Table 3.

Comparison between RDFNet and other object detection methods. The italic value indicates the top performance.

Methods	P	mAP@0.5	R	FPS
YOLOv5s	0.597	0.552	0.578	20.12
YOLOv5m	0.608	0.556	0.585	18.94
YOLOv5l	0.606	0.563	0.582	17.98
YOLOv3-tiny	0.589	0.5	0.566	22.74
Faster R-CNN	0.421	0.421	0.825	23.14
Mask R-CNN	0.42	0.42	0.824	17.89
RDFNet	0.623	0.569	0.579	20.24