Scale space reconstruction does not return a mesh

[oleg-alexandrov]

The "Scale-Space Surface Reconstruction" algorithm states that it creates a reconstructed mesh. It does not however return a triangulated mesh, rather some kind of object from which one has to figure out how to extract the mesh.

I looked in a bunch of places and all of them show the same thing.

https://doc.cgal.org/latest/Scale_space_reconstruction_3/Scale_space_reconstruction_3_2scale_space_8cpp-example.html#_a2 https://doc.cgal.org/latest/Scale_space_reconstruction_3/classCGAL_1_1Scale__space__surface__reconstruction__3.html#a300d0992b3be8caeacc1da30c5315eff

It would be helpful if the tutorial for this functionality would do the same as for the rest of examples. Input should be points, output should be a mesh the user can work with.

[stla]

You can easily create the mesh. See code below. Don't pay attention to the Rcpp::NumericMatrix.

Mesh3 SSSreconstruction_cpp(
  const Rcpp::NumericMatrix pts, const size_t scaleIterations,
  const unsigned nneighs, const unsigned nsamples,
  const bool separateShells, const bool forceManifold, const double borderAngle
) {

  std::vector<Point3> points = matrix_to_points3<Point3>(pts);

  SSS_reconstruction SSSR(points.begin(), points.end());
  SSS_smoother smoother(nneighs, nsamples);
  SSSR.increase_scale(scaleIterations, smoother);
  SSS_mesher mesher(
    smoother.squared_radius(), separateShells, forceManifold, borderAngle
  );
  SSSR.reconstruct_surface(mesher);

  Mesh3 mesh;
  for(
    SSS_point_iterator it = SSSR.points_begin(); it != SSSR.points_end(); ++it
  ) {
    mesh.add_vertex(*it);
  }
  for(
    SSS_facet_iterator it = SSSR.facets_begin(); it != SSSR.facets_end(); ++it
  ) {
    std::array<size_t, 3> face = *it;
    mesh.add_face(vxdescr(face[0]), vxdescr(face[1]), vxdescr(face[2]));
  }
  
  return mesh;
}

[oleg-alexandrov]

Thank you. This is helpful. Likely users will be grateful if the SSS_reconstruction class has an operator to return the mesh, maybe just the code you mention. And if that operator is added, the existing examples can be standardised to use it. I pointed out above two locations where the SSS reconstruction is used and would help if a mesh is returned. Another one is https://github.com/CGAL/cgal/blob/a8e87526348d2d1faa2ea893747314853ee6868d/Poisson_surface_reconstruction_3/examples/Poisson_surface_reconstruction_3/tutorial_example.cpp.

CGAL is a great library but its API takes a while to understand and the variety of structures used and approaches can be very confusing. I understand why, it is a heterogenous set of libraries by many people, and also very flexible, but it is hard to understand. Having at least an algorithm return a mesh rather than the user having to understand how to make that mesh, would help.

[afabri]

Note that scale space reconstruction creates a triangle soup, so maybe we should have a method that writes into a vector of points and a vector of vertex triples and the user then orients them and constructs a surface mesh as in this example of the Polygon Mesh Processing package.

[oleg-alexandrov]

That will work, I think. It will also make the example at https://github.com/CGAL/cgal/blob/c36d6df775f930880eae243385f33380d3ccc479/Poisson_surface_reconstruction_3/examples/Poisson_surface_reconstruction_3/tutorial_example.cpp#L13 have uniform treatment for the three kinds of surface reconstruction in it.

It would be nice if the example https://doc.cgal.org/latest/Scale_space_reconstruction_3/Scale_space_reconstruction_3_2scale_space_8cpp-example.html#_a2 is modified too.

Thank you for looking into this. Such small things make the library a lot more user-friendly.