search

qcraftai / pillarnext

PillarNeXt: Rethinking Network Designs for 3D Object Detection in LiDAR Point Clouds (CVPR 2023)
Other
176 stars 13 forks source link

About encode the pointcloud #15

Closed ouyangziyao closed 5 months ago

ouyangziyao commented 7 months ago

Hi, sir. The point cloud encoding into pillars seems to not use any sampling strategies such as random sampling to limit the number of points in each pillar and the number of pillars. Is it because not using sampling strategies makes the model perform better?

Konstantin5389 commented 5 months ago

we use dynamic encoding. it's a common way