CoChat
Sign Up
Cited by 46Industrial Vision Systems and Defect DetectionAdvanced Neural Network ApplicationsInfrastructure Maintenance and Monitoring

The DOI matches its Crossref record; 8 of the 8 references it cites were found in a registry or an index.

Toward surface defect detection in electronics manufacturing by an accurate and lightweight YOLO-style object detector

Open access
Wang, Jyunrong, Dai, Huafeng, Chen, Taogen, Liu, Hao, Zhang, Xuegang, Zhong, Quan, Lu, Rongsheng·2023·Scientific Reports
DOI 10.1038/s41598-023-33804-wPMC PMC10151317
Open this paper in CoChat — free accountOpen PDF

Abstract

In electronics manufacturing, surface defect detection is very important for product quality control, and defective products can cause severe customer complaints. At the same time, in the manufacturing process, the cycle time of each product is usually very short. Furthermore, high-resolution input images from high-resolution industrial cameras are necessary to meet the requirements for high quality control standards. Hence, how to design an accurate object detector with real-time inference speed that can accept high-resolution input is an important task. In this work, an accurate YOLO-style object detector was designed, ATT-YOLO, which uses only one self-attention module, many-scale feature extraction and integration in the backbone and feature pyramid, and an improved auto-anchor design to address this problem. There are few datasets for surface detection in electronics manufacturing. Hence, we curated a dataset consisting of 14,478 laptop surface defects, on which ATT-YOLO achieved 92.8% mAP0.5 for the binary-class object detection task. We also further verified our design on the COCO benchmark dataset. Considering both computation costs and the performance of object detectors, ATT-YOLO outperforms several state-of-the-art and lightweight object detectors on the COCO dataset. It achieves a 44.9% mAP score and 21.8 GFLOPs, which is better than the compared models including YOLOv8-small (44.9%, 28.6G), YOLOv7-tiny-SiLU (38.7%, 13.8G), YOLOv6-small (43.1%, 44.2G), pp-YOLOE-small (42.7%, 17.4G), YOLOX-small (39.6%, 26.8G), and YOLOv5-small (36.7%, 17.2G). We hope that this work can serve as a useful reference for the utilization of attention-based networks in real-world situations.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41598-023-33804-w

Cites

8 of 8 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Comparison among existing YOLO-style object detectors in terms of mAP0.5:0.95 on the COCO dataset and computational cost (GFLOPs). A smaller number of GFLOPs and a larger mAP0.5:0.95 are better. The x-axis represents mAP0.5:0.95, and the y-axis represents the computational cost (GFLOPs). The area of the circle representing each model is also proportional to its computational cost. This figure was plotted by Microsoft Office 2016.
Figure 1 Comparison among existing YOLO-style object detectors in terms of mAP0.5:0.95 on the COCO dataset and computational cost (GFLOPs). A smaller number of GFLOPs and a larger mAP0.5:0.95 are better. The x-axis represents mAP0.5:0.95, and the y-axis represents the computational cost (GFLOPs). The area of the circle representing each model is also proportional to its computational cost. This figure was plotted by Microsoft Office 2016.CC BY · PubMed Central
Figure 2 The prediction results of ATT-YOLO on various defect types. (a) Dirt, (b) plain particle, (c) collision, (d) scratch, (e) collision, and (f) edge particle. This figure was plotted by python with version 3.8.12.
Figure 2 The prediction results of ATT-YOLO on various defect types. (a) Dirt, (b) plain particle, (c) collision, (d) scratch, (e) collision, and (f) edge particle. This figure was plotted by python with version 3.8.12.CC BY · PubMed Central
Figure 3 Examples of each defect type. The defect types include (a) dirt, (b) scratches, (c) collision defects, (d) plain particles, (e) edge particles, and (f) defects of unknown type. This figure presents only the regions of interest of the original input images. This figure was plotted by python with version 3.8.12.
Figure 3 Examples of each defect type. The defect types include (a) dirt, (b) scratches, (c) collision defects, (d) plain particles, (e) edge particles, and (f) defects of unknown type. This figure presents only the regions of interest of the original input images. This figure was plotted by python with version 3.8.12.CC BY · PubMed Central
Figure 4 The appearance of a portable computer shell. The black regions represent regions of interest. The corresponding parts for the monitor and touchpad are considered background in the context of surface defect detection on portable computer shells. This figure was plotted by the Dahua-A5B57MG200 camera.
Figure 4 The appearance of a portable computer shell. The black regions represent regions of interest. The corresponding parts for the monitor and touchpad are considered background in the context of surface defect detection on portable computer shells. This figure was plotted by the Dahua-A5B57MG200 camera.CC BY · PubMed Central
Figure 5 The original sample images have a resolution of 5000 × 5000, captured by an industrial camera. To obtain high-quality images from these high-resolution industrial images, we use four high-resolution cameras to capture the top left, top right, bottom left, and bottom right sections of an original sample. This figure was plotted by the python with version 3.8.12.
Figure 5 The original sample images have a resolution of 5000 × 5000, captured by an industrial camera. To obtain high-quality images from these high-resolution industrial images, we use four high-resolution cameras to capture the top left, top right, bottom left, and bottom right sections of an original sample. This figure was plotted by the python with version 3.8.12.CC BY · PubMed Central
Figure 6 Illustration of dividing the original images into 1280 × 1280 sub-images. We split the original images into 1280 × 1280 sub-images along the green line. The two adjacent sub-images have an overlapping area of 32 pixels to prevent defects from being damaged during the process. This figure was plotted by the ImageLabeler (https://github.com/Manchery/ImageLabeler).
Figure 6 Illustration of dividing the original images into 1280 × 1280 sub-images. We split the original images into 1280 × 1280 sub-images along the green line. The two adjacent sub-images have an overlapping area of 32 pixels to prevent defects from being damaged during the process. This figure was plotted by the ImageLabeler (https://github.com/Manchery/ImageLabeler).CC BY · PubMed Central
Figure 7 The design of ATT-YOLO. ATT-YOLO uses a many-scale backbone and feature pyramid. In the backbone, ATT-YOLO uses only one self-attention module, which is located immediately behind the P7 module. The gray blocks represent elements from the original design of YOLOv5. This figure was plotted by the Microsoft Office2016.
Figure 7 The design of ATT-YOLO. ATT-YOLO uses a many-scale backbone and feature pyramid. In the backbone, ATT-YOLO uses only one self-attention module, which is located immediately behind the P7 module. The gray blocks represent elements from the original design of YOLOv5. This figure was plotted by the Microsoft Office2016.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

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?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 8 references

Research works printed but not linked: 34

  1. Redmon, J., Divvala, S., Girshick, R. & Farhadi, A. in Proceedings of the IEEE conference on computer vision and pattern recognition. 779–788.
  2. Redmon, J. & Farhadi, A. in IEEE Conference on Computer Vision & Pattern Recognition. 6517–6525.
  3. Tian, Z., Shen, C., Chen, H., He, T. J. I. T. o. P. A. & Intelligence, M. FCOS: A simple and strong anchor-free object detector. 1–1 (2020). · DOI
  4. Tian, Z., Shen, C., Chen, H. & He, T. in 2019 IEEE/CVF International Conference on Computer Vision (ICCV).
  5. Chen, Z. et al. Plant disease recognition model based on improved YOLOv5. Agronomy 12, 365 (2022). · DOI
  6. Dewi, C., Chen, R.-C., Yu, H. & Jiang, X. Robust detection method for improving small traffic sign recognition based on spatial pyramid pooling. J. Ambient Intell. Hum. Comput. 1–18 (2021). · DOI
  7. Dewi, C., Chen, R.-C., Zhuang, Y.-C. & Christanto, H. J. Yolov5 series algorithm for road marking sign identification. Big Data Cogn. Comput. 6, 149 (2022). · DOI
  8. Mekhalfi, M. L. et al. Contrasting YOLOv5, transformer, and EfficientDet detectors for crop circle detection in desert. IEEE Geosci. Remote Sens. Lett. 19, 1–5 (2021). · DOI
  9. Zhang, H., Tian, M., Shao, G., Cheng, J. & Liu, J. Target detection of forward-looking sonar image based on improved yolov5. IEEE Access 10, 18023–18034 (2022). · DOI
  10. Xu, Z., Zhang, Y., Cheng, J. & Ge, G. in Journal of Physics: Conference Series. 012038 (IOP Publishing).
  11. Yao, J. et al. A real-time detection algorithm for Kiwifruit defects based on YOLOv5. Electronics 10, 1711 (2021). · DOI
  12. Dewi, C., Chen, R.-C., Liu, Y.-T., Jiang, X. & Hartomo, K. D. Yolo V4 for advanced traffic sign recognition with synthetic training data generated by various GAN. IEEE Access 9, 97228–97242 (2021). · DOI
  13. Zhu, X., Lyu, S., Wang, X. & Zhao, Q. in Proceedings of the IEEE/CVF International Conference on Computer Vision. 2778–2788.
  14. Lin, T. Y., Maire, M., Belongie, S., Hays, J. & Zitnick, C. L. J. S. I. P. Microsoft COCO: Common Objects in Context. (2014). · DOI · Paper page
  15. Misra, D. Mish: A self regularized non-monotonic neural activation function. Preprint https://arxiv.org/abs/1908.086814, 10.48550 (2019).
  16. Du, S., Zhang, B., Zhang, P. & Xiang, P. in 2021 IEEE 2nd International Conference on Pattern Recognition and Machine Learning (PRML). 92–98 (IEEE).
  17. Wang, C. Y., Liao, H., Wu, Y. H., Chen, P. Y. & Yeh, I. H. in 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).
  18. He, K., Zhang, X., Ren, S. & Sun, J. Spatial pyramid pooling in deep convolutional networks for visual recognition. IEEE Trans. Pattern Anal. Mach. Intell. 37, 1904–1916 (2015). · DOI · Paper page
  19. Liu, S., Qi, L., Qin, H., Shi, J. & Jia, J. in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).
  20. Nair, V., Chatterjee, M., Tavakoli, N., Namin, A. S. & Snoeyink, C. Fast Fourier transformation for optimizing convolutional neural networks in object recognition. Preprint https://arxiv.org/abs/2010.04257 (2020). · arXiv
  21. Riaz, H. U. M., Benbarka, N. & Zell, A. in 2020 25th International Conference on Pattern Recognition (ICPR). 7833–7840 (IEEE).
  22. Wei, Y. et al. in Proceedings of the IEEE conference on computer vision and pattern recognition. 7268–7277.
  23. Carion, N. et al. in European conference on computer vision. 213–229 (Springer).
  24. Zhu, X. et al. in International Conference on Learning Representations.
  25. Ho, S.-Y., Shu, L.-S. & Chen, J.-H. Intelligent evolutionary algorithms for large parameter optimization problems. IEEE Trans. Evol. Comput. 8, 522–541 (2004). · DOI
  26. Ho, S. Y., Chen, J. H. & Huang, M. H. Inheritable genetic algorithm for biobjective 0/1 combinatorial optimization problems and its applications. IEEE Trans. Syst. Man Cybern. Part B Cybern. 34, 609–620 (2004). · DOI
  27. Ho, S.-Y. in Proc. Genetic and Evolutionary Computation Conference, Orlando, Florida, USA, 1999.
  28. Wang, J.-R. et al. ESA-UbiSite: accurate prediction of human ubiquitination sites by identifying a set of effective negatives. Bioinformatics 33, 661–668 (2017). · DOI
  29. Tsai, M.-J. et al. GREMA: modelling of emulated gene regulatory networks with confidence levels based on evolutionary intelligence to cope with the underdetermined problem. Bioinformatics 36, 3833–3840 (2020). · DOI
  30. Ho, S.-Y. & Huang, H.-L. Facial modeling from an uncalibrated face image using a coarse-to-fine genetic algorithm. Pattern Recogn. 34, 1015–1031 (2001). · DOI
  31. Yerukala Sathipati, S. & Ho, S.-Y. Identifying a miRNA signature for predicting the stage of breast cancer. Sci. Rep. 8, 1–11 (2018). · DOI
  32. Zhou, D., Fang, J., Song, X., Guan, C. & Yang, R. J. I. IoU Loss for 2D/3D Object Detection. (2019). · DOI
  33. Gevorgyan, Z. SIoU Loss: More Powerful Learning for Bounding Box Regression. Preprint https://arxiv.org/abs/2205.12740 (2022). · arXiv
  34. Du, S., Zhang, B. & Zhang, P. Scale-sensitive IOU loss: An improved regression loss function in remote sensing object detection. IEEE Access 9, 141258–141272 (2021). · DOI

Cited by

Sorted by citations

Metadata

Authors
Wang, J., Dai, H., Chen, T., Liu, H., Zhang, X., Zhong, Q., & Lu, R.
Affiliations
  • Tsinghua University (CN)
  • Hefei University of Technology (CN)
Corresponding author
Lu, Rongsheng
ORCID
Venue
Scientific Reports
Publisher
Springer Science and Business Media LLC
Published
2023-05-01
Volume
13
Issue
1
Pages
7062-7062
ISSN
2045-2322 (ISSN-L)
DOI
10.1038/s41598-023-33804-w
PMC
PMC10151317
OpenAlex
W4367598088
Cited by
46
References
8
Citations by year
2026: 10 · 2025: 8 · 2024: 24 · 2023: 4
FWCI
6.10
Type
journal article
Language
en
Open access
Licence
Locations
  • Scientific Reports (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • The School of Instrument Science and Opto-electronics Engineering at Hefei University of Technology
  • LCFC (Hefei) Electronics Technology Co., Ltd.
  • Hefei LCFC Information Technology Co., Ltd.
  • Anhui Province postdoctoral research personnel scientific research activity fund (2022B636)
Topics
  • Physical Sciences › Engineering › Industrial and Manufacturing Engineering › Industrial Vision Systems and Defect Detection
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
  • Physical Sciences › Engineering › Civil and Structural Engineering › Infrastructure Maintenance and Monitoring
Keywords
surface defect detection, electronics manufacturing, lightweight object detection, self-attention, feature pyramid, defect dataset, real-time inference, manufacturing quality control

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