Bug #1036

MakeClothes Average Child Vertex Groups

Added by jwc (Forum User) over 2 years ago. Updated over 2 years ago.

Status:FixedStart date:06/08/2016
Priority:NormalDue date:
Assignee:Thomas Larsson% Done:

0%

Category:MakeClothes
Target version:-

Description

Versions: MakeHuman 1.1.0, MakeClothes 1.1.0, OSX

Average Child has vertex 4258 (on the right side) included in Left Vertex Group and vertex 10890 (on the left side) included in Right Vertex Group

Can reproduce by:
Open Blender, Open MakeClothes
Set Type to "Average Child" and click Load Human Mesh
Select Human, Tab in edit mode
Select Left Vertex Group and examine highlighted mesh.

These two vertices are very close to the midline near bottom of the coccyx, on either side of index 4259.

Both of these vertices are also included the Mid Vertex Group, which is questionable.

screenshot_1_1465435710.png (224 KB) Rob Baer, 06/09/2016 03:29 AM

screenshot_1_1465437780.png (185 KB) Rob Baer, 06/09/2016 04:04 AM

1257
1258

Related issues

Related to MakeHuman - Bug #1034: Malassigned / malformed vertices in the armpit Accepted 06/01/2016

History

#1 Updated by Rob Baer over 2 years ago

Indeed there seems to be a small issue here. The two vetices seem transposed.

#2 Updated by Rob Baer over 2 years ago

It seems the average baby has 2 vertices on each side in about the same region that are similarly misassigned in left and right groups.

#3 Updated by jwc (Forum User) over 2 years ago

I'm not sure how extensive this is, it doesn't seem to appear on the "Average Female", but does appear with average baby.

The "Average Child with Helpers" mesh does not have the vertices on the wrong side. Instead it only includes these two vertices in the Mid Vertex Group, not in the Left or Right Groups.

I suspect it is the reason I am unable to create a smooth piece of clothing that works with the different poses on a young female mesh.

#4 Updated by Rob Baer over 2 years ago

Until this is fixed directly in MakeClothes you can probably just manually edit these errant vertices to belong to the right vertex groups. This should smooth your clothing fit.

Yes, base, average female, and average male seemed to have everything properly assigned as far as I can tell. I did not actually look carefully at the helper assignments in these regions yet. Hopefully they are ok. But using the helpers during clothes fitting will often lead to smoother/better fits as well, depending what you are trying to do.

#5 Updated by jwc (Forum User) over 2 years ago

Manually correcting the vertex assignment creates an article of clothing that properly follows the body shape for the different poses I tried.
Thank you, this solves the problem for me.

I looked at the tights helper for the average child.
It also has these vertices assigned to all three vertex groups, so some of the helpers have this issue.

If you wish, I could examine the 10 body meshes and 11(*5) helper meshes for all three vertex groups and send you a list with all of the discrepancies.
I don't know how to export vertex groups from Blender, otherwise I would make the changes and send them to you for review.

#6 Updated by Thomas Larsson over 2 years ago

MC assigns vertices on the wrong side to the Left and Right vertex group provided that they are very close to the middle. This slack was originally introduce to compensate for modelling errors, I think, but if the targets are perfectly symmetrical it could be removed.

The margin is hardcoded in lines 1654 and 1656 of makeclothes.py. If you replace 0.01 with 0.0001 there is no misassignment in the child.

#7 Updated by Aranuvir # over 2 years ago

Thomas Larsson wrote:

MC assigns vertices on the wrong side to the Left and Right vertex group provided that they are very close to the middle. This slack was originally introduce to compensate for modelling errors, I think, but if the targets are perfectly symmetrical it could be removed.

The margin is hardcoded in lines 1654 and 1656 of makeclothes.py. If you replace 0.01 with 0.0001 there is no misassignment in the child.

This makes sense. But shouldn't the underlying basemesh being assumed to be symmetrical, except it was connected to Joel's sync-tool? Perhaps we could have an option on the blender's properties panel for changing the margin.

#8 Updated by Rob Baer over 2 years ago

  • Related to Bug #1034: Malassigned / malformed vertices in the armpit added

#9 Updated by Thomas Larsson over 2 years ago

Aranuvir (Forum User) wrote:

This makes sense. But shouldn't the underlying basemesh being assumed to be symmetrical, except it was connected to Joel's sync-tool? Perhaps we could have an option on the blender's properties panel for changing the margin.

The basemesh has always been symmetric, I think. However, some of the macrotargets were not. This has probably been fixed by now, but when the code was written not all of them were symmetric. Anyway, assuming that all targets (only the four ones used by MC) are perfectly symmetric it should be safe to set the margin to something very small. Not exactly zero, because I'm not sure if Blender introduces rounding errors, but 1e-4 corresponds to 10 um which is about 1/10 of the width of a hair. Should be small enough.

#10 Updated by Aranuvir # over 2 years ago

  • Status changed from New to Fixed

Since there is a manual workaround a propose to close this task.

Also available in: Atom