Fuzzy output of the DMTet model

[Siziff]

Hello everyone, I'm trying to run the dmtet_tutorial.ipynb file on my data, but I get very poor quality output.

It's my input (I visualized a point cloud in the form of spheres in the Kaolin omniverse program so that it could be better seen in the picture)

It's my output after 15000 iterations

I tried to normalize the data in a different way, changed the learning rate and the number of iterations for training, but nothing helps, tell me what can be improved in the model, in the algorithm or in the data to get the best result.

[silence401]

same question

[otakudj]

Same question