Table 1.

Classification accuracy of different networks (%). (mA and OA denote the mean accuracy of each class of point cloud classification and the overall accuracy of point cloud classification, respectively.).

Method	Input	Points	mA	OA
Pointwise-CNN [23]	xyz	1k	81.4	86.1
ECC [18]	xyz	1k	-	87.4
PointNet [20]	xyz	1k	86.2	89.2
SCN [39]	xyz	1k	87.6	90.0
Kd-Net [14]	xyz	1k	86.3	90.6
PointNet++ [21]	xyz	1k	-	90.7
KCNet [34]	xyz	1k	-	91.0
Spec-GCN [15]	xyz	1k	-	91.5
PointCNN [25]	xyz	1k	88.1	92.2
DGCNN [35]	xyz	1k	90.2	92.2
GAPNet [17]	xyz	1k	89.7	92.4
Spec-GCN [15]	xyz+normal	1k	-	91.8
Pointconv [6] AGCN [41]	xyz+normal xyz+normal	1k 1k	- 90.7	92.5 92.6
PointNet++ [21]	xyz+normal	5k	-	91.9
SpiderCNN [7]	xyz+normal	5k	-	92.4
SO-Net [28]	xyz+normal	5k	90.8	93.4
DNet	xyz	1k	90.9	93.6