The Most Efficient Way to Convert Meshlib Data to VTK PolyData

[pty0220]

Hello,

I am currently working with both VTK and Meshlib, frequently switching between the two.

The issue I am facing is that the mesh types used by these two SDKs are different.
To overcome this, I am extracting faces and vertices as numpy arrays and then using these to create the respective mesh types for each SDK.

Is there a more efficient way to handle this type conversion?
Does Meshlib perhaps offer built-in support for VTK that could simplify this process?

Thank you for your time and help.

from meshlib import mrmeshpy
import meshlib.mrmeshnumpy as mrmeshnumpy

import vtk
from vtk.util.numpy_support import numpy_to_vtk, numpy_to_vtkIdTypeArray, vtk_to_numpy

import numpy as np

############################################################
### MeshLib -> VTK
############################################################

path = "path/your/mesh.stl"
mesh = mrmeshpy.loadMesh(path)

faces = mrmeshnumpy.getNumpyFaces(mesh.topology)
verts = mrmeshnumpy.getNumpyVerts(mesh)

points = vtk.vtkPoints()
points.SetData(numpy_to_vtk(verts, deep=1))

isize = vtk.vtkIdTypeArray().GetDataTypeSize()

req_dtype = np.int32 if isize == 4 else np.int64
n_elems = faces.shape[0]
n_dim = faces.shape[1]

cells = vtk.vtkCellArray()
cells.SetCells(n_elems,numpy_to_vtkIdTypeArray(np.hstack((np.ones(n_elems)[:, None] * n_dim, faces)).astype(req_dtype).ravel(), deep=1))

polyData = vtk.vtkPolyData()
polyData.SetPoints(points)
polyData.SetPolys(cells)

############################################################
### VTK -> MeshLib
############################################################

path = "path/your/mesh.stl"
reader = vtk.vtkSTLReader()
reader.SetFileName(path)

polyData = reader.GetOutput()
nFaces = polyData.GetNumberOfPolys()
faces = vtk_to_numpy(polyData.GetPolys().GetData())
faces = faces.reshape(nFaces, int(len(faces)/nFaces))
faces = faces[:, 1:]
faces = np.int32(faces)

verts = vtk_to_numpy(polyData.GetPoints().GetData())
verts = np.float32(verts)
verts = verts*np.array([-1,-1,1])

mesh = mrmeshnumpy.meshFromFacesVerts(faces, verts)

[Grantim]

Hello!

I think only other way is to export -> import witch is likely less efficient than numpy conversions way. We could add more import/export functions (to filesystem or to suggested data types), but we don't want to have vtk dependency.