About point cloud encoder

[filaPro]

Hi, @buaacyw

Thanks for sharing your wonderful research. Just small question. Do I understand correctly, that you use point cloud encoder frozen with exactly the same weights, provided by Michelangelo paper? So this part of the model is trained only on ShapeNet, not ShapeNet + Objavarse?

[buaacyw]

Hi, thanks a lot for your interest in our work! We used exactly the same weights from its github repo. Different from its paper, the encoder-decoder released by Michelangelo in its repo is in fact trained on Objaverse + ShapeNet.

[filaPro]

Got it, thanks.
May I ask how did you know that they trained on Objaverse + ShapeNet?)

[buaacyw]

Aha, I asked the first author of Michelangelo about this.