Lightweight Security Wear Detection Method Based on YOLOv5

Table 1

Comparison of results of multiple detection models on SWD dataset.

Model	Precision	Recall	mAP	FPS
Faster-RCNN	0.559	0.700	0.681	18.03
MobileNet-SSD	0.857	0.258	0.464	60.11
YOLO v5	0.877	0.809	0.849	83.33
DW-YOLO	0.869	0.760	0.805	90.90
DW-YOLO-attention	0.892	0.752	0.821	100.00