Closed ckxz closed 4 years ago
Looks like I updated this project and didn't check if it still works with the shapegan repository. Sorry! Will look into fixing this later today.
So sorry for not looking into this earlier :( It was simply a missing variable. Setting min_size
to 0 is fine, it just skips a sanity check that would throw out bad meshes. If you don't have super thin meshes or non-watertight meshes, you don't need that check.
Are you still trying to reproduce the shapegan project? I'd like to know how far you got.
Hey, no worries. I've been working with min_size = 0, as you suggest. I am currently training the point-gan. Let's see if it produces something interesting out of my dataset! I might bug you with other questions as I dig deeper in the project :)
I am trying to reproduce your work on ShapeGAN ([https://github.com/marian42/shapegan]) with my own set of .stl meshes. When producing sdf point clouds out of my meshes with mesh_to_sdf package: indeed, variable 'unit_sphere_sample_count' is not defined in surface_point_cloud.py. I get around this error by setting argument "min_size" to 0 in "sample_sdf_near_surface", although the BadMeshException would be really helpful. Could you please explain me what exactly the variable 'unit_sphere_sample_count' is and how should it be defined? Thanks in advance and keep up the awesome work!