An End-to-End Steel Surface Defect Detection Approach via Fusing Multiple Hierarchical Features
Open accessThe DOI matches its Crossref record; 43 of the 53 references it cites were found in a registry or an index.
Abstract
A complete defect detection task aims to achieve the specific class and precise location of each defect in an image, which makes it still challenging for applying this task in practice. The defect detection is a composite task of classification and location, leading to related methods is often hard to take into account the accuracy of both. The implementation of defect detection depends on a special detection data set that contains expensive manual annotations. In this paper, we proposed a novel defect detection system based on deep learning and focused on a practical industrial application: steel plate defect inspection. In order to achieve strong classification ability, this system employs a baseline convolution neural network (CNN) to generate feature maps at each stage, and then the proposed multilevel feature fusion network (MFN) combines multiple hierarchical features into one feature, which can include more location details of defects. Based on these multilevel features, a region proposal network (RPN) is adopted to generate regions of interest (ROIs). For each ROI, a detector, consisting of a classifier and a bounding box regressor, produces the final detection results. Finally, we set up a defect detection data set NEU-DET for training and evaluating our method. On the NEU-DET, our method achieves 74.8/82.3 mAP with baseline networks ResNet34/50 by using 300 proposals. In addition, by using only 50 proposals, our method can detect at 20 ft/s on a single GPU and reach…
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“There are some shortcomings in manual testing, such as low confidence and high labor intensity.”
From A New Steel Defect Detection Algorithm Based on Deep Learning (2021) · cites this paper as [5]
In the citing paper
“most classification tasks can only find the defects with the highest confidence level in the defect category, resulting in imprecise classification results [3].”
From Steel Surface Defect Recognition: A Survey (2022) · cites this paper as [3]
In the citing paper
“there are many types of steel surface defects and some of these defects may overlap”
From Steel Surface Defect Recognition: A Survey (2022) · cites this paper as [3]
In the citing paper
“Using manual inspection of steel surface defects, it is difficult to distinguish the type of defect and accurately detect the location of the defect [2].”
From A deep learning model for steel surface defect detection (2024) · cites this paper as [2]
In the citing 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.”
From A deep learning model for steel surface defect detection (2024) · cites this paper as [34]
1 more claim about this paper
Create a free account to see every claim other papers make about this one.
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 1×
“The third type is ResNet using residual blocks to make networks deeper without overfitting [23].”
MethodVery Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148
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×
“cessful networks are based on the sequential pipeline architecture [25], which establish the basic structure of CNN and prove the importance of depth of networks.”
BackgroundLeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126
Faster 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 · cited 6×
“Next, RPN [14] is employed to generate region proposals”
BackgroundMethodInfluentialFaster 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 · cited 6×
“Next, RPN [14] is employed to generate region proposals”
BackgroundMethodInfluentialImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333 · cited 1×
“As current research in transfer learning [15], the key to drive large networks is pretraining on ImageNet [22].”
BackgroundGradient-based learning applied to document recognition
Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998) · Proceedings of the IEEE · doi:10.1109/5.726791 · Cited by 58,805 · cited 1×
“combined? There are two essential conditions: nonadjacent, because adjacent layers have highly local correlation [32], and”
BackgroundYou 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 representative direct regression methods are YOLO [29] and SSD [30].”
MethodFully convolutional networks for semantic segmentation
Long, J., Shelhamer, E., & Darrell, T. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298965 · Cited by 38,834 · cited 2×
“The most famous region-based detectors are the “R-CNN family” [27] [28] [14].”
BackgroundMethod
Cited by
Items 1,301–1,338 of 1,338
Chazhoor, A. A. P., Hu, S., Gao, B., Ho, E. S. L., & Woo, W. L. (2026) · Multimedia Tools and Applications · doi:10.1007/s11042-026-21825-2
“1 [53].”
Huang, W., Ma, Y., Wang, B., Yang, F., He, W., & Chen, Y. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-31930-2_10
Unsupervised Transformer Framework for Real-World Furniture Surface Defect Detection
Wu, X., Wang, L., & Chen, X. (2026) · Smart innovation, systems and technologies · doi:10.1007/978-981-92-0607-0_25
Multi-Layer Feature Nonlinear Calibration for Few-Shot Steel Surface Defect Recognition
Yan, H., Wang, Y., Yuan, J., et al. (2026) · Youth Academic Annual Conference of Chinese Association of Automation (YAC) · doi:10.1109/yac71005.2026.11616078
“He et al. proposed a Multilevel Feature-fusion Network (MFN) to enhance fine-grained feature representation [9].”
CCP-YOLO: An Improved YOLOv11n Algorithm for Steel Surface Defect Detection
Xiao, L., Li, P., Wang, C., Zheng, H., & Xu, Y. (2026) · Sensors · doi:10.3390/s26154920
“He et al. [11] constructed a multi-level feature fusion network that strengthens the integration of features across different hierarchical levels; however, the method relies on fixed-weight fusion strategies and lacks an adaptive mechanism for modulating the importance of heterogeneous semantic…”
Cross-Domain Few-shot Surface Defect Detection with Memorable Contrastive Learning
Guo, B., Wang, X. B., & Yu, R. (2026) · Youth Academic Annual Conference of Chinese Association of Automation (YAC) · doi:10.1109/yac71005.2026.11616174
“We evaluated the perfor m ance of our m odel on seven public surface defect datasets, including Alu m iniu m [15], CID [16], Magnetic [17], MSD [18], NEUDET [19], PCB [20], SSGD [21], and one custo m surface defect dataset, CPSFSDET.”
Lin, H., Li, J., Liu, B., et al. (2026) · Railway Engineering Science · doi:10.1007/s40534-026-00445-4
A domain-adaptive contrastive and context-enhanced framework for steel surface defect detection
Gao, K., Wang, S., & Li, B. (2026) · Measurement Science and Technology · doi:10.1088/1361-6501/ae9570
“The effectiveness of the proposed method for steel surface defect detection is evaluated on two widely used public datasets, NEU-DET [13] and GC10-DET [34].”
SFEDet: edge-guided spatial-frequency fusion network for steel surface defect detection
Chen, W., Wei, J., Guo, X., & Chen, B. (2026) · Scientific Reports · doi:10.1038/s41598-026-64588-4
Lee, S., Gil, G., & Im, Y. (2026) · Measurement · doi:10.1016/j.measurement.2026.122725
Steel Surface Defect Detection Network with Multi-scale Feature Interaction and Geometric Perception
Zhang, W., & Ma, X. (2026) · International Conference on Artificial Intelligence and Electromechanical Automation (AIEA) · doi:10.1109/aiea70743.2026.11632582
GSLA-YOLOV11: lightweight steel surface defect detector with gated multi-branch aggregation
Chen, L., & wang, Y. (2026) · Measurement Science and Technology · doi:10.1088/1361-6501/ae9976
“This study uses the publicly available NEU-DET steel surface defect dataset released by Northeastern University [39].”
Direction-Context and Quality-Aware YOLOv8 for Steel Surface Defect Detection
Chen, H. (2026) · International Conference on Computer Vision, Image and Deep Learning (CVIDL) · doi:10.1109/cvidl70130.2026.11637634
Dual-Pooling Channel Attention for Data-Scarce Industrial Surface DefectClassificationPreprint
Sun, H. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7324967
Weng, G., Feng, M., Chen, Y., et al. (2026) · Signal Processing Image Communication · doi:10.1016/j.image.2026.117678
Wang, P., Li, C., Hu, C., et al. (2026) · Journal of King Saud University - Computer and Information Sciences · doi:10.1007/s44443-026-01196-2
“CNN-based one-stage detectors, particularly the YOLO family, have become the industrial mainstream because they offer a favourable speed–accuracy trade-off suited to real-time inspection (Bian et al. 2026; He et al. 2020; Wu et al. 2026; Chen et al. 2025).”
MBHS-DETR: a lightweight collaborative network for bearing surface defect detection
Wang, S., Li, J., Peng, G., Li, K., Wang, T., Liang, W., Yang, S., & Yang, S. (2026) · Multimedia Systems · doi:10.1007/s00530-026-02616-z
“He et al. [14] introduced an MFN-enhanced Faster R-CNN model that employs a multiscale feature pyramid to improve the detection performance of small surface defects.”
Min, K. C., & Kim, W. (2026) · IEEE Access · doi:10.1109/access.2026.3729942
“A large body of work targets surface defect detection and segmentation on steel, metal, and textured surfaces using deep networks [1], [2].”
Wang, Y., Xie, H., Yang, J., Sheng, Z., & Xing, F. (2026) · Optics & Laser Technology · doi:10.1016/j.optlastec.2026.116339
Dhrubo, A. F. H., Hossen, M. M., Zahid, R. R., Tisha, A. I., Sumona, M. A. S., Sajjad, M., Kabir, A. H., Qayum, M. A., & Khan, M. A. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7435056
ARC-Det: Adaptive Range-Aware Context Modeling for Printed Circuit Board Defect Detection
Du, X., & Wang, J. (2026) · Journal of Electronic Testing · doi:10.1007/s10836-026-06258-x
DMD-YOLO: a lightweight model for detecting surface defects on ceramic tiles
Gu, H., Chen, J., Fu, Z., et al. (2026) · Engineering Research Express · doi:10.1088/2631-8695/ae9fca
“To verify the robustness of the DMD-YOLO proposed in this paper under cross-domain conditions, the public dataset NEU-DET [44] for metal surface defects was selected as the external test set, and comparative experiments were conducted with mainstream YOLO series models such as YOLOv5, YOLOv8,…”
MSFUNet: A Lightweight Multi-Scale Fusion Network for Industrial Surface Defect Segmentation
Chen, X., Liu, B., Dong, H., et al. (2026) · International Conference on Computer Information Science and Application Technology (CISAT) · doi:10.1109/cisat71030.2026.11682588
“Industrial surface defect detection is a critical task in intelligent manufacturing and quality control [1].”
MRF-DETR: Multi-Domain Representation via Frequency Decoupling for Steel Surface Defect DetectionPreprint
Zhao, B., Guo, S., Jia, X., Liang, Z., & Hu, R. B. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7451284
Anusuya, S. (2026) · International Conference on Inventive Systems and Control (ICISC) · doi:10.1109/icisc69558.2026.11680997
“The transition to deep CNNs eliminated the need for manually designed pipelines, and ImageNet-pretrained backbones such as ResNet-50 and VGG-16 attained 94 – 97 percent classification accuracy on NEU-DET [1, 17].”
FH-Net: Frequency-domain and hierarchical dual enhancement for few-shot surface defect segmentation
Fan, L., Yang, Z., Guo, F., & Fang, P. (2026) · PLoS ONE · doi:10.1371/journal.pone.0358329
CBMDD-Net: An enhanced multiscale defect detection network for car body surfaces
He, Y., Gao, H., chen, G., et al. (2026) · Computers in Industry · doi:10.1016/j.compind.2026.104560
Self-citationMobile Phone Screen Defect Detection Based on an Improved D-FINEPreprint
Li, X., Huang, B., Jiang, J., Zhao, R., Tan, L., Mei, J., & Feng, Z. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7506850
A teacher-student framework for efficient utilization of unlabeled steel plate data
Ma, Z., Peng, T., & He, Y. (2026) · Displays · doi:10.1016/j.displa.2026.103703
Self-citationDSA-YOLO: Coupling Spectral and Spatial Cues for Lightweight Strip Steel Surface Defect Detection
Cheng, C., & You, Q. (2026) · IEEE Access · doi:10.1109/access.2026.3737532
“He et al. [13] proposed an end-to-end multi-feature fusion network (MFN) that aggregates hierarchical CNN features into a unified representation, establishing one of the earliest dedicated baselines on the NEU-DET benchmark.”
Chu, K.-L., & Chou, C.-W. (2026) · Journal of Intelligent Manufacturing · doi:10.1007/s10845-026-02989-y
A prototype-aligned multi-scale attention network for few-shot industrial surface defect detection
Huang, R., & Zhao, L. (2026) · PLoS ONE · doi:10.1371/journal.pone.0359400
“This experiment used two publicly available industrial surface defect detection datasets, NEU-DET [22,23] and GC10-DET [27], to train and evaluate the proposed model.”
Huang, X., Guangxu, 程. C., Jia, T., et al. (2026) · Laser & Optoelectronics Progress · doi:10.3788/lop252199
Niu, K., Liu, C., Fan, Y., et al. (2026) · Journal of Pipeline Science and Engineering · doi:10.1016/j.jpse.2026.100612
Dynamic scale fusion network for enhanced defect detection in aviation connectors
Ji, L., Huang, C., Wang, B., et al. (2026) · Signal Image and Video Processing · doi:10.1007/s11760-026-05737-0
Wang, D., Li, G., & Dai, C. (2026) · Wood Material Science and Engineering · doi:10.1080/17480272.2026.2740005
Holistic foreground–background guided feature learning for few-shot strip steel defect segmentation
Zhou, X., Wang, G., Wu, Y., et al. (2026) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2026.116365
GER-DETR: efficient steel surface defect detection via lightweight RT-DETR optimization
Liu, X.-J., & Liang, Z.-P. (2026) · Journal of Real-Time Image Processing · doi:10.1007/s11554-026-01987-w
“The NEU-DET dataset [34] contains six typical hot-rolled steel surface defects, including crazing (Cr), inclusion (In), patches (Pa), pitted surface (Ps), rolled-in scale (Rs), and scratches (Sc).”
Metadata
- Authors
- He, Y., Song, K., Meng, Q., & Yan, Y.
- Affiliations
- Northeastern University (CN)
- Loughborough University (GB)
- Venue
- IEEE Transactions on Instrumentation and Measurement
- Published
- 2020-04
- Volume
- 69
- Issue
- 4
- Pages
- 1493-1504
- ISSN
- 0018-9456 (ISSN-L)
- 1557-9662
- OpenAlex
- W2944303778
- Semantic Scholar
- 46d4a2b7ccf38b80b09574cf17544ff2297a8fbf
- Cited by
- 1,341
- References
- 53
- Citations by year
- 2026: 188 · 2025: 304 · 2024: 269 · 2023: 238 · 2022: 186 · 2021: 111 · 2020: 34
- FWCI
- 65.18
- Type
- journal article
- Language
- en
- Open access
- green
- figshare.com
- Licence
- Locations
- IEEE Transactions on Instrumentation and Measurement (publishedVersion)
- Figshare (submittedVersion, cc-by)
- Funders
- National Natural Science Foundation of China (51805078, 51374063)
- National Basic Research Program of China (2017YFB0304200)
- Fundamental Research Funds for the Central Universities (N170304014, N150308001)
- China Scholarship Council (201806085007)
- 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 › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Keywords
- multi-level feature fusion, hierarchical features, region proposal network, bounding box regression, NEU-DET dataset, real-time defect detection, industrial visual 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.
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.