Generating Semantic Labels for densified (Output) point cloud?

[Guru-Deep-Singh]

Dear Authors,
I plan to train a semantic segmentation model from the predicted (Output) densified point cloud. However I wanted to know if you can generate the semantic labels for the predicted cloud as well?

Even if I train a semantic network with the ground truth cloud of the network i.e the pointcloud uses as GT for the densification. I would still need the ground truth for the predicted cloud for evaluation purposes.

Thank You