Table 3.

Obstacle detection comparisons using radar images with different rendering radiuses on the nuScenes validation dataset [37]. The top sub-table is about average precision metrics. In addition, the bottom sub-table is about average recall metrics.

Radius	Backbone	Scale	$\mathit{AP}$(100)	${\mathit{AP}}^{0.50}$(100)	${\mathit{AP}}^{0.75}$(100)	${\mathit{AP}}^{\mathit{s}}$(100)	${\mathit{AP}}^{\mathit{m}}$(100)	${\mathit{AP}}^{\mathit{l}}$(100)
$r=1$	ResNet-50	800	70.1	90.4	77.1	52.3	68.4	81.0
$r=3$	ResNet-50	800	69.6	89.5	75.8	52.3	67.9	80.4
$r=5$	ResNet-50	800	64.5	87.1	70.0	47.1	63.2	75.2
$r=7$	ResNet-50	800	70.2	89.9	76.7	54.3	68.6	80.6
$r=9$	ResNet-50	800	69.7	89.9	75.7	52.8	68.5	79.8
$r=11$	ResNet-50	800	68.3	89.0	74.5	52.9	66.6	79.2
Radius	Backbone	Scale	$\mathit{AR}$ (1)	$\mathit{AR}$ (10)	$\mathit{AR}$ (100)	${\mathit{AR}}^{\mathbf{s}}$ (100)	${\mathit{AR}}^{\mathbf{m}}$ (100)	${\mathit{AR}}^{\mathbf{l}}$ (100)
$r=1$	ResNet-50	800	12.8	64.5	76.9	66.6	75.6	84.9
$r=3$	ResNet-50	800	12.7	64.5	76.6	65.2	75.5	84.7
$r=5$	ResNet-50	800	12.4	60.2	73.3	59.9	73.2	80.8
$r=7$	ResNet-50	800	12.7	65.2	77.4	66.8	76.3	85.0
$r=9$	ResNet-50	800	12.6	64.1	76.7	65.5	76.0	84.2
$r=11$	ResNet-50	800	12.6	63.1	76.3	65.5	75.2	84.2