Using the cutMesh function produces unexpected results

[leye123]

Hello, MeshLibTeam.

I tried using cutMesh to cut a model and found that it did not report any errors, and cutting lines appeared on the model but it did not actually cut. Meshlib version 3.1.1.92

Then I found that when I imported this model into the Inspector, there was no problem with the model.

def cut_mesh_and_get_parts(mesh, points):
      surfaceLine = numpy_to_mesh_tripoints(mesh, points)
      print(surfaceLine)
      SearchPathSettings = mm.SearchPathSettings(mm.GeodesicPathApprox.DijkstraBiDir, 100)
      contour_result = mm.convertMeshTriPointsToClosedContour(mesh, surfaceLine, SearchPathSettings)
      cut_params = mm.CutMeshParameters()
      cut_params.forceFillMode = mm.CutMeshParameters.ForceFill.All
  
      cut_result = mm.cutMesh(mesh, [contour_result], cut_params)
      edge_paths = cut_result.resultCut
  
      if not edge_paths:
          print("Error: (edgePaths is None)!")
          return None, None
      else:
          print(f"edgePaths size={len(edge_paths)}")
  
      edge_path = edge_paths[0]
  
      one_part = mm.fillContourLeft(mesh.topology, edge_path)  
      other_part = mesh.topology.getValidFaces() - one_part
  
      area_one = mesh.area(one_part)
      area_other = mesh.area(other_part)
      print(f"onePart: {area_one} | otherPart: {area_other}")
  
      if area_one < area_other:
          mesh_one = mesh.cloneRegion(other_part)
          mesh_other = mesh.cloneRegion(one_part)
      else:
          mesh_one = mesh.cloneRegion(one_part)
          mesh_other = mesh.cloneRegion(other_part)
  
      return mesh_one, mesh_other

Cut point data:
data.json

model:
model.zip

[Fedr]

It appears that your model has tunnels, so cutting it along the contour does not separate the mesh on two parts, and mm.fillContourLeft returns all triangles. Could you please attach the model before cutting?

[leye123]

sure, here is his original model.

source.zip

[leye123]

Hey, I tried the approach from #5045. By trying to cut through sparse points, I found that cutting is possible, but in #5049 it is said that the bug is resolved.

Below is the code I solved:

def prepPoints(sort_points, numpoints=100, s=0, k=1):
    tck, u = splprep(sort_points.T, s=s, k=k) 

    new_points = np.array(splev(np.linspace(0, 1, numpoints), tck)).T

    return new_points

# cut points
projectSort = prepPoints(projectSort, 100, 0, 1)
endcurve_faces, _ = cut_mesh_and_get_parts(endcurve_faces, projectSort)

[Grantim]

Hello!

The issue here is that your contour has self-intersections:

MeshLib/source/MRMesh/MRContoursCut.h

Line 76 in 97e0bd8

* \warning Input contours should have no intersections, faces where contours intersects (`bad faces`) will not be allowed for fill

But you have enabled ForceFill

MeshLib/source/MRMesh/MRContoursCut.h

Line 47 in 97e0bd8

All //< fills all faces with bad ones, but on bad faces triangulation can also be bad (may have self-intersections or tunnels)

It has lead to tunnels, so fillContourLeft leaks and fills all of the mesh.

You can check it with cut_result.fbsWithContourIntersections.any().

Using mm.fillContourLeftByGraphCut will prevent leakage even if tunnels exist but in general it is better to fix intersections on input contour, one way to do it is to sample it (as you did trying approach from #5045).

P.S.
The reason behind #5045 was a bug on our side during conversion and same "sampling" approach avoids it. #5049 fixed this bug. While "sampling" approach helps you, the bug is not related to your problem.

[leye123]

Thank you for your reply!