Table 7. Comparison with other methods.

Models	$\mathbf{m}\mathbf{A}\mathbf{P}\mathbf{@}\mathbf{0.5}\mathbf{/}\mathbf{\%}$	AR	$\mathit{P}\mathit{a}\mathit{r}\mathit{a}/{\mathbf{10}}^{\mathbf{6}}$	GFLOPs	F1-score/%	FPS
Faster R-CNN	73.2	52.1	28.3	940.9	60	11
Cascade R-CNN	54.8	61.2	87.9	110.6	56.4	24
SSD	72.7	54	26.2	62.7	61	76
EfficientDet	49.8	44.8	3.8	5.2	46	27
YOLOv3-tiny	76.1	68.5	8.7	13.0	71.2	220
YOLOv5n	80.6	76.1	1.7	4.2	77.1	218
YOLOv5s	85.2	82.6	7.02	15.8	83	156
YOLOv8n	84.2	77.9	3.0	8.1	80.3	243
YOLOv8n-BIFPN	84.5	80.5	3.0	8.1	81.3	238
YOLOv8n-GhostConv	83.6	81.4	2.8	7.7	80.5	256
YOLOv8n-CondConv	84.9	78.5	3.0	7.8	81.4	231
YOLOv8n-DWConv	85.6	78	3.0	7.9	81.8	237
OBC-YOLOv8	86	81.1	3.2	8.0	81.9	222