CGAL error subtracting STL from a cube
nophead:
Thanks for the suggestion. A convexity of 1000 does not fix the CGAL
error. The STL in question is a 3D scan of my face, and I have a large
beard. Perhaps convexity needs to be 10,000. Who knows.
Jon
On 10/31/2020 4:29 AM, nop head wrote:
I think you need to set convexity in the import statement to get it to
preview correctly.
On Sat, 31 Oct 2020 at 03:20, rickan <richard@kandarian.com
mailto:richard@kandarian.com> wrote:
None of my ideas were relevant.
But I made two attempts to get what I think you want using
different methods
and both look oddly fractured in preview and do not render with
thin shells.
The surreal fractures in preview seem identical with both methods and
independent of thickness. I mean when I change the file the
preview doesn't
change except for the thickness if I change that. The fractures do not
change. But they migrate and morph oddly as if not part of the
model when
you alter the POV.
The errors from render are in the scad file.
I set it up to generate the shell using scale instead of translate
and the
surreal fractures show up in preview even when the shell thickness
is 50%.
However! It renders at that thickness! (It did not render at 25%
and at
least one thinner value I tried, probably 5%.) I suppose you could
do a
binary search for a thickness that is thinner than 50% which
renders. This
has me thinking that the stl interferes with itself when the shell
thickness
is too thin, i.e. one or more of the craggy features, e.g. in the
beard
perforate the shell. That's just a hypothesis. Well, I'm pretty
sure it
happens if the shell is thin enough, but I don't know why that
would be a
problem.
I made a third attempt just to subtract the face from a cube and that
previews and renders fine. I would expect, from my limited
experience, that
any difference or intersection with the stl would fail, not just a
difference or intersection of the stl with itself depending on the
relative
scale.
I did do a 1% translation to avoid the pesky ~0 thickness
membrane. Maybe
that's big enough to cause interference between the two surfaces.
Maybe that
should be reduced to 0.1%. I kind of doubt that's the problem though.
HollowFaceRMK.scad
<http://forum.openscad.org/file/t1112/HollowFaceRMK.scad
<http://forum.openscad.org/file/t1112/HollowFaceRMK.scad>>
HollowFaceRMK.png
<http://forum.openscad.org/file/t1112/HollowFaceRMK.png
<http://forum.openscad.org/file/t1112/HollowFaceRMK.png>>
--
Sent from: http://forum.openscad.org/ <http://forum.openscad.org/>
_______________________________________________
OpenSCAD mailing list
Discuss@lists.openscad.org <mailto:Discuss@lists.openscad.org>
http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org
<http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org>
OpenSCAD mailing list
Discuss@lists.openscad.org
http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org
The convexity argument is only used for displaying in preview. It has
NOTHING to do with CGAL errors which are from render. If you're getting
CGAL errors it means CGAL thinks your geometry is not a manifold or is
somehow malformed, so the only fix is going to be to change the geometry.
jon_bondy wrote
nophead:
Thanks for the suggestion. A convexity of 1000 does not fix the CGAL
error. The STL in question is a 3D scan of my face, and I have a large
beard. Perhaps convexity needs to be 10,000. Who knows.
Jon
On 10/31/2020 4:29 AM, nop head wrote:
I think you need to set convexity in the import statement to get it to
preview correctly.
On Sat, 31 Oct 2020 at 03:20, rickan <
richard@
> <mailto:
richard@
>> wrote:
None of my ideas were relevant.
But I made two attempts to get what I think you want using
different methods
and both look oddly fractured in preview and do not render with
thin shells.
The surreal fractures in preview seem identical with both methods and
independent of thickness. I mean when I change the file the
preview doesn't
change except for the thickness if I change that. The fractures do
not
change. But they migrate and morph oddly as if not part of the
model when
you alter the POV.
The errors from render are in the scad file.
I set it up to generate the shell using scale instead of translate
and the
surreal fractures show up in preview even when the shell thickness
is 50%.
However! It renders at that thickness! (It did not render at 25%
and at
least one thinner value I tried, probably 5%.) I suppose you could
do a
binary search for a thickness that is thinner than 50% which
renders. This
has me thinking that the stl interferes with itself when the shell
thickness
is too thin, i.e. one or more of the craggy features, e.g. in the
beard
perforate the shell. That's just a hypothesis. Well, I'm pretty
sure it
happens if the shell is thin enough, but I don't know why that
would be a
problem.
I made a third attempt just to subtract the face from a cube and that
previews and renders fine. I would expect, from my limited
experience, that
any difference or intersection with the stl would fail, not just a
difference or intersection of the stl with itself depending on the
relative
scale.
I did do a 1% translation to avoid the pesky ~0 thickness
membrane. Maybe
that's big enough to cause interference between the two surfaces.
Maybe that
should be reduced to 0.1%. I kind of doubt that's the problem though.
HollowFaceRMK.scad
<http://forum.openscad.org/file/t1112/HollowFaceRMK.scad
> <http://forum.openscad.org/file/t1112/HollowFaceRMK.scad>>
HollowFaceRMK.png
<http://forum.openscad.org/file/t1112/HollowFaceRMK.png
> <http://forum.openscad.org/file/t1112/HollowFaceRMK.png>>
--
Sent from: http://forum.openscad.org/
<http://forum.openscad.org/>
_______________________________________________
OpenSCAD mailing list
Discuss@.openscad
<mailto:
Discuss@.openscad
>
http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org
<http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org>
OpenSCAD mailing list
Discuss@.openscad
Discuss@.openscad
--
Sent from: http://forum.openscad.org/
Yes!
Setting convexity to 2 removes nearly all the fractures. Setting it higher
seems to remove them all but makes preview extremely laggy on my computer.
Without the distraction of the fractures the holes in the shell become
obvious.
In the attached image of the stl model subtracted from the cube with
viewport settings
$vpt = [-1.97, -9.31, 0.42]; $vpr = 90.00, 0.00, 90.00]; $vpd = 2.30;
There is what looks like a separate blob. It is not separate, however, it is
a hook in the stl which becomes an empty loop in the subtraction. Maybe that
is part of the problem.
HollowFaceIsland.png
http://forum.openscad.org/file/t1112/HollowFaceIsland.png
Maybe there are other nearly invisible features, e.g., long extremely thin
spikes, which are causing problems also.
--
Sent from: http://forum.openscad.org/
On 2020-10-31 13:37, adrianv wrote:
The convexity argument is only used for displaying in preview. It has
NOTHING to do with CGAL errors which are from render. If you're
getting
CGAL errors it means CGAL thinks your geometry is not a manifold or is
somehow malformed, so the only fix is going to be to change the
geometry.
Indeed, the convexity parameter has nothing at all to do with geometric
calculations. In my opinion that parameter should not even exist, but
that's another story.
Where is the STL that you are struggling with?
Carsten Arnholm
Thanks, everyone, for helping out.
The STL is huge (>10MB) so I am sending it to individuals who
volunteer. I will send a copy to Carsten; if anyone else is curious,
please let me know
Jon
On 10/31/2020 8:52 AM, arnholm@arnholm.org wrote:
On 2020-10-31 13:37, adrianv wrote:
The convexity argument is only used for displaying in preview. It has
NOTHING to do with CGAL errors which are from render. If you're getting
CGAL errors it means CGAL thinks your geometry is not a manifold or is
somehow malformed, so the only fix is going to be to change the
geometry.
Indeed, the convexity parameter has nothing at all to do with
geometric calculations. In my opinion that parameter should not even
exist, but that's another story.
Where is the STL that you are struggling with?
Carsten Arnholm
OpenSCAD mailing list
Discuss@lists.openscad.org
http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org
On 31.10.2020 15:13, jon wrote:
Thanks, everyone, for helping out.
The STL is huge (>10MB) so I am sending it to individuals who
volunteer. I will send a copy to Carsten; if anyone else is curious,
please let me know
I got your file, thanks.
I ran it through polyfix which reported a number of issues (see below),
and saved it in AMF format, as that is supported in OpenSCAD since
2019.05. I then ran it through your .scad code using OpenSCAD version
2020.10.11.nightly (git b416aa2) under Ubuntu.
However, it failed:
ERROR: CGAL error in CGAL_Nef_polyhedron3(): CGAL ERROR: assertion
violation! Expr: e->incident_sface() != SFace_const_handle() File:
/usr/include/CGAL/Nef_S2/SM_const_decorator.h Line: 329, location unknown
So instead I ran your STL through polyfix again and saved it as OBJ, as
that is much more compact than AMF (but OBJ is not supported in OpenSCAD):
$ polyfix "Jon-mesh-export-3 - MM Repair.stl" -out=Jon_Face_polyfix.obj
Parameters:
input_file = Jon-mesh-export-3 - MM Repair.stl
out = Jon_Face_polyfix.obj
polyhedron 0 ================= volume=1992, dtol=0.01, atol=1e-06,
maxiter=10
iteration 0: vertices=898980 faces=299660
warning: nonmanifold edges: uc(1)=898980
merged 749167 vertices
removed 12 collapsed or zero area faces
removed 7 duplicate faces
removed 2 nonmanifold faces
total changes=749188
warning: nonmanifold edges: uc(1)=5 uc(3)=4 uc(4)=1
iteration 1: vertices=149813 faces=299639
warning: nonmanifold edges: uc(1)=5 uc(3)=4 uc(4)=1
split 3 faces
removed 1 duplicate face
removed 3 nonmanifold faces
removed 6 zero area faces
total changes=13
warning: nonmanifold edges: uc(4)=1
iteration 2: vertices=149813 faces=299634
warning: nonmanifold edges: uc(4)=1
removed 1 unused vertex
total changes=1
warning: nonmanifold edges: uc(4)=1
iteration 3: vertices=149812 faces=299634
warning: nonmanifold edges: uc(4)=1
total changes=0
warning: nonmanifold edges: uc(4)=1
Summary:
polyhedron 0: vertices=149812 faces=299634 : warning:
nonmanifold edges: uc(4)=1
Writing: Jon_Face_polyfix_1.obj
... polyfix finished, time used: 0d 00h 02m 36s
The summary says the result contains one edge with use-count=4, so it is
not perfectly repaired, but it might still work in AngelCAD. I wrote a
few lines of code to make the hollow face, but dropped the cube as it
did not seem to intersect with the Face (?).
See attached image of result.
If you allow I will upload the AngelCAD code and the resulting STL (26.1
MB) to a github gist, so you can be viewed and you can download it. Is
that ok?
Regards
Carsten Arnholm
On 31.10.2020 15:13, jon wrote:
Thanks, everyone, for helping out.
The script and files are here:
https://github.com/arnholm/hollow_face
You can get the file in 2 ways
Method 1: Go to the page above and find the green button that says
"Code" and select the "Download Zip" to get all files, including the STL
and OBJ files.
Method 2: Clone the repository using git to get all files.
$ git clone https://github.com/arnholm/hollow_face
Carsten Arnholm