Pointnet_Pointnet2_pytorch icon indicating copy to clipboard operation
Pointnet_Pointnet2_pytorch copied to clipboard

Published 20 hours ago • verified publisher icon yanx27

How to compute input channel of feature_propagation layer?

Open xuanzhezhao opened this issue 3 years ago • 1 comments

As mentioned in paper, we use skip connection in feature propagation layer, but I'm confused by this line in pointnet2_part_seg_msg.py : self.fp1 = PointNetFeaturePropagation(in_channel=150+additional_channel, mlp=[128, 128])

How do you get the value of 150 in this line? Thanks for your help.

Best, Xuanzhe

xuanzhezhao avatar Oct 02 '21 00:10 xuanzhezhao

It's 128+the number of object classes+6 In this case, there are 16 object classes.

glhr avatar Oct 10 '21 19:10 glhr