I was trying this, thinking it might be a good way to set up poly groups.
But in practice, it seems way too inaccurate to be useful (see pic).
Can someone explain what I’m missing?
Also both ‘polygroups from polypaint’ and ‘groupsloops’ seem to ignore
any symmetry settings!?
G.
Edit: OK It helps if you turn symmetry on :o.
I found the same problem when I tried it.
Hi, which brush do you use for painting? If I remember correctly they said in the R3 tutorial video that its the best to use the layer brushes. Ive tested it and it works fine for me with the LayerConst brush and RGB set to 100.
Hope that helps
Works fine for me too. Here some things that could help:
Next to the “from polypaint” button is a tolerance slider. Play with that one a little.
You have to make use of the “group loops” function to add smoothed buffers in between the polygroups, to get good results when manually painting polygroups. It will recalculate the geometry and smooth them out. Increase the polish value for more smoothing.
If you dont want to make make use of that feature, you are better manually defining the polygroups by hiding all but the exact portion you want grouped with the lasso selection tool, and hitting “group visible” in the polygroup menu.
Well. I was finally able to watch that video!
Agreed the use of the single layer brush helps. You can also adjust the
alpha curve to make the standard brush do the job better. But I still
can’t get accurate results.
And if you convert polygroups to polypaint, then back again. The result is
a real mess. I was so hoping I could now save polygroups to a texture
and reload them later.
I guess it’s not to be.
G.
Heres a couple tips tips I discovered.
and most importantly:
As long as I hit “group visible” first, then convert paint to group, I get pretty clean, accurate results with a tolerance value of about .7.
Thanks for the suggestions Spyndel. I had already figured out point (2)
from my experiments.
The problem here seems to be that ‘polypainting’ is essentially vertex painting,
while polygroups are effectively defined in terms of faces. A single vertex
can (and often does) belong to mutiple polygroups. I can see that deriving
groups from information stored solely in the vertices is a tricky problem.
G.
OK.
I think I’ve found the way to do the low-res polygroup export. (I’ll experiment
some more later - day job requires attention 
The trick (obvious really, I’d just missed it) is to divide the low res mesh
a couple of times (with SMT off) before doing the polygroups-to-polypaint.
Then a reverse polypaint to polygroups operation works a lot better (a
greater number of vertices to vote on the face assignment?). You
can then discard the higher res as appropriate.
[Edit: and as Spyndel says, a 0.7 tolerance looks about right]
G.
OK. So I figure I’ll back off and play with the tutorial video techniques
on polygroup and polypaint.
Only I pick a slightly different subject as you can see 
A quick sketch, a bit of ZSketching. Clean up the resulting mesh a bit.
I’m careful to keep X symmetry on when I do the remesh, and keep it on
for all editing and painting.
I use the layer brush, 100% RGB and focal shift zero.
Before I hit the polygroup from polypaint button, I run Smart Resym to
check everything is 100% sym’. It is.
Once again it goes wrong…See attached.
I can repaint over the problem area as many times as I like. It wont fix
the problem. I’ve also reset the PG with ‘group visible’ and tried again.
Same result.
I’ve also checked for stray masking problems. Nope.
Needless to say attempting ‘group loops’ at this point is pointless.
It just seems buggy.
G
I encountered a similar error. It resulted from me using the traditional “group visible” function to make my poly groups. Oddly, clearing the grouping and using the painting method fixed it for me.
Polygroups have always been subject to random weirdness. I would suggest the new Mirror Weld functions make it easier to deal with them by simply worrying about one “good” side of the model, then mirror the sculpting from one side to the other, like you would in a traditional modeling app.
Also, I have also documented some of the same sort of issues with polygroups in Zbrush in this thread:
http://www.zbrushcentral.com/showthread.php?t=76370&highlight=errors
One fairly reliable way I’ve found to “clean” a problematic mesh of an error thats bothering me, is to subdivide it , and delete the lower level of subdivision. Then, use the “Reconstruct Subdivision” command to reconstruct the mesh that was just deleted. Often, the reconstructed mesh will be “scrubbed” of the error that was troubling me.
Thanks Spyndel.
I finally got the mesh above polygrouped correctly by messing with the
tolerance slider. Success! My only excuse for not trying that before was
that I was working late last night, and completely forgot about it.
But I tried the subdiv-reconstruct trick as well. In this case it actually
seemed to make the problem worse (the area of dodgy polygroup got
larger).
Of course the fact that it had any effect at all is ‘interesting’.
G.
IMPORTANT NOTE ABOUT POLYGORUPS FROM POLYPAINT OR MASKING.
Polygroups from paint or masking are additive, it does not destroy your exiting polygroups, it adds to it.
This allows you to add partial polygroups without losing the existing polygroups assignmets.
However, when you want to start ‘fresh’ and overwrite the existing groups, you may…
option one: Press “GroupVisible” followed by ‘From PolyPaint’ or ‘From Masking’ .
or
option two: Option+Click ‘From PolyPaint’ or ‘From Masking’.