Open matfcp opened 3 months ago
Hi!
I am currently working with point cloud pose regression and classification and I implemented PointNet as the basis for my architecture.
My question is: why use 2D convolutions instead of 1D convolutions? What are the reasons and advantages for that approach?
Thanks in advance!
Hi!
I am currently working with point cloud pose regression and classification and I implemented PointNet as the basis for my architecture.
My question is: why use 2D convolutions instead of 1D convolutions? What are the reasons and advantages for that approach?
Thanks in advance!