ma-xu/pointMLP-pytorch

ScanObjectNN evaluation

Closed this issue · 2 comments

EternityDS commented

Hi,

Thanks for your great work! I really like this idea and your brilliant paper-writing! I have a quick question about ScanObjectNN evaluation.

I notice this dataset has many variants shown in their paper, such as *objectdataset.h5, *objectdataset_augmentedrot_scale75.h5. And I think the result in PointMLP is based on *objectdataset_augmentedrot_scale75.h5.

May I ask is this the default setting to use this variant for ScanObjectNN classification task? And do you think all the results here use the same variant?

Thank you!

ma-xu commented

@EternityDS Thanks for your interest!
It may not be the default setting, but most people use this one by default(because of the difficulty).
They are using the same variant. I double-checked the papers and codes.

Also, you can present the results of all the variants as in some papers.

EternityDS commented

Thanks for your reply!