ma-xu/pointMLP-pytorch

GEOMETRIC AFFINE MODULE

Closed this issue · 4 comments

qingfeng0830 commented

Regarding the part of the geometric mapping module, why does the point cloud show a normal distribution after such an operation?

ma-xu commented

Thanks for pointing out this. We notice that this presentation is not right (it will not affect the performance), it is not a normal distribution. We will revise this sentence. Thanks again for pointing out this error.

qingfeng0830 commented

Thanks for your answer. And what is the reason for this transformation? Is the original irregular local point cloud mapped to a relatively regular distribution?

ma-xu commented

The reason behind this module is that we want to know if there is a simple method to make up the loss of local operators (besides the MLP extractor). After several experiments, we get this module.

Is the original irregular local point cloud mapped to a relatively regular distribution?
It's hard to tell if this could be considered as a reason.

ma-xu commented

@qingfeng0830 I will close this issue since there is no further discussion. Feel free to reopen it if necessary.