defect Detection: Status and Challenges
Wang, Y., Zhou, M., Zhang, C., & Sedraoui, K. (2026) · Processes · doi:10.3390/pr14081305 · Cited by 3
Improved YOLOv8 for Strip Steel Surface Defect Detection
Su, Y., Luo, L., Deng, Y., et al. (2025) · China Automation Congress (CAC) · doi:10.1109/cac67268.2025.11486903
“The X-SDD steel-strip defect dataset [12] addresses critical industrial quality inspection challenges, including severe class imbalance (ranging from 63 to 397 samples per class), limited data volume (1,360 grayscale 128×128px images), and significant intra-class variations[13].”
Self-citationWei, J., Li, Z., Liu, Y., & Feng, X. (2026) · Journal of Intelligent Manufacturing · doi:10.1007/s10845-026-02882-8
Self-citationOptimized Deep Learning Model for Effective Strip Steel Defect Detection Based on Data Fusion
Wang, J. J., Fu, Z. J., Wan, L. X., & Park, J. J. (2026) · Data & Metadata · doi:10.56294/dm2026794
DIDW-YOLOv11: The Steel Surface Defect Detection Method Based on Improved YOLOv11 Network
Jiang, J., Zhang, Y., Xue, Z., & Wang, C. (2026) · Electronics · doi:10.3390/electronics15122593
Liu, K., Huang, C., & Huang, S. (2026) · Chinese Control and Decision Conference (CCDC) · doi:10.1109/ccdc69976.2026.11560303
“The extracted feature vectors are normalized to unit length to ensure stable and consistent similarity computation: where input f denotes the Euclidean norm of the feature vector, All normalized features from IID-R are computed offline and indexed using FAlSS[13], which supports efficient large-scale similarity search…”
Rethinking Efficiency: A Structural Compression Framework for Domain-Specific Visual RecognitionPreprint
Abazi, Y. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7160880
Surface Defect Detection via Dynamic Contrastive Graph Neural Networks
Iqbal, M. A., Genlot, C., Pao, H.-K., & Supriyono, J. (2026) · IEEE International Conference on Big Data and Artificial Intelligence (BDAI) · doi:10.1109/bdai70753.2026.11655113
“We evaluated the proposed model on three publicly available datasets for SDD applications: the Magnetic-tile-defect datasets (MTD [35], the NEU metal surface defect datasets (NEU-DT) [36], and the Xsteel surface defect dataset (X-DD) [37].”
Setyorini, F., Fariza, A., Wijaya, R. A., et al. (2026) · International Electronics Symposium (IES) · doi:10.1109/ies71033.2026.11702658
“This research aims to: 1) Analyze the performance of MobileNetV2 [9], Mobile-ViT [10], and Swin Transformer [8] in classifying seven X-SDD defect classes [11].”
InfluentialIndustrial Surface Defect Detection via Multi-Compartment Spiking Neural Networks
Fan, S., Guo, B., & Luo, Z. (2026) · International Conference on Advanced Algorithms and Neural Networks (AANN) · doi:10.1109/aann71145.2026.11709394
“X-SDD Dataset [11] : The X-SDD database contains 1,360 images across seven defect categories.”
Metadata
- Authors
- Feng, X., Gao, X., & Luo, L.
- Affiliations
- Northeastern University (CN)
- Corresponding author
- Gao, Xianwen
- Venue
- Symmetry
- Publisher
- MDPI AG
- Published
- 2021-04-17
- Volume
- 13
- Issue
- 4
- Pages
- 706
- ISSN
- 2073-8994 (ISSN-L)
- OpenAlex
- W3153037112
- Cited by
- 158
- References
- 40
- Citations by year
- 2026: 22 · 2025: 27 · 2024: 50 · 2023: 29 · 2022: 23 · 2021: 7
- FWCI
- 14.10
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Symmetry (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- National Natural Science Foundation of China (61573087, 61573088)
- Topics
- Physical Sciences › Engineering › Industrial and Manufacturing Engineering › Industrial Vision Systems and Defect Detection
- Physical Sciences › Engineering › Mechanical Engineering › Non-Destructive Testing Techniques
- Physical Sciences › Engineering › Civil and Structural Engineering › Infrastructure Maintenance and Monitoring
- Keywords
- hot-rolled strip, surface defect detection, defect classification, benchmark dataset, RepVGG, spatial attention mechanism, steel quality inspection
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.