Chinese Journal of Liquid Crystals and Displays, Volume. 39, Issue 2, 237(2024)
Defect detection algorithm of improved YOLOv5s solar cell
Fig. 8. Dataset defect type.（a）Cell with broken areas；（b）Cell with obvious bright areas；（c）Cell with black or gray border areas；（d）Scratchy cell；（e）No-electricity，showing the black area of the cell.
Fig. 10. Comparison of detection results before and after improvement.（a）Original drawing；（b）YOLOv5s detection effect drawing；（c）Improved YOLOv5s detection effect drawing.
Table 1. Experimental environment configuration
View tableView in ArticleTable 1. Experimental environment configuration
名称 环境配置 操作系统 Windows 10 64位 GPU NVIDIA GeForce RTX 2080Ti 内存 16 GB Python Python 3.8.2版本 深度学习框架 Pytorch1.11.0
Table 2. Ablation experiments
View tableView in ArticleTable 2. Ablation experiments
CoT CBAM CARAFE WIoU Precision Recall mAP@0.5 检测耗时/ms × × × × 0.861 0.849 0.868 11.9 √ × × × 0.829 0.865 0.895 13.1 × √ × × 0.865 0.87 0.885 11.6 × × √ × 0.871 0.868 0.883 12.3 × × × √ 0.817 0.874 0.873 11.5 √ √ × × 0.863 0.878 0.877 15.7 √ √ √ × 0.879 0.881 0.887 17.4 √ √ √ √ 0.916 0.89 0.901 13.1
Table 3. Comparison of detection performance of different attention mechanisms
View tableView in ArticleTable 3. Comparison of detection performance of different attention mechanisms
算法模型 Param/106 Precision Recall mAP@0.5 YOLOv5s 7.02 0.861 0.849 0.868 YOLOv5s_SE 7.05 0.791 0.864 0.848 YOLOv5s_CA 7.04 0.865 0.835 0.864 YOLOv5s_NWD 7.02 0.838 0.876 0.875 YOLOv5s_NAMAttention 7.02 0.855 0.816 0.84 YOLOv5s_CBAM 6.7 0.865 0.87 0.885
Table 4. Comparison experiments
View tableView in ArticleTable 4. Comparison experiments
Model Precision Recall mAP@0.5 FPS Parameters/M faster-rcnn 0.831 0.668 0.773 18 53.8 EfficientDet 0.883 0.561 0.714 15 70.6 SSD 0.857 0.506 0.749 26 57.9 detr 0.801 0.641 0.729 23 99.3 YOLOv3 0.833 0.772 0.843 46 61.5 YOLOv7 0.873 0.784 0.833 40 36.5 YOLOv7_tiny 0.865 0.815 0.868 46 6.0 YOLOV5s 0.861 0.849 0.868 84 7.02 本文算法 0.916 0.89 0.901 76 7.8
Get Citation
Copy Citation Text
Xueling PENG, Shanling LIN, Zhixian LIN, Tailiang GUO. Defect detection algorithm of improved YOLOv5s solar cell[J]. Chinese Journal of Liquid Crystals and Displays, 2024, 39(2): 237
Category: Research Articles
Received: Aug. 5, 2023
Accepted: --
Published Online: Apr. 24, 2024
The Author Email: Shanling LIN (sllin@fzu.edu.cn)
