请问一下kitti的背景类为什么不需要训练

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

Open wfs123456 opened 1 year ago

wfs123456 commented 1 year ago

您好，请问训练kitti时忽略背景类的训练，预测的时候背景类全部预测为前景，这样如何评估呢？另外如果我想加上训练背景类的话，semantickitti.yaml中 dataset_params中的ignore_label如何设置，设置为-1吗