cleaver software mailing list

Text archives Help


Re: [Cleaver] Coarser tets along homogeneous boundaries


Chronological Thread 
  • From: Julien Finet <julien.finet@kitware.com>
  • To: Jonathan Bronson <bronson@sci.utah.edu>
  • Cc: "cleaver@sci.utah.edu" <cleaver@sci.utah.edu>, Ricardo Ortiz <ricardo.ortiz@kitware.com>
  • Subject: Re: [Cleaver] Coarser tets along homogeneous boundaries
  • Date: Fri, 28 Feb 2014 08:30:35 -0500

Hi Jonathan,

Thanks for your answer, we are looking forward to meet Cleaver2 :-)

Julien.


On Thu, Feb 27, 2014 at 4:48 PM, Jonathan Bronson <bronson@sci.utah.edu> wrote:
Hi Julien,

I'm glad it's working out well for you!  I'm also sorry about
the confusion with the behavior of tet sizes around those
smooth boundaries.

I agree the expectation as a user should be for coarser tets
along that smooth boundary. The are two reasons the current
release doesn't behave this way:

First,  the output mesh always refines to the voxel level along
material boundaries. For materials that go to the border this
isn't by itself a problem. But unfortunately the current release
also cannot handle material transitions on the boundary. So
the smooth plane that cuts across the border forces padding
to be used, which puts a 2-material interface around the
boundary of the domain. This leads to that refinement you see.

That deficiency is a side-effect of some early design decisions
and would be pretty tricky to fix. We welcome outside help on
this, but there are no plans to address it.

However, there is good news. A new adaptive version of Cleaver,
that we're calling Cleaver2  is in the final stages of being completed.
It not only handles material interfaces on the boundaries, but can
generate coarse tets along internal material interfaces as well. We're
very excited about the release and I hope to put out some information
on it soon.

Best,
    Jonathan


On Thu, 27 Feb 2014 11:01:24 -0500
 Julien Finet <julien.finet@kitware.com> wrote:
Hi,

We've done further tests with Cleaver (still loving it :-) ) and we have
noticed the following issue.

Feeding cleaver with a 30x30x30 cube with 2 label regions separated along a
plane, please see below the generated mesh (the tets (red and blue) are
intersected with a slice of the volume. Please disregard the odd looking
intersections :-), nevertheless we can observe:
- a) the generated tet mesh is coarse in the homogeneous areas
- b) the generated tet mesh is detailed along the jagged
green-cream labels boundaries.
- c) the generated tet mesh is detailed along the smooth air(black)-green
label boundaries.

I understand a) and b). I also understand that air is treated as a label (I
know Ricardo previously mentioned it).
However, as a naive user, I would expect to have a coarser mesh along the
smooth (!=jagged) label boundaries ? Would it be a hard feature to add into
Cleaver ?

Thanks for your insights,
Julien.

[image: Inline image 1]





Archive powered by MHonArc 2.6.18.

Top of page