Table 3 Performance comparison on the...

Table 3

Performance comparison on the KITTI test set for all classes 3D detection in the moderate difficulty and also the comparison of the 3D mean average precision (mAP). All the 3D detection results are evaluated with average precision and calculated by 40 recall positions.

Models	3D Mod.			3D mAP
Models	Car	Ped.	Cyc.	3D mAP
PointRCNN, [25]	75.64	39.37	58.82	60.33
Point-GNN, [27]	79.47	43.77	63.48	63.90
3DSSD, [33]	79.57	44.27	64.10	65.01
IA-SSD, [35]	80.32	41.03	66.25	64.39
SECOND, [32]	75.96	35.52	60.82	59.29
PointPillars, [15]	74.31	41.92	58.65	60.65
TANet, [17]	75.94	44.34	59.44	61.71
Part-A², [26]	78.49	43.35	63.52	63.99
SVGA-Net, [11]	80.47	40.39	62.28	62.92
PV-RCNN, [24]	81.45	43.29	63.71	64.74
SA-Det3D, [1]	81.46	40.89	68.54	65.04
PDV, [12]	81.86	40.56	67.81	65.31
Ours	82.00	42.25	66.46	65.64

Models	3D Mod.			3D mAP
Models	Car	Ped.	Cyc.	3D mAP
PointRCNN, [25]	75.64	39.37	58.82	60.33
Point-GNN, [27]	79.47	43.77	63.48	63.90
3DSSD, [33]	79.57	44.27	64.10	65.01
IA-SSD, [35]	80.32	41.03	66.25	64.39
SECOND, [32]	75.96	35.52	60.82	59.29
PointPillars, [15]	74.31	41.92	58.65	60.65
TANet, [17]	75.94	44.34	59.44	61.71
Part-A², [26]	78.49	43.35	63.52	63.99
SVGA-Net, [11]	80.47	40.39	62.28	62.92
PV-RCNN, [24]	81.45	43.29	63.71	64.74
SA-Det3D, [1]	81.46	40.89	68.54	65.04
PDV, [12]	81.86	40.56	67.81	65.31
Ours	82.00	42.25	66.46	65.64

Note: Our model’s results are shown in bold, and the best results are underlined.

[ViewLarge]

Sharing Unavailable