A deep learning model for steel surface defect detection
Open accessThe DOI matches its Crossref record; 43 of the 44 references it cites were found in a registry or an index; 3 of the 16 sources its claims cite were checked for support.
Abstract
Industrial defect detection is a hot topic in the field of computer vision. It is a challenging task due to complex features and many categories of industrial defects. In this paper, a deep learning model based on the multiscale feature extraction module is introduced for steel surface defect detection. The main focus on the feature extraction capability of the model and feature fusion capability to improve the accuracy of the model for steel surface defect detection. First, to improve the feature extraction ability of the model, a multiscale feature extraction (MSFE) module is introduced. The MSFE module can effectively extract multiscale features through three branches that have different convolution kernel sizes. Second, an efficient feature fusion (EFF) module is proposed to optimize feature fusion by adding features from the backbone network to the neck network. Third, this paper puts forward a new Bottleneck module by reducing the normalization layer and activation function in the original Bottleneck module. Finally, the backbone network is deepened to further enhance the feature extraction ability of the model. Extensive experiments are conducted on the public NEU-DET dataset. The experimental results validate the effectiveness of the designed modules and the proposed model. Compared with other state-of-the-art methods, the proposed model achieves optimal accuracy(73.08% mAP@0.5) while maintaining a small number of parameters.
Overview
CoChat Summary
## TL;DR
This paper proposes a YOLOv5s-based steel surface defect detector with four changes: an MSFE module, an EFF module, a new Bottleneck, and a deeper backbone. On the NEU-DET dataset, the final model reports the best results among the compared methods, with 73.08% mAP@0.5 and 37.57% mAP@0.5:0.95.
## Key findings
- On the NEU-DET dataset, the proposed model achieved 73.08% mAP@0.5 and 37.57% mAP@0.5:0.95, which the paper states are the best among the compared object detectors.
- Compared with YOLOv5s, the final model improved by 2.21% in mAP@0.5 and 2.55% in mAP@0.5:0.95.
- Ablation results reported these gains over the baseline YOLOv5s: MSFE added +0.81% mAP@0.5 and +1.00% mAP@0.5:0.95; EFF added +0.61% and +1.12%; the new Bottleneck added +0.23% and +0.05%; deepening the backbone added +0.56% and +0.38%.
- The final model has 9.54M parameters and 18 GFLOPs; the baseline YOLOv5s has 7.04M parameters and 15.9 GFLOPs.
## Methods
- The baseline is YOLOv5s, modified with four components: MSFE for multiscale feature extraction, EFF for feature fusion, a simplified Bottleneck with fewer normalization/activation layers, and a deeper backbone with stage ratio changed from 1:2:3:1 to 2:2:4:2.
- Experiments use the public NEU-DET dataset, which contains 1,800 images across 6 defect categories; the paper splits it 8:2 into 1,448 training images and 352 test images.
- Main analysis consists of comparison experiments against YOLOv3, ScaledYOLOv4-csp, YOLOv5s/m/l, and YOLOv7, plus ablation studies evaluating each proposed module; evaluation uses mAP@0.5, mAP@0.5:0.95, and FLOPs.
## Limitations
- The paper does not report cross-dataset validation or external test-set evaluation beyond NEU-DET.
- The text makes evident that the dataset is split only once into train/test; no cross-validation or statistical significance testing is stated in the paper.
- Not stated in the paper.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
2 Valid · 2 Not valid · 7 Not checked · 37 Own results and interpretations
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“Therefore, PANet has a better fusion effect than FPN.”
What the sources say
- Not checked — no full text of the sourcecites [22] Liu S, Qi L, Qin H, Shi J, Jia J (2018) Path aggregation network for instance segmentation. In: IEEE Conference on computer vision and pattern recognition. pp 8759–8768 source
- Could not verifycites [36] Lin TY, Dollár P, Girshick R, He K, Hariharan B, Belongie S (2017) Feature pyramid networks for object detection. In: IEEE conference on computer vision and pattern recognition. pp 2117–2125 sourceExperimental — accuracy not yet measured
The ablation experiments (§5.1.1, pp. 5–6) compare FPN with single-scale baselines and variants of FPN. Searches for PANet and a PANet–FPN feature-fusion comparison found no passage in our copy assessing whether PANet fuses features better than FPN. (LLM, conf=0.77)
plainstates a cause- 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.
In the paper
“There are many categories of steel surface defects and the size of each defect category varies widely. The average size of each category in the NEU-DET [34] dataset is shown in Table1.”
What the sources say
- Not checked — no full text of the sourcecites [34] He Y, Song K, Meng Q, Yan Y (2019) An end-to-end steel surface defect detection approach via fusing multiple hierarchical features. IEEE Trans Instrum Meas 69(4):1493–1504 source
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 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
“It is clear that the proposed method achieves better performance at mAP@0.5 and mAP@0.5:0.95.”
- Also stated in: § Conclusion · boosted
“The experimental results demonstrate that the proposed method obtains optimal scores in mAP@0.5 and mAP@0.5:0.95.”
boostedstates a cause- 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: § Conclusion · boosted
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
“Compared with other state-of-the-art methods, the proposed model achieves optimal accuracy(73.08% mAP@0.5) while maintaining a small number of parameters.”
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 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 number of parameters of the model reduces about 1M.”
plainstates a cause- 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.
14 more claims
Create a free account to see all 48 claims and every source verdict.
- Probably not a substantial claim. 3
- It may state two separate claims as one. 25
- Borderline: one of the tests for a substantial claim is unclear. 1
Citations
Cites
You Only Look Once: Unified, Real-Time Object Detection
Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.91 · Cited by 53,803 · cited 1×
“The well-known one-stageobjectdetectionmodelsincludeSSD[12],YOLOseries [13–16], and RetinaNet [17].”
Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation
Girshick, R., Donahue, J., Darrell, T., & Malik, J. (2014) · 2014 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2014.81 · Cited by 32,684 · cited 1×
“The SOTA two-stage object detection models include R-CNN [9], Fast R-CNN [10], and Faster R-CNN [11].”
(2016) · Cited by 31,590
Feature Pyramid Networks for Object Detection
Lin, T.-Y., Dollar, P., Girshick, R., He, K., Hariharan, B., & Belongie, S. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.106 · Cited by 30,939
Focal Loss for Dense Object Detection
Lin, T.-Y., Goyal, P., Girshick, R., He, K., & Dollar, P. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.324 · Cited by 30,581 · cited 1×
“The well-known one-stageobjectdetectionmodelsincludeSSD[12],YOLOseries [13–16], and RetinaNet [17].”
Girshick, R. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.169 · Cited by 28,105
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 1×
“The well-known one-stageobjectdetectionmodelsincludeSSD[12],YOLOseries [13–16], and RetinaNet [17].”
YOLO9000: 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
MobileNets: 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
YOLOv3: An Incremental Improvement
Redmon, J., & Farhadi, A. (2018) · arXiv · doi:10.48550/arxiv.1804.02767 · Cited by 17,141
Research works printed but not linked: 2
- Ren S, He K, Girshick R, Sun J (2015) Faster R-CNN: towards real-time object detection with region proposal networks. Adv Neural Inf Process Syst 28
- He Y, Song K, Meng Q, Yan Y (2019) An end-to-end steel surface defect detection approach via fusing multiple hierarchical features. IEEE Trans Instrum Meas 69(4):1493–1504 · DOI · Paper page
Cited by
Deep Learning and Computer Vision Techniques for Enhanced Quality Control in Manufacturing Processes
Islam, M. R., Zamil, M. Z. H., Rayed, M. E., Kabir, M. M., Mridha, M. F., Nishimura, S., & Shin, J. (2024) · IEEE Access · doi:10.1109/access.2024.3453664 · Cited by 220
“Li et al. [61] unveiled a DL model specifically for steel surface defect detection, incorporating a Multiscale Feature Extraction (MSFE) module that utilizes various convolutional kernel sizes for superior feature extraction across different scales.”
MethodESMNet: An enhanced YOLOv7-based approach to detect surface defects in precision metal workpieces
Xu, H., Han, F., Zhou, W., Liu, Y., Ding, F., & Zhu, J. (2024) · Measurement · doi:10.1016/j.measurement.2024.114970 · Cited by 98
Steel surface defect detection algorithm in complex background scenarios
Zhao, B., Chen, Y., Jia, X., & Ma, T. (2024) · Measurement · doi:10.1016/j.measurement.2024.115189 · Cited by 84
Xie, W., Ma, W., & Sun, X. (2024) · Neurocomputing · doi:10.1016/j.neucom.2024.128775 · Cited by 75
REDef-DETR: real-time and efficient DETR for industrial surface defect detection
Li, D., Jiang, C., & Liang, T. (2024) · Measurement Science and Technology · doi:10.1088/1361-6501/ad60ea · Cited by 39
Metadata
- Authors
- Li, Z., Wei, X., Hassaballah, M., Li, Y., & Jiang, X.
- Affiliations
- Qilu University of Technology (CN)
- Prince Sattam Bin Abdulaziz University (SA)
- Queen's University (CA)
- Corresponding author
- Jiang, Xuesong
- Venue
- Complex & Intelligent Systems
- Published
- 2023-08-11
- Volume
- 10
- Issue
- 1
- Pages
- 885-897
- ISSN
- 2198-6053 (ISSN-L)
- 2199-4536
- OpenAlex
- W4385767366
- Cited by
- 122
- References
- 33
- Citations by year
- 2026: 34 · 2025: 56 · 2024: 31 · 2023: 1
- FWCI
- 16.14
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- cc-by
- creativecommons.org/licenses/by/4.0 (tdm)
- creativecommons.org/licenses/by/4.0 (vor)
- Locations
- Complex & Intelligent Systems (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Topics
- Physical Sciences › Engineering › Industrial and Manufacturing Engineering › Industrial Vision Systems and Defect Detection
- Physical Sciences › Engineering › Civil and Structural Engineering › Infrastructure Maintenance and Monitoring
- Physical Sciences › Engineering › Mechanical Engineering › Non-Destructive Testing Techniques
- Keywords
- steel surface defect detection, industrial defect detection, multi-scale feature extraction, feature fusion, lightweight neural networks, NEU-DET dataset, convolutional neural network architecture
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: A deep learning model for steel surface defect detection. https://app.cochat.ai/papers/10.1007/s40747-023-01180-7
- The paper
- Li, Z., Wei, X., Hassaballah, M., Li, Y., & Jiang, X. (2024). A deep learning model for steel surface defect detection. Complex & Intelligent Systems. https://doi.org/10.1007/s40747-023-01180-7
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.