Is there a way to load .stl file with finer detail

[liruiw]

When initializing PolyData, it seems like there's some sampling on the original stl file. I wonder if there's a way to turn off the sampling, and get a very fine mesh of the original stl file as a numpy array

[akaszynski]

It should be resampling and there’s no option to turn it on or off. Can you upload the file here for me to check it?

You can also check the number of faces and points in the polydata object and compare that with another mesh reading program.

[liruiw]

mesh = vtkInterface.PolyData(filename)
points = mesh.points
It's just these two lines of code. I've also been playing with vtk.STLReader and stl2numpy, they all provide the same number of points. In my case, I'd like to have a fine point cloud data in order to project those 3d points to image plane. If there's no option to adjust the resampling frequency, it seems like I have to repopulate the mesh myself. So, I wonder if the face and points returned by your program can be used to populate more points.

[akaszynski]

See the Subdivide function

mesh = vtkInterface.PolyData(filename)
help(mesh.Subdivide)

[liruiw]

It's throwing the exception Exception: Subdivision filter must be one of the following: 'butterfly', 'loop', or 'linear'. And in vanilla vtk, it's saying "vtkLoopSubdivisionFilter (0x55f8a8aeb420): Dataset is non-manifold and cannot be subdivided. Edge shared by 3 cells." So subdivision seems not working.

[akaszynski]

VTK requires the surface to be manifold for subdivision. Check out pymeshfix if you'd like to make it manifold.

[liruiw]

Thank you, the pymeshfix package is working well. The problem is that after repairing, the manifold seems to contain only half of the size of the original model. In my case, before fixed it has around 7000 points, and after repairing, the top half (3500 points) is reserved. Is there a way to work around that?

[akaszynski]

Do you mean that the face order is reversed so that the face normals are pointed inward? If so, you can correct the face ordering with:

mesh.GenerateNormals(auto_orient_normals=True)

[liruiw]

Oh sorry. The face order looks fine to me. I mean the model size is reduced by half after repairing.

[akaszynski]

Are you sure those points weren’t dupicates?

…

On Sun, Jul 1, 2018 at 09:52, Leroy357 ***@***.***> wrote: Oh sorry. The face order looks fine to me. I mean the model size is reduced by half after repairing. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <https://github.com/akaszynski/vtkInterface/issues/15#issuecomment-401590768>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ALbTP4Rj4-uiDSBCTQOZWQwYJAXK1MNsks5uCH_UgaJpZM4U6lkz> .

[liruiw]

mesh = vtkInterface.PolyData(abs_name)
meshfix = pymeshfix.MeshFix(mesh)
meshfix.Repair()
mesh = meshfix.mesh

Above is the code that I run. Those points were not duplicated (I print numberofpoints after first line, turns out to be 14224 points, and after the last line it becomes 3159). Do you know what's causing this? Also I wonder the meaning of the numbers printed when I used the meshfix. Thanks!

[akaszynski]

Can you zip the mesh and upload it?

[liruiw]

mesh.zip
Here it is. Thanks for your help.
Also I wonder if a z-buffer is implemented in python vtk or 3rd party libraries.

[akaszynski]

It looks like the input mesh contains several parts and isn't a single piece. Additionally, the part contains several self-intersections that can't be repaired automatically with pymeshfix. I'd try outputting or generating the mesh as a single object.

Also, when debugging, you should use mesh.Plot() in order to visualize the mesh after repairing it to see what changed in each step. pymeshfix removed all but one of the four isolated components, which is why only a quarter of the points remained after the "repair".

Z-buffer is implemented in VTK.