Table 2 Time statistics of different networks using the same inputs.

From: An efficient point cloud semantic segmentation network with multiscale super-patch transformer

Method

Data representation

Parameter size

Training time (min)

Inference time (min)

PointNet8

Sampling points

1.93M

105

33

PointNet++9

Sampling points

0.97M

557

185

DGCNN27

Sampling points

0.99M

175

45

SPGraph28

Super-points

0.25M

2471

864

Ours

Super-patches

5.12M

15

6