Table 1 Classification result on ModelNet40.

From: Frame points attention convolution for deep learning on point cloud

Methods

Accuracy

#params

Runtime (ms)

Kd-Net18

91.8

2.0 M

–

PointNet19

89.2

3.48 M

21

PointNet++20

90.7

1.48 M

33

DGCNN25

92.2

1.84 M

101

RGCNN42

90.5

2.24 M

97

3DTI-NET43

91.7

2.6 M

91

PointConv32

82.8

19.0 M

97

PointCNN30

92.2

0.45 M

103

KCNET44

91.0

0.9 M

–

3DmFV-Net15

91.6

4.6 M

44

KPConv33

92.9

6.15 M

121

RS-CNN38

92.6

1.28 M

66

PointBERT35

93.2

–

–

PT34

93.7

–

–

FPAC

93.9

1.94 M

62

  1. The best performance for each metric is shown in bold.
Back to article page