Research on Pedestrian Detection Algorithm Based on MobileNet-YoLo

Table 8

Comparison of target detection results of different algorithms.

Models	Basic network	mAP (%)	Speed (FPS)	Params (MB)
YoLov3	Darknet53	78.30	14.25	61.63
YoLov4	CSPDarknet53	82.30	20.37	64.04
MobileNetv3-YoLov3^1	MobileNetv3	69.94	46.92	23.29
MobileNetv3-YoLov4^2	MobileNetv3	71.87	49.04	11.41
YoLov4-tiny	CSPDarknet53_tiny	68.17	107.76	5.92
YoLov5s	CSPDarknet53	72.43	67.14	7.12
YoLov5x	CSPDarknet53	81.18	10.09	87.37
MobileNet-YoLo	MobileNetv3	73.17	73.39	3.23

¹MobileNetv3-YoLov3 and²MobileNetv3-YoLov4 are models for adjusting the backbone feature extraction network.