改进YOLOv5s算法的钢材表面缺陷检测

doi:10.16180/j.cnki.issn1007-7820.2024.12.008

Abstract

Abstract:

In view of the problems such as low accuracy and slow recognition speed of existing steel surface defect detection methods, a defect detection method based on improved YOLOv5s(You Only Look Once version 5s) is proposed in this study. CBAM(Convolutional Block Attention Module) attention mechanism is introduced into the backbone feature extraction network to pay attention to the information of important regions and improve the model's learning ability of target defects. In order to improve the regression speed and localization accuracy of the target frame, the EIoU(Efficient Intersection over Union) boundary frame loss function combining distance loss and width-height loss are used to calculate the loss value. Transfer learning is used to accelerate the convergence speed of the model to improve the accuracy of defects detection. The experimental results on the data set NEU-DET show that compared with the original YOLOv5s network, the improved YOLOv5s network of the accuracy rate of the data set increased by 6.3 percentage point, the recall rate increased by 9.2 percentage point, and the mAP(mean Average Precision) reached 81.7%, which indicates the improved method has good performance for the detection of steel surface defects. The size of the steel surface defect detection model with the improved YOLOv5s algorithm is only 13.8 MB, which improves the detection accuracy on the basis of real-time performance and facilitates the deployment of the model in practical applications.

Key words: deep learning, convolutional neural networks, YOLOv5s, attention mechanism, loss function, transfer learning, target detection, defect detection

CLC Number:

TP391.41

CUI Jingnan, HUANG Chunyan, LI Yanling. Improvement of YOLOv5s Algorithm for Steel Surface Defect Detection[J].Electronic Science and Technology, 2024, 37(12): 48-55.

Figures/Tables 14

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

Table 1.

Figure 6.

Figure 7.

Table 2.

Table 3.

Figure 8.

Table 4.

Table 5.

Figure 9.

References 19

[1]	Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks[J]. Advances in Neural Information Processing Systems, 2012, 25(2):196-210.
[2]	胡江宇, 贾树林, 马双宝. 基于改进级联Faster R-CNN的PCB表面缺陷检测算法[J]. 仪表技术与传感器, 2022(7):106-110,126.
	Hu Jiangyu, Jia Shulin, Ma Shuangbao. PCB surface defect detection algorithm based on improved cascaded Faster R-CNN[J]. Instrument Technique and Sensors, 2022(7): 106-110,126.
[3]	杨莉, 张亚楠, 王婷婷, 等. 基于改进Faster R-CNN的钢材表面缺陷检测方法[J]. 吉林大学学报(信息科学版), 2021, 39(4):409-415.
	Yang Li, Zhang Yanan, Wang Tingting, et al. New method for steel surface defect detection based on improved Faster R-CNN[J]. Journal of Jilin University(Information Science Edition), 2021, 39(4):409-415.
[4]	刘琪, 雷景生. 基于改进深度网络的钢材表面缺陷检测[J]. 计算机工程与设计, 2022, 43(9):2654-2661.
	Liu Qi, Lei Jingsheng. Surface defect detection of steel based on improved deep network[J]. Computer Engineering and Design, 2022, 43(9):2654-2661.
[5]	吴珊, 周凤. 基于改进SSD的小目标检测算法[J]. 计算机工程, 2023, 49(7):179-188,195. doi: 10.19678/j.issn.1000-3428.0065253
	Wu Shan, Zhou Feng. Small target detection based on improved SSD algorithm[J]. Computer Engineering, 2023, 49(7):179-188,195. doi: 10.19678/j.issn.1000-3428.0065253
[6]	方叶祥, 甘平, 陈俐. 金属表面缺陷检测的改进YOLOv3算法研究[J]. 机械科学与技术, 2020, 39(9):1390-1394.
	Fang Yexiang, Gan Ping, Chen Li. Improved YOLOv3 algorithm for detection of metal surface defect[J]. Mechanical Science and Technology for Aerospace Engineering, 2020, 39(9):1390-1394.
[7]	王新宇, 赵静文, 刘翔, 等. 融合坐标注意力机制的YOLOv3肺结节检测算法[J]. 电子科技, 2024, 37(6):1-6.
	Wang Xinyu, Zhao Jingwen, Liu Xiang, et al. YOLOv3 lung nodule detection based on coordinate attention[J]. Electronic Science and Technology, 2024, 37(6):1-6.
[8]	Redmon J, Divvala S, Girshick R, et al. You only look once:Unified,real-time object detection[C]. Las Vegas: Conference on Computer Vision and Pattern Recognition,2016:779-788.
[9]	Redmon J, Farhadi A. YOLO9000:Better,faster,stronger[C]. Honolulu: Conference on Computer Vision and Pattern Recognition,2017:6517-6525.
[10]	Farhadi A, Redmon J. YOLOv3:An incremental improvement[C]. Berlin:Computer Vision and Pattern Recognition,2018:1-6.
[11]	Bochkovskiy A, Wang C Y, Liao H Y M. YOLOv4:Optimal speed and accuracy of object detection[EB/OL].(2020-04-23)[2023-04-14] https://arxiv.org/abs/2004.10934.
[12]	Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]. Salt Lake City: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2018:7132-7141.
[13]	Wang Q L. ECA-Net:Efficient channel attention for deep convolutional neural networks[C]. Seattle: IEEE/CVF Conference on Computer Vision and Pattern Recognition,2020:11531-11539.
[14]	Woo S, Park J, Lee J Y, et al. CBAM:Convolutional block attention module[C]. Munich: Proceedings of the European Conference on Computer Vision,2018:3-19.
[15]	Rezatofighi C, Seyed H, Hamid P, et al. Generalized intersection over union:A metric and a loss for bounding box regression[C]. Long Beach: IEEE/CVF Conference on Computer Vision and Pattern Recognition,2019:658-666.
[16]	Zheng Z, Wang P, Liu W, et al. Distance-IoU loss:Faster and better learning for bounding box regression[EB/OL].(2019-11-19)[2023-04-14] https://arxiv.org/abs/1911.08287.
[17]	Zheng Z, Wang P, Ren D, et al. Enhancing geometric factors in model learning and inference for object detection and instance segmentation[J]. IEEE Transactions on Cybernetics, 2021, 52(8):8574-8586.
[18]	Zhang Y F, Ren W, Zhang Z, et al. Focal and efficient IoU loss for accurate bounding box regression[J]. Neurocomputing, 2022, 50(6):146-157.
[19]	Yu H. An end-to-end steel surface defect detection approach via fusing multiple hierarchical features[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69(4):1493-1504. doi: 10.1109/TIM.2019.2915404

实验环境	配置
操作系统	Windows 11
CPU	AMD Ryzen 75 800 H with Radeon Graphics
GPU	NVIDIA GeForce RTX 3060 Laptop GPU(6 GB)
深度学习框架	PyTorch 1.10.2
开发环境	Python 3.8
显卡加速	CUDA 11.3

参数	数值
学习率(lr)	0.010 0
学习率动量(momentum)	0.937 0
权重衰减系数(weight_decay)	0.000 5
批处理大小(batch size)	12
迭代次数(epoch)	200

实验	CBAM	EIoU	迁移学习	准确率 /%	召回率 /%	mAP /%
YOLOv5s	-	-	-	67.2	70.0	70.7
YOLOv5s-C	√	-	-	63.0	73.1	71.4
YOLOv5s-CE	√	√	-	68.8	73.5	72.0
YOLOv5s-CEQ	√	√	√	73.5	79.2	81.7

算法	mAP/%	模型大小 /MB	AP/%
算法	mAP/%	模型大小 /MB	裂纹	夹杂物	斑块	麻点	轧入氧化皮	划痕
Faster R-CNN	77.2	108	37.8	86.1	92.3	88.5	63.5	95.1
SSD	72.3	93.1	36.3	81.9	91.3	83.9	62.1	78.2
YOLOv4	69.1	244.0	35.1	77.6	90.2	78.4	51.2	82.2
YOLOv8s	71.3	21.4	30.7	65.8	92.2	97.2	51.0	90.9
YOLOv8l	71.4	83.6	29.4	71.3	90.1	95.9	51.1	90.5
YOLOv5s	70.7	14.1	33.9	67.4	92.1	96	50.8	83.9
YOLOv5s-CEQ	81.7	13.8	45.1	82.9	91.7	97.4	78.6	94.1

Improvement of YOLOv5s Algorithm for Steel Surface Defect Detection

RichHTML

PDF (PC)

Like

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 14

References 19

Related Articles 15

Metrics

Comments

Recommended 0

网络层序号	连接层序号	模块参数	模块
0	-1	3 520	Conv
1	-1	18 560	Conv
2	-1	18 816	C3
3	-1	19 426	CBAM
4	-1	73 984	Conv
5	-1	115 712	C3
6	-1	76 642	CBAM
7	-1	295 424	Conv
8	-1	625 152	C3
9	-1	304 738	CBAM
10	-1	1 180 672	Conv
11	-1	1 182 720	C3
12	-1	1 215 586	CBAM
13	-1	656 896	SPPF
14	-1	131 584	Conv
15	-1	0	Upsample
16	[-1, 9]	0	Concat
17	-1	361 984	C3
18	-1	33 024	Conv
19	-1	0	Upsample
20	[-1, 6]	0	Concat
21	-1	90 880	C3
22	-1	147 712	Conv
23	[-1, 18]	0	Concat
24	-1	296 448	C3
25	-1	590 336	Conv
26	[-1, 14]	0	Concat
27	-1	1 182 720	C3
28	[21, 24, 27]	29 667	Detect

[1]	KUAI Xinchen, LI Ye. Hybrid Image Super-Resolution Reconstruction with Multiple and Multi-Scale Attention [J]. Electronic Science and Technology, 2024, 37(9): 34-42.
[2]	HE Zhiqiang, SUN Zhanquan. Swin-Transformer-Based Carotid Ultrasound Image Plaque Segmentation [J]. Electronic Science and Technology, 2024, 37(9): 48-56.
[3]	LI Jun, LIU Hu, YANG Haima, WANG Yuan, XU Wencheng, HUANG Hongxin. Research on AGV Path Fusion of Object Detection and DWA Algorithm [J]. Electronic Science and Technology, 2024, 37(7): 33-42.
[4]	CAO Chunping, XU Zhihua. A Self-Supervised CT Image Classification Method Incorporating Intra-Slice Semantic and Inter-Slice Structural Features [J]. Electronic Science and Technology, 2024, 37(7): 43-52.
[5]	HE Xing, HUANG Yongming, ZHU Yong. Pavement Pothole Detection Method Based on Improved YOLOv5 [J]. Electronic Science and Technology, 2024, 37(7): 53-59.
[6]	WANG Xinyu, ZHAO Jingwen, LIU Xiang, SHI Yunyu, SHE Yunlang. YOLOv3 Lung Nodule Detection Based on Coordinate Attention [J]. Electronic Science and Technology, 2024, 37(6): 1-7.
[7]	YE Yuxin, JU Zhiyong, LAI Ying. Traffic Sign Detection Algorithm Incorporating Receptive Field Enhancement Module and Attention Mechanism [J]. Electronic Science and Technology, 2024, 37(6): 8-16.
[8]	YUAN Ao, QI Jinpeng, JIA Can, XUE Yuxin, GUO Yangyang. A Classification Method of Time Series Pathological Data Segments Based on Deep Learning [J]. Electronic Science and Technology, 2024, 37(6): 84-91.
[9]	ZHU Zihao, SONG Yan. Lightweight Capsule Network Fusing Attention and Capsule Pooling [J]. Electronic Science and Technology, 2024, 37(5): 1-8.
[10]	XIA Rongcheng, LIU Deer. Vehicle Detection and Analysis in Urban Waterlogging Area Based on Deep Learning [J]. Electronic Science and Technology, 2024, 37(5): 18-24.
[11]	LIANG Chenye, ZHANG Xuanxiong. Research on Multiclass Garbage Classification Algorithm Based on Improved MobileNet Network [J]. Electronic Science and Technology, 2024, 37(4): 38-46.
[12]	MA Wenjie, ZHANG Xuanxiong. Research on Blind Roads and Obstacle Recognition Algorithm Based on Deep Learning [J]. Electronic Science and Technology, 2024, 37(3): 75-83.
[13]	WANG Xiaozhu,YU Lianzhi. Small Object Detection Based on Convolution and Self-Attention of Aggregation [J]. Electronic Science and Technology, 2024, 37(2): 14-22.
[14]	MO Zhuorui, HUANG Qianghao, ZHANG Lin, CAO Yuqi, GE Weiting, YU Minghui. Grading and Diagnostic Method for Colorectal Cancer Immunohistochemical Images [J]. Electronic Science and Technology, 2024, 37(12): 24-31.
[15]	LIANG Shuifen, GUO Zhiyong. Advancements of Deep Learning in Imaging Diagnosis of Kidney Stone [J]. Electronic Science and Technology, 2024, 37(12): 67-72.