How to keep boundaries and features when using stl generalOffsetMesh to repair stl？

[WHUT666]

How to keep the border and feature when using stl generalOffsetMesh to repair stl, thank you very much！
stl file：
qlktgy4.zip

from meshlib import mrmeshpy as mm

# load mesh
mesh = mm.loadMesh(r"G:\case\stloffset\qlktgy4.stl")

# heal mesh (your mesh has self-intersectoins and not connected parts)
goParams = mm.GeneralOffsetParameters()
goParams.signDetectionMode = mm.SignDetectionMode.HoleWindingRule
mesh = mm.generalOffsetMesh(mesh,0,goParams)

mm.saveMesh(mesh, r"G:\case\stloffset\resMesh.stl")

[Grantim]

Hello! As far as it is voxel based algorithm it has voxel size precision, to make precision better you can use smaller voxel like this;

goParams.voxelSize = mm.suggestVoxelSize(mesh,1e9) # using approx voxels size to fit box of 1'000'000'000 voxels

[WHUT666]

Thank you for your reply. This method takes too long. Is there a way to add special edge or boundary information when reconstructing the point cloud?

[Grantim]

Best way is to find balanced voxel size that will meet both speed and precision requirements also you can try this #3026 (comment)

[WHUT666]

Thank you very much！

[Fedr]

The other option is to set GeneralOffsetParameters::mode = Mode::Sharpening

Please give it a try.

[WHUT666]

Cool！This works well, except for some small geometric features.Can it be further repaired?

from meshlib import mrmeshpy as mm

# load mesh
mesh = mm.loadMesh(r"G:\case\stloffset\qlktgy4.stl")

# heal mesh (your mesh has self-intersectoins and not connected parts)
goParams = mm.GeneralOffsetParameters()
goParams.signDetectionMode = mm.SignDetectionMode.HoleWindingRule
goParams.mode=mm.GeneralOffsetParametersMode.Sharpening
mesh = mm.generalOffsetMesh(mesh,0,goParams)

mm.saveMesh(mesh, r"G:\case\stloffset\resMesh.stl")

[Fedr]

Cool! It looks that the groove in highlighted area is below voxel size. So the only way of improvement is decreasing the voxel size.

[WHUT666]

It takes several attempts to reduce the size, Is there an adaptive voxel size interface or any subsequent development plans?Is there an adaptive voxel size interface or any subsequent development plans?

[Grantim]

Hello! Now we don't have adaptive voxel size algorithms. Also, we plan to expose CUDA versions of the offset algorithms to python in the future.