My own data are only raw lidar points without 3D semantic and 3D bounding box labels. Can I still try your code to generate the groundtruth?

[Rango-T10000]

Really thanks for your excellent work.

My own data are only raw lidar points without 3D semantic and 3D bounding box labels. To generate the dense occ label(occ or non-occ), the naive method is to overlap a sequence of lidar frames like the description in your paper, but you also said this such a straightforward solution is only applicable to completely static scenes and ignores moving objects.

Therefore, according to your paper, "movable objects are cut out from the LiDAR points according to 3D bounding box labels!". And it can solve above problem. The prerequisites of your method to generate the occ label are lidar points with 3D semantic and 3D bounding box labels.

I am ok with no semantic, but I still want to generate the dense occ label(occ or non-occ), could you suggest if I can still use your code? I am not sure whether it's work or not. Really thanks.

And it seems someone met the same problem:#96 (comment).