The 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.
Deep Metallic Surface Defect Detection: The New Benchmark and Detection Network
Open accessAbstract
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
Surface Defect Detection Methods for Industrial Products: A Review
Chen, Y., Ding, Y., Zhao, F., Zhang, E., Wu, Z., & Shao, L. (2021) · Applied Sciences · doi:10.3390/app11167657 · Cited by 457
“com/ujik132016/industrial-metallic-surface-dataset (accessed on 21 June 2021) GC10-Det [113] Metal https://www.”
RDD-YOLO: A modified YOLO for detection of steel surface defects
Zhao, C., Shu, X., Yan, X., Zuo, X., & Zhu, F. (2023) · Measurement · doi:10.1016/j.measurement.2023.112776 · Cited by 386
Defect Detection Methods for Industrial Products Using Deep Learning Techniques: A Review
Saberironaghi, A., Ren, J., & El-Gindy, M. (2023) · Algorithms · doi:10.3390/a16020095 · Cited by 317
“com/c/severstal-steel-defect-detection/data (accessed on 2 February 2023) GC10–Det [116] Metal https://kaggle.”
A New Steel Defect Detection Algorithm Based on Deep Learning
Zhao, W., Chen, F., Huang, H., Li, D., & Cheng, W. (2021) · Computational Intelligence and Neuroscience · doi:10.1155/2021/5592878 · Cited by 259
“Lv et al. [16] proposed a new end-to-end defect detection network (EDDN) to deal with different types of steel surface defects.”
BackgroundDCC-CenterNet: A rapid detection method for steel surface defects
Tian, R., & Jia, M. (2022) · Measurement · doi:10.1016/j.measurement.2021.110211 · Cited by 222
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Deep Metallic Surface Defect Detection: The New Benchmark and Detection Network. https://app.cochat.ai/papers/10.3390/s20061562
- The paper
- Lv, X., Duan, F., Jiang, J.-J., Fu, X., & Gan, L. (2020). Deep Metallic Surface Defect Detection: The New Benchmark and Detection Network. Sensors. https://doi.org/10.3390/s20061562
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.