Bug #1034

Malassigned / malformed vertices in the armpit

Added by Joel Palmius over 2 years ago. Updated almost 2 years ago.

Status:AcceptedStart date:06/01/2016
Priority:NormalDue date:
Assignee:Joel Palmius% Done:

0%

Category:Modelling
Target version:MakeHuman 1.1.2

Description

From http://www.makehumancommunity.org/forum/viewtopic.php?p=34333

In some poses vertices in the armpits very visibly ends up in the wrong place, even for bundled clothing.

It is unclear whether this is only for makeclothes or if there is a problem with the rigging / weight painting.

arm3.png (237 KB) Joel Palmius, 06/01/2016 08:31 AM

screenshot_1_1464842286.png (129 KB) Rob Baer, 06/02/2016 06:39 AM

screenshot_1_1465052852.png (513 KB) Rob Baer, 06/04/2016 05:08 PM

1254
1255
1256

Related issues

Related to MakeHuman - Bug #1036: MakeClothes Average Child Vertex Groups Fixed 06/08/2016

History

#1 Updated by Joel Palmius over 2 years ago

#2 Updated by Joel Palmius over 2 years ago

  • Description updated (diff)

#3 Updated by Rob Baer over 2 years ago

It also happens with some topologies (male generic, for example) - but not so clearly with others. With male generic it can be seen in t-pose as well as fly.

#4 Updated by Rob Baer over 2 years ago

  • File screenshot_1_1465052852.png added
  • Category changed from MakeClothes to Modelling
  • Assignee changed from Thomas Larsson to Anonymous

tested r2065 stable
Although this could be a pose-skeleton weighting issue, some further testing suggests that some clothes and topologies simply need to have some smoother edge-loops or updated geometry in the armpit area. Here's what I see with the benchmark pose when I look at various items including th base mesh and helper geometry.

I think this is more in Manuel's wheelhouse than Thomas's

#5 Updated by Aranuvir # over 2 years ago

Is this related to the margin issue (#1036)?

#6 Updated by Rob Baer over 2 years ago

  • Related to Bug #1036: MakeClothes Average Child Vertex Groups added

#7 Updated by Rob Baer over 2 years ago

Well, yes and no. They are, strictly speaking, not directly related, but I do think both are probably a function of MakeClothes vertex group assignment. I added the cross-reference anyway.

The issue here (#1034) is about vertex group assignment on a final asset. I imagine this problem arises at a MakeClothes level and not a rig weighting level, but this remains to be proven. Manuel probably used the automatic vertex assignment when making the offending assets and didn't realize the assignment of a few errant vertices in the underarm had jumped. Only proxies or clothes where such automatic assignment errors were not corrected by hand will be affected.

Issue #1036 is potentially more serious because it is the collection of "human gold standards" that MakeClothes uses to master all proxy group vertex mapping that are in error. Here it's the mater templates that are flawed. All proxy assets based on these human reference standards will be affected in the crotch area. The vertex groups for baby and average child and the helpers for the average child (and perhaps other helpers) need editing.

#8 Updated by Joel Palmius about 2 years ago

  • Assignee changed from Anonymous to Joel Palmius
  • Target version changed from MakeHuman 1.1.1 to MakeHuman 1.1.2

#9 Updated by Joel Palmius about 2 years ago

  • Target version changed from MakeHuman 1.1.2 to MakeHuman 1.1.1

Changing my mind. I'll post on the forums and see if someone is willing to take a look at this for 1.1.1.

#10 Updated by Jonas Hauquier about 2 years ago

  • Status changed from New to Accepted

#11 Updated by Jonas Hauquier about 2 years ago

Making proxies compatible with different poses is a tough issue.
Perhaps one that can be improved with proper tools. Some idea I had a while back, is to make MakeClothes (or some plugin within MH) cycle through a set of benchmark poses and even allow in-line modifications to the fitting to fix them.
One could also test whether the new MakeClothes algorithm actually fixes most of these issues. The fitting for this new algorithm is actually already implemented in MH by me but it doesn't work properly yet, and still needs some debugging.

#12 Updated by Joel Palmius almost 2 years ago

  • Target version changed from MakeHuman 1.1.1 to MakeHuman 1.1.2

Not happening for 1.1.1.

Also available in: Atom