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
857 stars 180 forks source link

Cartesian and cylinder #127

Open RuixiangXue opened 2 years ago

RuixiangXue commented 2 years ago

Hi,thanks for your great work! Cylinder voxelization is a great idea especially for lidar point cloud,I want to adopt this method into my work,but I'm confused that public available dataset such as SemanticKITTI provide cartesian coordinate data,so here is regular pipline

1.transform cartesian into cylinder coordinats 2.quantize cylinder coodinate(how to judge the best precision with minium error of coordinates transformation?) 3.input cylinder coordinates into network 4.transform cylinder coordinates back to cartesian coordinates

xinge008 commented 2 years ago

how to judge the best precision with minium error of coordinates transformation? In my opinion, these hyper-parameters are closely related to your deploy environments, and usually need a lot of ablation studies.

RuixiangXue commented 2 years ago

Thank u for your reply!

xinge008 @.***> 于2022年5月19日周四 16:45写道：

how to judge the best precision with minium error of coordinates transformation? In my opinion, these hyper-parameters is closely related to your deploy environments, and usually need a lot of ablation studies.

— Reply to this email directly, view it on GitHub https://github.com/xinge008/Cylinder3D/issues/127#issuecomment-1131416584, or unsubscribe https://github.com/notifications/unsubscribe-auth/AULA2FVDUD7ROINLGZPU4ULVKX5RZANCNFSM5ULBZQZQ . You are receiving this because you authored the thread.Message ID: @.***>