Real-Time Detection of Insulators and Drop Fuses Based on Improved YOLOv4

Table 4

Evaluation of models.

Model	Clustering algorithm	Precision (%)	Recall (%)	mAP (%)	(%)	(%)	(%)
YOLOv4	K-means	84.8	93.5	93.3	95.4	91.0	93.4
YOLOv4	K-means++	85.8	93.3	93.2	95.4	91.1	93.0
YOLOv4	FCM	85.6	94.1	94.2	95.3	93.3	94.1
YOLOv4	DCN	88.2	94.3	93.5	96.0	92.8	91.7
YOLOv4	SDCN	90.0	94.9	94.2	96.1	94.9	91.7
Improved YOLOv4	K-means	89.2	96.7	95.9	97.0	95.9	94.8
Improved YOLOv4	K-means++	86.0	96.2	96.0	96.9	95.9	95.4
Improved YOLOv4	FCM	88.0	96.7	97.0	97.7	97.4	95.8
Improved YOLOv4	DCN	89.2	93.7	94.2	94.8	94.2	93.8
Improved YOLOv4	SDCN	89.8	93.6	94.0	94.1	94.3	93.7