How do I make my own training set And train

[wangzaichao]

How do I make my own training set And train

[SajjadPSavoji]

I have a similar question too, Is the code to generate the synthetic dataset public? I'm unsure what was the preprocessing they've done on CADs. In the Supplementary Material they say

Objaverse assets vary extremely in the object size and mesh complexity. Therefore, we further normalize the objects and remove the disconnected components automatically based on the mesh edge connectivity graph, to make the objects suitable for learning pose estimation.

what is the exact normalization steps they perform?