CGAL users discussion list

["Sebastien Loriot (GeometryFactory)" <>]

On 07/28/2014 05:11 PM, hcherkaoui wrote:
> Thanks for the reply  Sebastien,
>
> So I work on the binding and now it works fine, it pass all my tests and
> finally paraview confirm the good reconstruction. I've also bind
> Alpha_Shape_3 and like the Poisson process it works fine.
>
> However, I bind these process for my work so especially for Alpha_Shape_3
> it's far from the work done in the old project 'cgal-python'
> (http://cgal-python.gforge.inria.fr/).
> The aim of my work it to provide python module to do this:
>
> python-list-of-Point_3 => | my module (poisson or alpha_shape_3)| =>
> python-Polyhedron_3
>
> So my work could interest some peoples but may be not the official
> repository of cgal-binding, am I right?
> I produced hight-level-functions, to be clear this is the little API
> provided by my two bindings:
>
> CGAL.CGAL_Alpha_shape_3 :
>          Alpha_shape_3_reconstruction(points, alpha_coef)
>          \points: list of Point_3
>          \alpha_coef: python float, coef in front of the optimal value of
> alpha (.find_optimal_alpha(1))
>                            pass to the .set_alpha() methode
>          \return a Polyhedron_3
>
> CGAL.CGAL_Mesh_process_3
>          Computes_reconstruction_error(points, Polyhedron)
>          \points python-list of Point_3
>          \Polyhedron the re-construction polyhedron
>          \return a list of 3 python-floats
>                               [avg_distance of the point set,
>                                max distance points set <-> surface of the
> polyhedron (in % of avg_spacing)
>                                avg distance points set <-> surface of the
> polyhedron (in % of avg_spacing)]
>
>          Poisson_reconstruction(points,
>                                          normals,
>                                          sm_angle,
>                                          sm_radius,
>                                          sm_distance,
>                                          solver_name)
>          \points python-list of Point_3
>          \normals python-list returned by jet or pca_estimate_normals
>          \sm_angle python-float Min triangle angle in degrees
>          \sm_radius python-float triangle size w.r.t. pt set average spacing
>          \sm_distance python-float Surface Approximation error w.r.t. point
> set average spacing
>          \solver_name "Eigen - built-in CG" or "Eigen - built-in simplicial
> LDLt" for the compatibility of the
>           native solver in the machine, have to fixe this part...
>          \return a Polyhedron_3
>
>          Write_off_polyhedron(filename, Polyhedron)
>          \filename the name of the .off file produice (have to pass
> "my_file.off", with the .off)
>          \Polyhedron the re-construction polyhedron
>          \return None
>
> Could something like this interest the cgal-binding repos? or 'How' could
> something like this, finally interest the cgal-binding repos (with some
> changes) ?

I integrate a package in the cgal-bindings when the whole API of the
package is available.

> Then, I have two questions linked to my work, I want to produce an even
> easier API for my work, something like:
>
> my_polyhedron = CGAL.CGAL_Mesh_process_3.Poisson_reconstruction(points,
> normals, nb_facets)
>
> where nb_facets would be the final numbers of facets of the reconstructions
> polyhedron. But I find it hard to predict the good sm_angle, sm_radius and
> sm_distance for an appropriate nb facets... Could you give me some idea for
> the issue?
>
> Then, the second question is: how can we produce low-nb-facets polyhedron
> with Poisson reconstruction?
> I find it hard to produce 10_facets-polyhedrons or 50_facets-polyhedrons
> with that algorithm...Could you give me some idea for this problem?

You cannot predict the number of facets easily.
What you can do is make a first try with a bad approximation factor (0),
and no constraints on the shape of triangles, look at the result and
if needed ask for better quality iteratively depending on your time budget.

Sebastien.

> Thank Sebastien
>
> Regards
>
>
>
> --
> View this message in context: 
> http://cgal-discuss.949826.n4.nabble.com/Python-binding-of-Poisson-reconstruction-function-tp4659400p4659639.html
> Sent from the cgal-discuss mailing list archive at Nabble.com.