Table 2.

Object detection performance on our dataset under normal conditions with $IoU=0.5$.

Sensor	Method	${\mathit{AP}}_{3\mathit{D}}$ (%)	${\mathit{AP}}_{\mathit{BEV}}$ (%)
LiDAR	CT3D [60]	87.22	90.81
	SASA [61]	86.94	90.98
	TED [62]	85.83	91.42
	CPD [63]	87.95	92.45
	LISO [64]	88.64	92.88
LiDAR + RGB camera	CLOCs [7]	83.34	87.60
	EPNet [27]	85.15	88.85
	FocalsConv [5]	87.34	91.63
	SFD [8]	88.53	91.76
	GraphR-CNN [28]	86.67	90.93
	MoME [9]	91.02	93.63
	EPR [10]	90.65	94.38
	Ours	91.79	95.17