Detection and Classification of Surface Defects on Solar Cell Panels Based on Deep Learning

Fig. 8. Classification of crack defects. (a) Edge fine defect; (b) large area crack defect; (c) edge bevel defect; (d) star-shaped defect; (e) edge star-shaped defect

Download full size

Fig. 9. Training results of YOLOv5s model for solar cell dataset. (a) Bondary box loss in training set; (b) confidence loss in training set; (c) classification loss in training set; (d) precision; (e) recall; (f) bondary box loss in validation set; (g) confidence loss in validation set; (h) classification loss in validation set; (i) mAP@0.5; (j) mAP@0.5∶0.95

Download full size

Fig. 10. Detection results of YOLOv5s model for solar cell dataset

Download full size

Table 1. Experimental environment configuration
View table
Table 1. Experimental environment configuration
Parameter Configuration
Operating system Windows 11
GPU NVIDIA GeForce RTX 4060
Memory 8 GB
Python 3.8.19
Deep learning framwork PyTorch 2.2

Table 2. Ablation experimental design
View table
Table 2. Ablation experimental design
Serial number YOLOv5s SiLU CBAM C3TR CARAFE WIoUv3
1 √
2 √ √
3 √ √ √
4 √ √ √ √
5 √ √ √ √ √
6 √ √ √ √ √ √

Table 3. Results of ablation experiments

View table

Table 3. Results of ablation experiments

Serial number	P /%	R /%	mAP@0.5 /%	FPS /（frame·s^-1）	Params /10⁶
1	84.6	94.0	89.1	114.746	6.97
2	86.8	95.0	90.0	120.645	7.02
3	81.4	98.0	93.8	97.171	7.26
4	86.1	97.0	94.9	89.044	8.11
5	84.5	97.0	95.7	75.412	8.25
6	89.7	98.0	95.9	75.133	8.25

Table 4. Comparasion of experimental results for different models

View table

Table 4. Comparasion of experimental results for different models

Model	P /%	R /%	mAP@0.5 /%	FPS /（frame·s^-1）	Params /10⁶
YOLOv5n	89.6	95.0	87.2	80.330	1.76
YOLOv8s	87.9	98.0	88.9	78.125	11.12
YOLOv8m+GAMAttention	83.9	95.0	93.0	36.630	29.94
YOLOv5s	84.6	94.0	89.1	114.746	6.97
YOLOv9	76.7	74.5	85.0	42.880	15.51
YOLOv5s-cc	85.3	95.0	92.1	64.065	13.69
YOLOv5s-CW	89.7	98.0	95.9	75.133	8.25

Tools

Get Citation

Copy Citation Text

Junbo Tu, Jialin Zeng, Yuexin Tang, Chenxi Wu, Xiaoyu Liu. Detection and Classification of Surface Defects on Solar Cell Panels Based on Deep Learning[J]. Laser & Optoelectronics Progress, 2025, 62(2): 0237013

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites