Lightweight Detection Algorithm of Power Equipment Based on Improved YOLOv5

Table 1. mAP values of different devices compared before and after model improvement
View table
View in Article
Table 1. mAP values of different devices compared before and after model improvement
类别精确率/(0.5%)
Mobilenetv3‑YOLOv5 MEGI‑YOLOv5
绝缘子 85 96.42
避雷器 70 75.37
断路器 75 82.33
隔离开关 80 92.73
发热区域 80 91.32

Table 2. Model comparison experiment
View table
View in Article
Table 2. Model comparison experiment
类别精确率/(0.5%) 参数量/10⁶ 平均帧率/s
YOLOv5s 90.03 7.2 41.20
Mobilenetv3‑YOLOv5 87.04 6.8 80.03
MEGI‑YOLOv5 95.47 5.6 56.32
SSD 90.01 26.2 47.29
YOLOv4 95.03 64.4 138.33
F‑RCNN 95.45 137 80.04

Table 3. The results of model testing obtained by ablation experiment
View table
View in Article
Table 3. The results of model testing obtained by ablation experiment
原始模型 +ECA +迁移学习 +标签平滑尺寸/m
精确率/
(0.5%)
平均帧率/s
√ 25.84 87.04 80.03
√ √ 27.02 89.14 77.46
√ √ √ 27.02 90.95 77.46
√ √ √ √ 27.02 95.55 78.20

Tools

Get Citation

Copy Citation Text

Xuqing LI, Guangya LI, Zhiyi ZHANG, Ziyi WANG. Lightweight Detection Algorithm of Power Equipment Based on Improved YOLOv5[J]. Optoelectronic Technology, 2024, 44(1): 47

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites