The labeled points are randomly selected in each batch. How do you ensure that the labeled points in each scene are fixed?

[Dyb3438]

Hi @QingyongHu, thanks for your awesome work!

I have a question about the setup of your experiment.

Since the labeled points are randomly and temporarily selected while each batch data is generating, how do you ensure the labeled points in the same scene are fixed in multiple training epochs?

The randomly selected annotation points code is here: https://github.com/QingyongHu/SQN/blob/main/main_SemanticKITTI.py#L94