Questions on using labels or not

[jsongnjupt]

Hi author, first, great work.

I have a question after I read your codes.
You mentioned in the paper that your point cloud segmentation is unsupervised, why I find using labels from label path in the codes, e.g., in "data_prepare_SemanticKITTI.py", "initialSP_prepare_SemanticKITTI.py".

In other words, if we have point cloud data without labels, can the codes run successfully since there is no label path?

Thx.

[zhang-zihui]

Hi, we only use the labels to delete those points that do not belong to the meaningful categories. For SemanticKiTTI, we exclude points that are not in the 20 classes. This is because the evaluation of this benchmark only considers these 20 classes.