Train torch model

[samsgates]

can you please provide any one sample image / 3d mesh which is used for torch trained model, did you used same vertices count (3d mesh vertices) in trained model?

[AaronJackson]

The original models do have a fixed number of vertices, but the network was not trained on these. Instead if was trained on a voxelised version of the model. There is no way to preserve the correspondence when voxelisation is performed.

[samsgates]

thanks for your response. now i understand about technology limitation, do you have any idea about preserver the correspondence in output (or) convert voxel model to correspondence model (to maintain same vertices count) i just try to move forward output model to animation stage. if you have any thoughts please share