Table 8.

Comparison table of training results of different algorithms.

Model	Image Size	Params/MB	Model Size/MB	P/%	R/%	F1-Score/%	mAP0.5/%	mAP@0.5:0.95/%	FPS
SSD	512 × 512	41.18	200	83.5	78.9	41.8	77.5	/	48.5
Faster-RCNN	1000 × 600	60.17	420	87.5	82.3	44.5	82.6	/	11.6
YOLOv3	640 × 640	61.95	123.5	86.5	84.5	46.8	92.3	60.7	37.59
ML-YOLOv3	640 × 640	18.15	37	89.5	85.8	47.5	90.2	58.2	88.88
YOLOv4	640 × 640	52.92	112.6	87.2	85.1	47.4	91.5	58.2	45.3
YOLOv5s	640 × 640	7.06	14.4	90.3	86.2	48.2	89.8	57.1	133.33
YOLOv5u	640 × 640	6.52	11.4	89.8	86.2	47.5	88.5	56.2	136.5
YOLOv5s-DM	640 × 640	7.06	14.4	90.5	86.1	48.8	90.2	56.5	133.33
YOLOv6s	640 × 640	/	38.16	89.5	85.8	48.3	90.9	57.9	79
YOLOv7-w6	640 × 640	69.83	140.1	90.1	87.2	48.5	91.7	58.8	55.4
YOLOv5s-Improved	640 × 640	5.06	12.5	91.2	87.1	51.5	92.4	59.2	142.66