Deep Metallic Surface Defect Detection: The New Benchmark and Detection Network
Open accessThe DOI matches its Crossref record; 42 of the 42 references it cites were found in a registry or an index; 2 of the 6 sources its claims cite were checked for support.
Abstract
Metallic surface defect detection is an essential and necessary process to control the qualities of industrial products. However, due to the limited data scale and defect categories, existing defect datasets are generally unavailable for the deployment of the detection model. To address this problem, we contribute a new dataset called GC10-DET for large-scale metallic surface defect detection. The GC10-DET dataset has great challenges on defect categories, image number, and data scale. Besides, traditional detection approaches are poor in both efficiency and accuracy for the complex real-world environment. Thus, we also propose a novel end-to-end defect detection network (EDDN) based on the Single Shot MultiBox Detector. The EDDN model can deal with defects with different scales. Furthermore, a hard negative mining method is designed to alleviate the problem of data imbalance, while some data augmentation methods are adopted to enrich the training data for the expensive data collection problem. Finally, the extensive experiments on two datasets demonstrate that the proposed method is robust and can meet accuracy requirements for metallic defect detection.
Overview
CoChat Summary
## TL;DR
The paper introduces GC10-DET, a new real-industry metallic surface defect dataset with 10 defect types and 3,570 grayscale images, and proposes an end-to-end defect detector (EDDN) based on SSD. Experiments on GC10-DET and NEU-DET show the method is robust, improves multi-scale defect detection, and achieves the best mAP on GC10-DET.
## Key findings
- GC10-DET contains 10 defect types and 3,570 grayscale images collected in a real industrial setting; NEU-DET has 6 defect types and 1,800 grayscale images.
- On NEU-DET, the proposed method achieves the best mAP of 0.724 and the best results on most defect categories; SSD300 is slightly better on Sc recall (0.990 vs. 0.981) and on In AP (0.796 vs. 0.763) and Rs AP (0.621 vs. 0.581).
- On GC10-DET, the proposed method achieves the best mAP and the best results on most categories; SSD300 is slightly better on Rp recall (0.667 vs. 0.333) and Faster-RCNN is better on Cr recall (1 vs. 0.857).
- Runtime is reported as 27 ms/image on NEU-DET and 33 ms/image on GC10-DET for the proposed method; the paper says this is fast enough for real-time use, though YOLO-V2 is slightly faster on GC10-DET testing set time (4.49 s vs. 8 s).
## Methods
- The paper proposes EDDN, an end-to-end detector built on Single Shot MultiBox Detector (SSD) with VGG16, multi-scale feature maps, default boxes, and non-maximum suppression.
- It introduces hard negative mining, limiting negatives to at most a 3:1 ratio with positives, and applies data augmentation using full images, random patches, and patches with minimum Jaccard overlaps of 0.1, 0.3, 0.5, 0.7, or 0.9.
- Evaluation uses Recall, AP, and mAP on NEU-DET and GC10-DET, with comparisons against SSD, Faster-RCNN, YOLO-V2, YOLO-V3, and traditional LBP/HOG methods.
## Limitations
- The paper states that existing defect datasets are generally unavailable for deployment because of limited data scale and defect categories, and that metallic defect detection is difficult due to illumination, light reflection, and metal material effects.
- It notes that traditional methods are sensitive to lighting and background clutter and need repeated parameter tuning; this is a limitation of prior work, not the proposed method.
- Evident from the text: some reported gains are modest or mixed across categories, since other methods outperform the proposed method on a few defect classes (for example, SSD300 and Faster-RCNN on some NEU-DET and GC10-DET classes).
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures





Verification
Claims the paper makes
2 Valid · 1 Not valid · 1 Not checked · 29 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“we propose an end-to-end metallic surface defect network based on the Single Shot MultiBox Detector [6].”
- Also stated in: § 5. Conclusions · plain
“we propose an end-to-end defect detection and classification network based on the Single Shot MultiBox Detector.”
§ Deep Metallic Surface Defect Detection: The New Benchmark and Detection Network > 1. Introduction
What the sources say
- Supportedcites [6] Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C.Y., and Berg, A.C. (2016, January 8–16). Ssd: Single Shot Multibox Detector. Proceedings of the European conference on computer vision, Amsterdam, The Netherlands. source
The citation identifies the Single Shot MultiBox Detector as the basis for the citing authors’ proposed network. The cited paper introduces SSD (Introduction, p. 2) and describes its single-network object detection method (p. 1); it need not describe the citing authors’ metallic-surface defect network. (LLM, conf=0.93)
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: § 5. Conclusions · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Besides, the proposed method shows the best mAP accuracy.”
- Also stated in: § 4.4. Accuracy Comparisons with Deep Methods · plain
“the proposed method shows the best mAP accuracy.”
§ Deep Metallic Surface Defect Detection: The New Benchmark and Detection Network > 4. Experiments > 4.4. Accuracy Comparisons with Deep Methods
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: § 4.4. Accuracy Comparisons with Deep Methods · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“we contribute a new dataset called GC10-DET for large-scale metallic surface defect detection.”
- Also stated in: § 1. Introduction · plain
“We contribute a new dataset named”
- Also stated in: § 4.1. Datasets · plain
“GC10-DET is the surface defect dataset collected in a real industry. It contains ten types of surface defects”
§ Deep Metallic Surface Defect Detection: The New Benchmark and Detection Network > Abstract
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.
- Rests on an assumption not shown here.
- Also stated in: § 1. Introduction · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“the mAP can reach 0.724.”
§ Deep Metallic Surface Defect Detection: The New Benchmark and Detection Network > 4. Experiments > 4.4. Accuracy Comparisons with Deep Methods
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.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“the YOLO methods are difficult to distinguish the six types of defects.”
§ Deep Metallic Surface Defect Detection: The New Benchmark and Detection Network > 4. Experiments > 4.4. Accuracy Comparisons with Deep Methods
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.
7 more claims
Create a free account to see all 33 claims and to verify the 1 reference still open.
- Probably not a substantial claim. 8
- It may state two separate claims as one. 14
Citations
Cites
ImageNet 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 · cited 1×
“Since the introduction of AlexNet [26], convolutional neural networks have been successfully deployed to detect surface defects.”
MethodFaster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Ren, S., He, K., Girshick, R., & Sun, J. (2017) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2016.2577031 · Cited by 73,472
Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Ren, S., He, K., Girshick, R., & Sun, J. (2015) · arXiv · doi:10.48550/arxiv.1506.01497 · Cited by 70,812
Microsoft COCO: Common Objects in Context
Lin, T.-Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D., Dollár, P., & Zitnick, C. L. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-10602-1_48 · Cited by 45,460 · cited 1×
“The pre-trained models are often trained on ImageNet [34] and MS COCO [35] datasets.”
MethodImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380 · cited 1×
“The pre-trained models are often trained on ImageNet [34] and MS COCO [35] datasets.”
Method(2016) · Cited by 31,590
SSD: Single Shot MultiBox Detector
Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C.-Y., & Berg, A. C. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46448-0_2 · Cited by 22,128 · cited 15×
“The proposed method can obtain the best results on the defects of Pu, In, and Rp, while the SSD300 is slightly higher than proposed on Wl (0.974 vs. 0.885), Ss, (0.689 vs. 0.650) and Wf (1 vs. 0.919).”
MethodInfluentialYOLO9000: Better, Faster, Stronger
Redmon, J., & Farhadi, A. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.690 · Cited by 19,892 · cited 8×
“Therefore, the SSD method utilizes VGG16 as the pre-trained model, YOLO-V2 uses Darknet19 model, YOLO-V3 uses Darknet53 model, and Faster R-CNN adopts Resnet50 model.”
MethodInfluentialYOLOv3: An Incremental Improvement
Redmon, J., & Farhadi, A. (2018) · arXiv · doi:10.48550/arxiv.1804.02767 · Cited by 17,141
OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks
Sermanet, P., Eigen, D., Zhang, X., Mathieu, M., Fergus, R., & LeCun, Y. (2013) · arXiv · doi:10.48550/arxiv.1312.6229 · Cited by 2,680 · cited 1×
“In [29], the OverFeat network [30] was implemented to detect 5 different types of surface defects.”
Method
Cited by
Items 201–300 of 592
DDR-YOLO: An Efficient Detection Model Based on YOLO11s for Steel Surface Defect
Ren, X., Liu, Z., Cheng, H., & Chen, L. (2025) · IEEE Access · doi:10.1109/access.2025.3599978 · Cited by 12
Wang, Y., Sun, H., Luo, K., Zhu, Q., Li, H., Sun, Y., Wu, Z., & Wang, G. (2025) · Scientific Reports · doi:10.1038/s41598-025-16619-9 · Cited by 15
“The dataset used in this study, GC10-DET+, is primarily derived from the GC10-DET dataset 37 .”
EL-DETR: a lightweight steel surface defect detection model
Su, F., & Peng, M. (2025) · Measurement science and technology · doi:10.1088/1361-6501/ae17e0 · Cited by 9
“Two steel surface defect datasets, NEU-DET [40] and GC10-DET [41], are employed to evaluate the performance of the EL-DETR model.”
Zeng, Q., Wei, D., Zhang, X. J., Gan, Q., Wang, Q., & Zou, M. (2023) · International Conference on New Trends in Computational Intelligence (NTCI) · doi:10.1109/ntci60157.2023.10403686 · Cited by 9
“In this paper, we evaluate the proposed method on two public datasets for strip surface defect detection, Northeastern University Strip Surface Defect Dataset (NEU-DET) [8] and Industrial Metal Surface Dataset (GC10-DET) [9].”
MethodYOLOv8n-GAM: an improved surface defect detection network for hot-rolled strip steel
Xu, H., Xiao, X., Zhao, Z., & Wang, Z. (2024) · Engineering Research Express · doi:10.1088/2631-8695/ad5417 · Cited by 9
Research on the Detection of Steel Plate Defects Based on SimAM and Twin-NMF Transfer
Zou, Y., Zhang, G., & Fan, Y. (2024) · Mathematics · doi:10.3390/math12172782 · Cited by 8
Defective Surface Detection based on Improved Faster R-CNN
Duan, H. M., Huang, J., Liu, W., & Shu, F. (2022) · IEEE International Conference on Industrial Technology (ICIT) · doi:10.1109/icit48603.2022.10002822 · Cited by 11
Continual learning for surface defect segmentation by subnetwork creation and selection
Dekhovich, A., & Bessa, M. A. (2024) · Journal of Intelligent Manufacturing · doi:10.1007/s10845-024-02393-4 · Cited by 11
“For instance, after Song and Yan (2013) proposed their NEU-DET dataset with Hot Rolled Steel Strip Surface defects, containing six types of defects, other groups collected datasets with either different defect categories or a more significant number of defects, e.g., GC10-DET (Lv et al., 2020) and X-SDD (Feng et al., 2…”
Li, X., Xu, Y., Ji, Z., Zhang, Y., Chen, M., Wang, S., & Zhao, Q. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.127949 · Cited by 12
A novel context-adaptive multi-scale defect detection method for object surface defects
Zhang, H., Liu, J., Pan, D., Jiang, Z., & Dong, J. (2025) · Neurocomputing · doi:10.1016/j.neucom.2025.130927 · Cited by 12
Data expansion for industrial defect detection with diffusion prior
Tai, Y., & Zhang, Z. (2025) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2025.111835 · Cited by 8
Lei, J., Wang, J., Li, Y., Zhang, M., Zhao, R., & Sun, H. (2025) · Measurement Science and Technology · doi:10.1088/1361-6501/aded2b · Cited by 9
“And the strip steel dataset GC10-DET [47].”
Balancing complexity and accuracy for defect detection on filters with an improved RT-DETR
Zhang, M., Wei, X., Liu, G., et al. (2025) · Scientific Reports · doi:10.1038/s41598-025-13960-x · Cited by 7
“To further challenge our model and rigorously evaluate its robustness and generalization capabilities, we incorporate the GC10-DET public dataset 20 .”
Shi, W., Dai, J., Li, C., & Niu, N. (2025) · Mathematics · doi:10.3390/math13172769 · Cited by 7
Shi, Y., Zhan, X., Li, Y., Wu, Z., Zhang, W., & Li, H. (2025) · Measurement · doi:10.1016/j.measurement.2025.119602 · Cited by 11
Liu, L., Sun, S.-Q., & Karimi, H. R. (2025) · Information Fusion · doi:10.1016/j.inffus.2025.104041 · Cited by 17
Zhao, H. M., Liu, J., Zhang, H., Wang, Q., Shen, X., & Zhang, X. (2024) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2024.3355984 · Cited by 7
“In order to validate the effectiveness of the method, we performed validation on two publicly available industrial datasets, including the NEU dataset [37] and GC10-DET dataset [38].”
Zhang, H., Liu, H., Guo, R., Liang, L., Liu, Q., & Ma, W. (2024) · Sensors · doi:10.3390/s24144630 · Cited by 12
“Figure 6 displays a partial sample of this dataset, which comprises 10 hot-rolled defects (6 types are sourced from the NEU-CLS dataset [50] and 4 types are sourced from the X-SDD dataset [51]) and 10 cold-rolled defects (sourced from the GC10-DET dataset [52]), each with 50 samples.”
MethodEfficient minor defects detection on steel surface via res-attention and position encoding
Wu, C., & He, T. (2024) · The Visual Computer · doi:10.1007/s00371-024-03583-0 · Cited by 10
“…validity, we performed experiments on two distinct datasets: the NEU-DET dataset for steel surface defects, hosted by Northeastern University [35], and the GC10-DET dataset, provided by Lv Xiaoming et al. [36] The experiments were executed using a PyTorch-based network on an NVIDIA A800 GPU.”
FPDNet: A fast and high-precision detection network for hot-rolled strip surface defects
Jiang, F., Lü, K. L., Wang, W., Pan, X., Zhang, J., Chen, P., & Wang, B. (2025) · Measurement · doi:10.1016/j.measurement.2025.117309 · Cited by 14
CSAMamba-YOLOv8: a method for detecting defects on steel surfaces
Zhao, Z., & Liu, X. (2025) · Signal Image and Video Processing · doi:10.1007/s11760-025-04044-4 · Cited by 7
“The object detection algorithms include Faster R-CNN, SSD, and YOLOv8n, while the defect detection algorithms consist of EDDN [27], DDN [26], Improved Faster R-CNN [28], and Improved YOLOv7-tiny.”
Xu, H., Zhang, Z., Ye, H., Song, J., & Chen, Y. (2025) · Electronics · doi:10.3390/electronics14102029 · Cited by 12
Qiu, J. W., Shi, H., Hu, Y. H., & Yu, Z. (2023) · Sensors · doi:10.3390/s23239360 · Cited by 14
A Semi-Supervised Deep Learning Approach for the Classification of Steel Surface Defects
Mayuravaani, M., & Manivannan, S. (2021) · International Conference on Information and Automation for Sustainability (ICIAfS) · doi:10.1109/iciafs52090.2021.9606143 · Cited by 9
Li, H., Yi, Z., Wang, Z., Wang, Y., Ge, L., Cao, W., Mei, L., Yang, W., & Sun, Q. (2024) · Processes · doi:10.3390/pr12102134 · Cited by 6
Li, Z., Lai, X., Lou, S.-T., & Li, L. (2025) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2025.3575135 · Cited by 7
“To further substantiate the robustness of the proposed method, we assess its performance on four publicly available industrial defect datasets (NEU-DET [14], GC10-DET [33], PCB [34], and Tianchi-Fabric [35]) commonly used for benchmarking.”
Zhang, H., Liu, H., Guo, R., Zheng, C., Liu, Q., Liang, L., Ma, W., Liu, D., & Zhang, Y. M. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3598394 · Cited by 17
“5(a), FSC-20 comprises ten types of hot-rolled defects (from X-SDD [34] and NEU-CLS [35]) and ten types of cold-rolled defects (from GC10-DET [36]), with 50 samples for each defect category.”
Li, L., Hu, S., Hou, Y., Ke, Y., & Chen, Z. (2025) · Measurement Science and Technology · doi:10.1088/1361-6501/ae08dc · Cited by 7
“The detailed process of MSPM is illustrated in equation (10): To comprehensively evaluate the defect detection performance and versatility within the ADMA-YOLO, this study selected two public datasets, NEU-DET [36] and GC10-DET [37], for verification.”
Optimizing surface defect detection with YOLOv9: the role of advanced backbone models
Zeng, Z., Wang, H., Yao, C., Dong, Z., & Cai, S. (2025) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2025.1675154 · Cited by 7
“The GC10-DET dataset (Lv et al., 2020), released by Tianjin University in 2020, extends the evaluation to a broader range of metallic surface defect types.”
Li, S., Chen, Z., Zhang, X., Yang, T., & Gong, D. (2025) · Measurement Science and Technology · doi:10.1088/1361-6501/ae140a · Cited by 7
“For further evaluation, we also utilized the GC10-DET dataset [37], a real-world industrial dataset collected for SDD.”
Xu, Z., & Wang, N. (2025) · Electronics · doi:10.3390/electronics14244968 · Cited by 6
“Extensive experiments were conducted on the NEU-DET and GC10-DET datasets [10] using RAC-RTDETR.”
Cylinder Liner Defect Detection and Classification based on Deep Learning
Gao, C., Hao, F., Song, J., Chen, R., Wang, F., & Liu, B. (2022) · International Journal of Advanced Computer Science and Applications · doi:10.14569/ijacsa.2022.0130818 · Cited by 8
“To address the problem that the existing defect datasets are generally unavailable for on-site deployment due to the limitation of data scale and defect types, Lv Xiaoming et al established a dataset named GC10-DET using a linear array image acquisition system to collect images [24].”
MethodResearch on Bearing Surface Scratch Detection Based on Improved YOLOV5
Jia, H., Zhou, H., Chen, Z., Gao, R., Lu, Y., & Yu, L. (2024) · Sensors · doi:10.3390/s24103002 · Cited by 8
Zhang, L., Fu, Z., Guo, H., Feng, Y., Sun, Y., & Wang, Z. (2024) · Electronics · doi:10.3390/electronics13183721 · Cited by 6
AMC‐YOLO: Improved YOLOv8‐based defect detection for cigarette packs
Dong, P. M., Wang, Y., Yu, Q., et al. (2024) · IET Image Processing · doi:10.1049/ipr2.13272 · Cited by 6
A high-precision and real-time lightweight detection model for small defects in cold-rolled steel
Chen, S., Jiang, S., Wang, X., Ye, K., Sun, J., & Hua, C. (2024) · Journal of Real-Time Image Processing · doi:10.1007/s11554-024-01606-6 · Cited by 8
“A notable dataset for cold-rolled strip steel defects is GC10-DET, contributed by Lv Xiaoming [25] including 10 common surface defects in cold-rolled steel.”
FMR‐YOLO: An improved YOLOv8 algorithm for steel surface defect detection
Ni, Y., Wu, Q., & Zhang, X. (2025) · IET Image Processing · doi:10.1049/ipr2.70009 · Cited by 9
Peculiarities and Applications of Stochastic Processes with Fractal Properties
Amosov, O. S., & Amosova, S. G. (2021) · Sensors · doi:10.3390/s21175960 · Cited by 6
“Methods of image recognition using photo- and camcorder-sourced images are used to detect defects [30,31].”
MethodGAN-based Data Augmentation for Metal Surface Defect Detection Using Convolutional Neural Networks
Tseng, L.-S., Wu, C., Chen, Y. H., & Tsai, C.-H. (2023) · International Symposium on Computer, Consumer and Control · doi:10.1109/is3c57901.2023.00029 · Cited by 6
An Image-Based Hairline Crack Identification Method for Metal Parts
Hao, C., He, Y., Li, Y., Niu, X., & Wang, S. (2023) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2023.3324689 · Cited by 12
“Although there are public steel sheet datasets for defect detection, such as NEU-CLS [28] and GC10 [29], they cannot obtain satisfactory results for metal parts’ crack detection.”
BackgroundClassification of Steel Strip Surface Defects Based on Optimized ResNet18
Hao, Z., Ren, F., Kang, X., Ni, H., Lv, S., & Wang, H. (2021) · IEEE International Conference on Agents (ICA) · doi:10.1109/ica54137.2021.00018 · Cited by 6
Fang, W., Yang, Y. Q., Zhang, W., Wang, T., Feng, J., & Liu, G. (2025) · Measurement Science and Technology · doi:10.1088/1361-6501/ae06bf · Cited by 8
“This study utilizes two steel surface defect datasets: NEU-DET from Northeastern University [36] and the GC10-DET dataset [37], both specifically designed for surface defect detection in steel materials.”
A lightweight detection network integrating multi-scale semantic refinement for steel strip defects
Bian, B., Chen, L., Han, Z., Wu, S., Li, W., & Chen, H. (2025) · Robotics and Computer-Integrated Manufacturing · doi:10.1016/j.rcim.2025.103188 · Cited by 9
Zhang, Y., Sun, Y.-N., Huang, Q., & Han, Y. (2025) · Measurement Science and Technology · doi:10.1088/1361-6501/ae26a5 · Cited by 10
“(2) GC10-DET [56]: sourced from practical industrial environments, this dataset is widely utilized for detecting surface defects.”
Wu, M., Peng, J., Yu, X., Xu, H., & Sun, H. (2026) · PLoS ONE · doi:10.1371/journal.pone.0339617 · Cited by 5
“To evaluate the effectiveness of our proposed framework, we conduct comprehensive experiments using the NEU-DET [37] dataset for initial module verification, followed by generalization validation on the GC10-DET [38] dataset.”
Shih, Y.-H., & Chen, J. (2026) · Machine Vision and Applications · doi:10.1007/s00138-025-01784-8 · Cited by 5
Image Enhancement of Steel Plate Defects Based on Generative Adversarial Networks
Jie, Z., Zhang, H., Li, K., et al. (2024) · Electronics · doi:10.3390/electronics13112013 · Cited by 5
“Test comparisons show the model’s FID performance on generated defect images in Tables 1 and 2 below.”
TAANet: A Task-Aware Attention Network for Weak Surface Defect Detection
Cui, L., Xie, S., Chen, E., Jiang, X., Wang, Z., Guo, X., & Xu, M. (2024) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2024.3457936 · Cited by 8
“We evaluate the proposed methods on two publicly available steel surface defect detection datasets: NEU-DET [45] and GC10-DET [46].”
MethodGe, Y., Li, Z., Yue, X., et al. (2025) · IEEE Transactions on Emerging Topics in Computational Intelligence · doi:10.1109/tetci.2024.3523776 · Cited by 11
“To evaluate the effectiveness of our proposal, extensive experiments are performed on created datasets Dish-F2 and Dish-S2, a dish dataset Dish-10, and four open datasets Dish-20 [32], [6], Crack [33] and GC10-DET [34].”
Defect attention-based lightweight real-time adaptable surface defect analysis
Delibaşoğlu, İ., DAŞ, D. B., & Daş, O. (2025) · Ironmaking & Steelmaking Processes Products and Applications · doi:10.1177/03019233241296959 · Cited by 7
“MultiBox Detector [43].”
Jiang, S., Liu, M., Liu, Y., Ma, Y., & Wang, Y. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3547122 · Cited by 6
“The same situation occurs in the GC10-DET [53] dataset, especially the insignificant defects at the edges of the image (see row five in Fig.”
InfluentialLabel recognition on metal surfaces in semiconductor industry by YOLO object detection model
Yeh, W., Liao, S.-Y., & Huang, C. (2025) · The International Journal of Advanced Manufacturing Technology · doi:10.1007/s00170-025-15489-9 · Cited by 9
“Lv et al. proposed a new benchmark and detection network for metal surface inspection, demonstrating that deep learning detection networks can effectively address various metal surface defects [7].”
Gan, Y., Ren, X., Liu, H., Chen, Y., & Lin, P. (2025) · Nondestructive Testing And Evaluation · doi:10.1080/10589759.2025.2515572 · Cited by 10
“…of the proposed PV-DETR model and to reduce potential dataset-specific bias, an additional evaluation is performed on the publicly available GC10-DET dataset [40], which consists of 2,293 images covering 10 types of steel surface defects, providing a more diverse and robust benchmark.”
A novel steel surface defect detector with wavelet upsampling and frequency-domain attention
Gan, Y., Ren, X., Liu, H., Lin, P., & Chen, Y. (2025) · Nondestructive Testing And Evaluation · doi:10.1080/10589759.2025.2566787 · Cited by 8
“To further assess the generalisation of the proposed algorithm and reduce dataset-specific bias, the GC10-DET dataset [39] is additionally employed.”
BiSPD-YOLO: Surface Defect Detection Method for Small Features and Low-resolution Images
Yan, S., Chen, G., Gao, A., Liu, C., & Xiong, Z. (2023) · IEEE/ASME International Conference on Advanced Intelligent Mechatronics · doi:10.1109/aim46323.2023.10196194 · Cited by 6
“In the same year, Lv et al. [12] proposed an EDDN defect detection model based on SSD to be applied to metal surface defect detection at different scales, and proved that the model has good robustness and detection accuracy on GC10-DET dataset.”
MethodSurface defect detection using distributed features
Le, T. V., Joshua, J. D., Kim, T.-H., Lee, J., Kim, S. H., & Lee, J. Y. (2025) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2025.110655 · Cited by 7
Application of Task-Aligned Model Based on Defect Detection
Hung, M.-H., Ku, C.-H., & Chen, K.-Y. (2023) · Automation · doi:10.3390/automation4040019 · Cited by 5
“[11] proposed an end-to-end defect detection and classification network based on the single shot multiBox detector Ref.”
MethodWang, R., Du, W., & Jiang, Q. (2025) · Complex & Intelligent Systems · doi:10.1007/s40747-025-01874-0 · Cited by 5
Casting defect-detection method based on multi-image feature of fusion artificial intelligence model
Yang, D., Jiang, H., He, Y., Wang, Y., Liu, Z., Zhang, X., & Cheng, H. (2025) · Applied Intelligence · doi:10.1007/s10489-025-06429-x · Cited by 5
“Furthermore, we conducted experiments on another real-world industrial dataset, GC10-DET [52], to evaluate the defect-detection performance of the proposed method.”
DG-Net: a steel surface defect detection method using dynamic receptive field and gradient attention
Zhang, Y., Zhou, H., Chen, Z., & Shi, Y. (2025) · Engineering Research Express · doi:10.1088/2631-8695/adea31 · Cited by 3
CTC-YOLO: an improved YOLOv11 algorithm for steel surface defect detection
Yang, L., Li, Z., Hu, X., Shao, M., Zhao, Y., & Zhou, C. (2025) · Engineering Research Express · doi:10.1088/2631-8695/adf9bf · Cited by 4
Guan, L., Zhang, H., Zhou, Y., Du, X., & Li, M. (2025) · PLoS ONE · doi:10.1371/journal.pone.0331025 · Cited by 7
“The proposed YOLOv11-WBD model was evaluated on two datasets: NEU-DET [22] and GC10-DET [23].”
Yan, H., Zhang, Z., Peng, T., et al. (2025) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.114606 · Cited by 4
Li, J., He, X., Chen, X., Kong, D., Huang, T., & Song, P. Y. (2025) · Measurement · doi:10.1016/j.measurement.2025.119390 · Cited by 5
An Efficient Lightweight Method for Steel Surface Defect Detection
Zheng, A., Jiang, X., & Liu, W. (2025) · Sensors · doi:10.3390/s25247527 · Cited by 4
“To verify the effectiveness of the proposed LCED-YOLO network, we conduct experimental verification on the NEU-DET [45] and GC10-DET [46] datasets.”
A Cross-Scale Spatial–Semantic Feature Aggregation Network for Strip Steel Surface Defect Detection
Xu, C., Sun, Y., Huang, L., & Guo, H. (2025) · Materials · doi:10.3390/ma18245567 · Cited by 5
“To validate the effectiveness of the proposed method, we conduct extensive experiments on two challenging steel surface defect datasets: NEU-DET [45–47] and GC10-DET [48].”
Dong, F., Zhang, S., chen, X., Mu, D., Yu, X., & Shen, L. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3632285 · Cited by 4
“To verify the e ffi cacy and ideal detection capabilities of the investigated Vanillanet-BGA network, this study conducts strip steel defects detection experiments on NEU-DET [49] and GC10-DET [50] datasets.”
CFL-YOLO: A modified YOLOv8 for detection of steel surface defects
Ren, J., Li, Z., Zhou, Z., Yang, M., Yu, L., He, H., & Liu, T. (2026) · Measurement · doi:10.1016/j.measurement.2026.120509 · Cited by 9
Wang, Y., Lu, Y., Zhu, F., Du, G., & Li, Z. (2026) · Scientific Reports · doi:10.1038/s41598-025-31683-x · Cited by 6
“Experiments were conducted on the NEU dataset, comparing classical models (VGG16, ResNet18), efficient models (EfficientNet-B3, DenseNet121), deep residual models (ResNet-101), and our proposed method (Our work).”
Vengaloor, R. P., & Muralidhar, R. (2023) · The International Arab Journal of Information Technology · doi:10.34028/iajit/21/1/9 · Cited by 10
Li, X., Xu, C., Li, J., Li, Y., & Zhou, X. (2024) · Ironmaking & Steelmaking Processes Products and Applications · doi:10.1177/03019233241293880 · Cited by 5
FFDDNet: Flexible Focused Defect Detection Network
Lin, Z., Li, Z., Yu, J., Hu, M., & Wang, X. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3551459 · Cited by 10
“In order to verify the effectiveness of our proposed method, we conduct experiments on two publicly available steel surface defect detection datasets, NEU-DET [36] and GC10-DET [41].”
CFC-CenterNet: A hybrid approach for steel surface defect detection
Ying, H., Song, M., Xue, Z., Xiao, S., & Yu, J. (2025) · Measurement · doi:10.1016/j.measurement.2025.118891 · Cited by 8
Surface defect detection using deep learning
Piwal, H., Dhokale, M., Biswas, R., Raut, S., & Malge, A. (2023) · AIP conference proceedings · doi:10.1063/5.0129207 · Cited by 4
Classification of Surface Defects in Steel Sheets Using Developed NasNet-Mobile CNN and Few Samples
Kateb, Y., Khebli, A., & Meglouli, H. (2024) · Revue d intelligence artificielle · doi:10.18280/ria.380232 · Cited by 7
NCAT12-DET: A New Benchmark Dataset for Surface Defect Detection and a Comparative Study
Gyimah, N. K., Akinie, R., Yan, X., Nabil, M., Gupta, K. D., Homaifar, A., Hemmati, V., & Opoku, D. (2024) · IEEE Access · doi:10.1109/access.2024.3402668 · Cited by 4
“…to different defect categories may exhibit some similarities in their appearance and texture, as seen in defect categories Rs and Cr. 3) GC10-DET [11] This dataset, with its 3,570 grayscale images sourced from real-world industrial steel plates, represents a significant advancement over the…”
BackgroundMethodSemi-SSDDet: An adaptive semi-supervised steel surface defect detection
Liu, G., Chu, M., & Gong, R. (2025) · Signal Image and Video Processing · doi:10.1007/s11760-025-04416-w · Cited by 4
Research on YOLOX-based tire defect detection method
Qinghua, Q., Xu, X., Liu, Z., et al. (2022) · CAA International Conference on Vehicular Control and Intelligence (CVCI) · doi:10.1109/cvci56766.2022.9964564 · Cited by 4
Metal Surface Defect Detection using Object Detection Models
Shah, H., & Patil, V. (2023) · International Conference on Intelligent and Innovative Technologies in Computing, Electrical and Electronics (IITCEE) · doi:10.1109/iitcee57236.2023.10090466 · Cited by 5
Li, H., Yi, Z., Mei, L., Duan, J., Sun, K., Li, M., Yang, W., & Wang, Y. (2024) · Processes · doi:10.3390/pr12050931 · Cited by 7
Research on Brake Pad Surface Defect Detection Method based on Deep Learning
Cong, Z., Li, X., & Huang, Z. (2023) · International Conference on Advances in Electrical Engineering and Computer Applications (AEECA) · doi:10.1109/aeeca59734.2023.00149 · Cited by 4
“Xiaoming Lv et al. contributed a new dataset GC10-DET for large-scale metal surface defect detection [7], and they propose a novel end-to-end defect detection network (EDDN) based on the Single Shot MultiBox Detector.”
MethodST-YOLO: An Improved Metal Defect Detection Model Based on Yolov5
Sun, S., Deng, M., Luo, J., Zheng, X., & Pan, Y. (2024) · Asia Conference on Algorithms, Computing and Machine Learning (CACML) · doi:10.1145/3654823.3654852 · Cited by 5
“(3) Utilization of the public dataset NEU-DET[11] and GC10-DET[5], with experimental results indicating that the proposed model is more effective than other YOLO models.”
MethodResultAn Improved YOLOv8 Algorithm for Steel Surface Defect Detection
Wang, W., Xiao, H., Zhang, Q., et al. (2024) · Asian Conference on Artificial Intelligence Technology (ACAIT) · doi:10.1109/acait63902.2024.11022247 · Cited by 4
“To validate the effectiveness of our model using the NEU-DET [34] and GC10-DET [35] datasets.”
A Review and Benchmark on State-of-The-Art Steel Defects DetectionPreprint
Chazhoor, A. A. P., Ho, E. S. L., Woo, W. L., & Gao, B. (2022) · SSRN Electronic Journal · doi:10.2139/ssrn.4121951 · Cited by 5
Global Attention Module and Cascade Fusion Network for Steel Surface Defect DetectionPreprint
Liu, G., Chu, M., Gong, R., et al. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4818932 · Cited by 4
Wen, X., Zhao, W., Yu, Z., Zhao, J., & Song, K. (2023) · Coatings · doi:10.3390/coatings13071171 · Cited by 4
“Each defect category includes both the training set D_Train (250 normal images) and the test set D_Test (50 abnormal images and their abnormal area truth maps, 30–80 normal images).”
Unsupervised selective labeling for semi-supervised industrial defect detection
Ge, J., Qin, Q., Song, S., Jiang, J., & Shen, Z. (2024) · Journal of King Saud University - Computer and Information Sciences · doi:10.1016/j.jksuci.2024.102179 · Cited by 4
“Validation experiments were conducted on various datasets, including the Automotive Sealing Rings Defect Dataset, the Metallic Surface Defect Dataset and the PCB Defect Dataset, to demonstrate the effectiveness and broad applicability of the USCL algorithm.”
InfluentialSelf-citationWang, Z., & Song, Y. (2025) · Measurement Science and Technology · doi:10.1088/1361-6501/add611 · Cited by 3
“(5) Our approach demonstrate superior performance, outper-forming the current mainstream detection models on the NEU-DET [12], GC-10 [13], and TCAL datasets [14].”
Anomaly-Focused Augmentation Method for Industrial Visual Inspection
de León, M. A. L., Hameed, I. A., & Koziorek, J. (2025) · IEEE Access · doi:10.1109/access.2025.3570075 · Cited by 7
Li, C., Huang, H., Mo, H., Wen, Z., Zhou, S., & Zhu, Z. (2025) · IEEE Access · doi:10.1109/access.2025.3581313 · Cited by 5
Zhang, K., Liu, C., & Tang, L. (2025) · IEEE Transactions on Image Processing · doi:10.1109/tip.2025.3580320 · Cited by 3
“To verify the e ff ectiveness of the proposed method, we conducted experiments on two benchmark datasets: GC10 [37] and NEU-CLS-64 [38].”
Ding, H., Chen, J., Ye, H., & Chen, Y. (2025) · Applied Sciences · doi:10.3390/app15147887 · Cited by 4
An on-line global–local defect detection framework for wide cold-rolled strip steel
Jiang, P., Xu, Z., Fan, W., & Zhang, J. (2025) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2025.111872 · Cited by 3
A multi-stage dynamic self-distillation network for industrial defect detection
Zhao, H., Liu, J., Wang, Q., Wen, Z., & Shen, X. (2025) · Advanced Engineering Informatics · doi:10.1016/j.aei.2025.103921 · Cited by 3
Su, J., Luo, Q., Wang, Y. W., Yang, C., Gui, W., & Heikkilä, J. (2025) · Information Fusion · doi:10.1016/j.inffus.2025.103940 · Cited by 8
PANDA: Progressive Adaptive Network for Defect-Aware Few-Shot Segmentation
Gao, M., Ma, Y.-F., Jiang, S., et al. (2026) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2025.3640631 · Cited by 3
Huang, W., Zhao, J., Wei, R., et al. (2026) · Information Sciences · doi:10.1016/j.ins.2026.123092 · Cited by 3
DD-Net: A defect detection model for carbon fiber-reinforce thermoplastic prepreg surface
Hong, J., Hu, H., Zhang, H., Song, K., & Zhou, Q. (2026) · Measurement · doi:10.1016/j.measurement.2026.120313 · Cited by 3
WaveMamba-YOLO: Combining frequency awareness and state-space modeling for defect localization
Ge, A., Lv, Y., & Huang, J. (2026) · PLoS ONE · doi:10.1371/journal.pone.0344940 · Cited by 3
“The GC10-DET dataset [44] is a practical industrial benchmark derived from real steel sheet production, encompassing 3,570 grayscale images and ten distinct defect categories.”
Su, S., Shen, J., Lin, P.-Y., Tang, W.-H., Zhang, W.-J., & Chen, Z. (2026) · Italian National Conference on Sensors · doi:10.3390/s26082311 · Cited by 4
“During the steel production process, due to rolling temperature fluctuations, equipment wear, and uneven cooling, various surface defects are inevitable [2].”
Metadata
- Authors
- Lv, X., Duan, F., Jiang, J.-J., Fu, X., & Gan, L.
- Affiliations
- Tianjin University (CN)
- The State Key Lab of Precision Measuring Technology and Instruments, Tianjin University, Tianjin 300072, China
- Corresponding author
- Duan, Fajie
- Venue
- Sensors
- Publisher
- MDPI AG
- Published
- 2020-03-11
- Volume
- 20
- Issue
- 6
- Pages
- 1562
- ISSN
- 1424-8220 (ISSN-L)
- PMC
- PMC7146379
- OpenAlex
- W3012374719
- Cited by
- 643
- References
- 38
- Citations by year
- 2026: 188 · 2025: 188 · 2024: 125 · 2023: 81 · 2022: 40 · 2021: 18 · 2020: 3
- FWCI
- 18.23
- Type
- journal article
- Language
- en
- Open access
- gold
- dx.doi.org
- Licence
- Locations
- Sensors (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- National Natural Science Foundations of China (No.61905175, 51775377, 61971307)
- National key research and development plan (2017YFF0204800)
- the TianJin Natural Science Foundations of China (No. 17JCQNJC01100)
- Equipment pre-research field fund (61405180505, 61400040303)
- Science and Technology on Underwater Information and Control Laboratory (6142218081811)
- 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
- Single Shot MultiBox Detector, hard negative mining, class imbalance, data augmentation, industrial inspection, object detection
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.
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.