search

xinge008 / Cylinder3D

Rank 1st in the leaderboard of SemanticKITTI semantic segmentation (both single-scan and multi-scan) (Nov. 2020) (CVPR2021 Oral)
Apache License 2.0
858 stars 180 forks source link

Pointwise Cylinder Refinement Model #151

Closed L-Reichardt closed 2 years ago

L-Reichardt commented 2 years ago

Hello,

I have a question about the Pointnet that generates the Cylinder features. Is there any particular reason why the input tensor has the dimension (N, 9) not (N, 3)? Do I understand correctly that this is contains the (polar coordinate points centered in voxels)(points in polar coordinates)(pointcloud)?

Also why is the data shuffled?

L-Reichardt commented 2 years ago

No need for answer anymore.