. 2024 Jan 30;10:e1830. doi: 10.7717/peerj-cs.1830

Table 2. Performance comparison of different feature fusion modules in the helmet detection dataset.

Model	Ghost	CA	SimSPPF	mAP@0.5	Parameters/million	t/s
YOLOv5s				0.841	7.01	0.011
Version-1	$\sqrt$			0.825	5.00	0.012
Version-2		$\sqrt$		0.855	7.05	0.014
Version-3			$\sqrt$	0.849	7.01	0.011
Version-4	$\sqrt$	$\sqrt$		0.830	5.04	0.014
Version-5	$\sqrt$		$\sqrt$	0.848	5.00	0.012
Version-6		$\sqrt$	$\sqrt$	0.853	7.04	0.014
YOLOv5s (ours)	$\sqrt$	$\sqrt$	$\sqrt$	0.872	5.04	0.014

Note:

The best results are in bold.