Carbon Fiber Defect Detection Based on Terahertz Technology and YOLOv5s

Table 1. Experimental environment configuration
View table
Table 1. Experimental environment configuration
Name Parameter
System Windows 10
Operational memory 16 GB
Graphics processing unit NVIDIA GeForce RTX 3080Ti
Environment configuration Python 3.8+PyTorch 1.9.0+CUDA 11.6

Table 2. Comparison of detection effects of different models
View table
Table 2. Comparison of detection effects of different models
Item Inclusion 1 Inclusion 2 Hole Tear 1 Tear 2
Original image
YOLOv5s
YOLOv5s-SMALL-CBAM-SIoU

Table 3. Results of ablation experiment

View table

Table 3. Results of ablation experiment

Model	Number of parameters /10⁶	P /%	R /%	mAP@0.5 /%	Detection time /ms
YOLOv5s	7.018	81.7	77.8	84.7	25.236
YOLOv5s-SMALL	7.162	91.4	77.3	89.0	27.919
YOLOv5s-SMALL-CBAM	7.222	90.0	84.9	91.2	30.863
YOLOv5s-SMALL-CBAM-SIoU	7.222	92.3	80.8	91.5	29.414

Table 4. Comparison of detection results of different models

View table

Table 4. Comparison of detection results of different models

Model	AP /%			mAP@0.5 /%	Number of parameters /10⁶	Detection time /ms
Model	Inclusion	Hole	Tear	mAP@0.5 /%	Number of parameters /10⁶	Detection time /ms
YOLOv5n	85.4	81.8	67.4	78.2	1.768	15.156
YOLOv5s	92.8	93.2	68.2	84.7	7.018	25.236
YOLOv7	92.4	89.9	89.3	90.5	37.205	38.325
YOLOv8n	89.3	84.2	74.3	82.6	3.152	18.847
YOLOv8s	90.6	87.8	88.7	89.0	11.137	32.631
YOLOv5s-SMALL-CBAM-SIoU	93.5	93.0	88.1	91.5	7.222	29.414

Tools

Get Citation

Copy Citation Text

Leijun Xu, Yafei Zhou, Jianfeng Chen, Xue Bai. Carbon Fiber Defect Detection Based on Terahertz Technology and YOLOv5s[J]. Laser & Optoelectronics Progress, 2025, 62(6): 0615013

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites