Mesh not loaded correctly about trimesh HOT 5 CLOSED

Hey- what's probably happening here is that the automatic generation of face normals through cross products (since OFF doesn't include normals) detected a zero- area face and removed it.

You can verify this by checking the datastore (before accessing mesh.faces):

print(m._data['faces'].shape)

The reason face normals are auto-generated and degenerate faces are removed when you access mesh.faces is so that there is no chance of getting an inconsistent shape between mesh.face_normals and mesh.faces

I agree the silent alteration is potentially questionable, however the alternatives seems to me to be bad:

• have mesh.face_normals potentially include NaN entries (would dump warnings everywhere and be annoying).
• don't do the check when accessing mesh.faces (which would mean mesh.faces and mesh.face_normals potentially have different shapes depending on access order)
• keep doing the same thing but output a log.warning when there's a degenerate face (this could be OK)
• do the check but raise an error when there's a degenerate face (annoying)

I'm open to suggestions :)

from trimesh.

I see. Thanks a lot for the clarification :))
I think putting warning should be good. At least one knows what happens.
But do you mean that after calling once m.faces(), there is no way to get the original set of faces?

The reason I need original faces is that I am using a mesh segmentation algorithm which outputs a label for each face. Missing faces causes discrepancy between labels and faces. How I fix this now is that I access faces directly via m._data['faces'].

from trimesh.

Cool, I added the warning. Oh, the other thing you could do (I forgot I added a flag for this- I should work on docs at some point):

mesh = trimesh.load('blah.off')
mesh._validate = False #turn off face/normal validation

print(mesh.faces.shape)

from trimesh.

Awesome. Thank you!

from trimesh.

NP.

from trimesh.