CGAL users discussion list

["Sebastien Loriot (GeometryFactory)" <>]

I loaded the model in the polyhedron demo, converted it into nef and I 
got 10002 vertices.
Then I ran the convex decomposition which indeed took some time but
provided a convex decomposition that is valid but obviously not the
minimal one.

The kind of model the algorithm was written for is more for mechanical
parts rather than "noisy" or smooth surfaces.
IMO, it should not be used in such cases.

Sebastien.

On 09/11/2014 08:56 PM, Dimitris Tzionas wrote:
> Some additional info, sorry for the 2nd email:
>
> POL.is_pure_triangle = 1 -> that means that luckily I don't have
> degenerate triangles.
>
>
> Also, step 3 (CGAL::convex_decomposition_3(NEF))  finished after much
> time (~1 hour?) and returned 25642 convex parts,
>
> while the original OFF file (attached) has 20000 triangles (problem
> correlated with what I reported about 2x the number of vertices).
>
>
> OFF
>
> 10002 20000 0
>
>
> 59.297 -0.455242 -70.6708
>
> 66.3921 -1.96008 -58.9775
>
>
> On Thu, Sep 11, 2014 at 8:54 PM, Dimitris Tzionas 
> <
> <mailto:>>
>  wrote:
>
>     Some additional info, sorry for the 2nd email:
>
>     POL.is_pure_triangle = 1 -> that means that luckily I don't have
>     degenerate triangles.
>
>
>     Also, step 3 (CGAL::convex_decomposition_3(NEF))  finished after
>     much time (~1 hour?) and returned 25642 convex parts,
>
>     while the original OFF file (attached) has 20000 triangles (problem
>     correlated with what I reported about 2x the number of vertices).
>
>
>     OFF
>
>     10002 20000 0
>
>
>     59.297 -0.455242 -70.6708
>
>     66.3921 -1.96008 -58.9775
>
>     ...
>
>
>
>
>     On Thu, Sep 11, 2014 at 8:09 PM, Dimitris Tzionas
>     
> <
>  
> <mailto:>>
>  wrote:
>
>         Hi all,
>
>         In a physics simulation I have a problem with concave meshes, so
>         it seems that I really have to do decomposition in convex
>         mesh-parts.
>
>         Since CGAL covered successfully past needs, I was hoping to use
>         this again.
>
>         It seems that it is doable only by using Nef_polyhedron_3.
>         http://doc.cgal.org/latest/Convex_decomposition_3/index.html
>
>         In order to have this representation:
>         - I first read a .off file (ideally this should happen from
>         memory, but my code is slower than CGAL's built-in reader from
>         file) in a Polyhedron_3. My mesh is closed and depicted fine
>         with the Polyhedron viewer demo.
>         - then I convert to a Nef_polyhedron_3.
>         - Finally I try to use CGAL::convex_decomposition_3
>
>         - The first part is very fast.
>
>         std::ifstream  file("/home/dimitris/Model_Hand_R.off");
>
>         file  >>  POL;
>
>         However instead of 10002 vertices I noticed that I get
>
>         POL.size_of_vertices = 20004
>
>         POL.is_closed = 1
>
>
>         - The second part takes ~15sec for 20k vertices.
>
>         Nef_polyhedron_3  NEF(  POL  );
>
>         NEF.is_empty: 0
>
>         Nef vertices: 20004
>
>         Nef edges: 60000
>
>         Nef facets: 40000
>
>         Nef volumes: 3
>
>
>         - The third part takes forever and never returns anything.
>
>         CGAL::convex_decomposition_3(  NEF  );
>
>
>         Could it be that I'm doing something very obviously wrong?
>
>         Could I get best practices hints for this problem?
>
>         What is a sensible run-time for CGAL's approach?
>
>
>         Some extra info on the current setup:
>
>
>         typedef  CGAL::Exact_predicates_exact_constructions_kernel        
> Kernel;
>
>         typedef  CGAL::Polyhedron_3<Kernel>                               
> Polyhedron;
>
>         typedef  CGAL::Nef_polyhedron_3<Kernel,  CGAL::SNC_indexed_items>  
> Nef_polyhedron_3;
>
>
>         Thank you in advance for potential hints,
>
>         Dimitris