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
859 stars 180 forks source link

validation Iou #134

Closed fengjiang5 closed 2 years ago

fengjiang5 commented 2 years ago

Hello, it's a so exciting work on point cloud segmentation. And I'am try to use it as my backbone, so I have read your paper so many times. But still I have some question about the results on validation. I see your result on validation is 65.9, and no iou for each class. Can you announce them if possible. Thanks a lot. Hope for your reply.

xinge008 commented 2 years ago

valset_results The detailed results on valset.

fengjiang5 commented 2 years ago

Thanks a lot for your reply. Maybe the valid set has few motorcyclist and other-ground.

YEOMJINSEOP commented 10 months ago

Thansk for sharing helpful issues. However, i'm curious why validation results show "bus" instead of "other-vehicle"? In the cylinder3D paper, there's no "bus" label reported on semanticKITTI dataset. In config/label_mapping/semantic-kitti.yaml, "bus" is mapped to "other-vehicle" Could you tell me why "bus" result is shown in this code?