Custom generated dataset with multiple pointclouds

[toobeeh]

Hello! This is the first time i'm working with torchpoints3d and i'm not quite sure if the way I imageine Datasets is correct.

I have annotated pointclouds that get annotated and generated by code, and i want to train a model with them. I wrote a pytorch geometric dataset that loads and processes the saved poinclouds, but now i'm not sure how to handle the data.

As i understand it, the dataset returns a Data object per index that holds information to a single point. My dataset has many pointclouds, so at some point there would be another Data object with the same coordinates but a different label. How should this be handled? Should the pointclouds be combined to a single big cloud?

Thanks in advance!

[Pippo809]

Seconding this, is it still open?