cleaver software mailing list

Text archives Help


Re: [Cleaver] Cleaver fails on 1px wide materials


Chronological Thread 
  • From: "Jonathan Bronson" <bronson@sci.utah.edu>
  • To: Joshua Levine <levinej@clemson.edu>,Julien Finet <julien.finet@kitware.com>
  • Cc: "cleaver@sci.utah.edu" <cleaver@sci.utah.edu>
  • Subject: Re: [Cleaver] Cleaver fails on 1px wide materials
  • Date: Mon, 17 Mar 2014 20:29:38 -0600

Hi,

I apologize for the confusion. Let me see if I can clarify.

First, if you're using NRRD as input files, Cleaver will
use the format specified in the file. If it calls for node
centered, cleaver interprets it as node centered. Same for
cell centered. The default setting in the class is only for
Cleaver created data, data that doesn't specify, or for user
requested interpretation. The default choice was made because
of our frequent use of image data.

As for the loss of 1-voxel thick features. A while back, this
was expected behavior with no known solution. The reason behind
it is pretty technical, but it's related to the way we populate
the background lattice and the way we ensure cuts are consistent
across dual vs. primal edges.

With that said, it's been a good while since this particular issue
has been looked into. A lot of the code has been changed since then,
including a new volume interpolation abstraction. So there might be
a workaround now. I'll need to look carefully through the code to know
any more.

Best,
Jonathan



On Mon, 17 Mar 2014 21:31:35 -0400
Joshua Levine <levinej@clemson.edu> wrote:
To follow up on this, Julien and I did some debugging and it has to do
with the initial CellCentered vs. NodeCentered modes for the
FloatField type. I was actually a bit surprised that the default was
CellCentered.

Jonathan -- do you have any feedback here. The net result was that n
Julien's data, the interpolation completely loses the 1-voxel thick
material. I would have thought (erroneously) that in NodeCentered
type, actually the dual mesh could have been populated with the input
data values, instead of interpolating the once-interpolated values.

Josh

On Mon, Mar 17, 2014 at 2:04 PM, Julien Finet <julien.finet@kitware.com> wrote:
Hi,

As discussed, hre [1] you will find data that fails with Cleaver. There are
3 materials, 1pixel wide cylinder inside a 5pixel wide cylinder. The 1 pixel
wide cylinder does not show up in the final tetmesh.

Thanks,
Julien.
[1]: http://packages.kitware.com/folder/258




Archive powered by MHonArc 2.6.18.

Top of page