search

QingyongHu / SpinNet

[CVPR 2021] SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration
MIT License
266 stars 34 forks source link

Comparisons with a recent work DIP #1

Open yiming107 opened 3 years ago

yiming107 commented 3 years ago

Hi, very nice work. I'm working on a related topic and before your method I was checking DIP, a recent work published in ICPR ([43] in your paper). My impression was that they are quite effective in generalisation too (It achieves almost the same performance in ETH dataset as SpinNet based on the reported numbers). Well, I notice their work was not in comparison in your work, just wondering if there are any particular reasons why DIP is not compared in the tables (I meant tables for 3DMatch and ETH)? Thanks a lot.

QingyongHu commented 3 years ago

Hi @yiming107, thanks for pointing out this in our paper. We will update the paper in the next version, and to see if we could add other generalization results of DIP (e.g. 3DMatch->KITTI, KITTI->3DMatch).

Thanks again for your interest in our work!