search

charlesq34 / pointnet

PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation
Other
4.66k stars 1.44k forks source link

Using conv2d rather than linear layer #169

Closed Steve-Tod closed 5 years ago

Steve-Tod commented 5 years ago

Hi, I read your paper and your code and I got one question. Why do you use conv2D instead of linear to implement mlp? Like here: https://github.com/charlesq34/pointnet/blob/d64d2398e55b24f69e95ecb549ff7d4581ffc21e/models/pointnet_cls.py#L44

luis-gonzales commented 5 years ago

@Steve-Tod It's simply a convenient way to do the weight-sharing of the MLPs (across number of samples)

Steve-Tod commented 5 years ago

I got it, Thanks.