pointnet2 icon indicating copy to clipboard operation
pointnet2 copied to clipboard

Published 20 hours ago • verified publisher icon charlesq34

Regarding 1x1 convolution filter

Open yash1112 opened this issue 4 years ago • 0 comments

Why is a 1x1 convolution filter used here, while shared mlp is mentioned in the original paper? How is the model able to learn the features with just a 1x1 filter, which cannot detect spatial structures?

yash1112 avatar Jul 28 '20 11:07 yash1112