← All ideas
  • Published 15 Mar 2019

Better Uwrapping Algorithm

Better Uwrapping For Blender .

BetterUVAlghorithm.jpg

78.6%

Development Status: Incomplete

Upvote 11 Downvote 3
giakaama
15 Mar 2019

12 Comments

jason

"Make it better" is not an actionable proposal. Describe specifically what is wrong with Blender's default unwrap versus how it should be. Even better, provide links to whitepapers or research papers that show promising new unwrap algorithms that you feel the Blender developers should implement.

7
giakaama Author

As you can see in the Picture, Maya has a better algorithm than blender's one .

2
Alex-D

Sorry for double post but for some reason I can't edit my initial answer:

I already proposed a similar thing a few months ago: Link

There was a summer of code project from Aurel Gruber in 2016 - the SLIM algorithm. He stopped working on it, though and I've heared it's been in a buggy state ever since.

Personally I's also like a few more relax algorithms, as well as the option to iterate manually and over selections of island.

Select UV-Island or polygons -> click [relax] -> select algorithm used as well as iterations. Repeat for as many islands and polys or UDIMS or arbitrary selections as needed.

1
Alex-D
1
giakaama Author

I think we need a UV Room Code Quest :) or UV Improvement Coding Week Fund :D .

Alex-D

@giakaama

RizomUV have the forumla down pretty good, I think: Here's the trailer from the 2018 Release.

Notice those features:

  • individual UV islands can be relaxed iteratively and by diffrent methods
  • lines can be straightened resp. squares can be aligned into a box shape.
  • UV Islands can be combined and sligned on the UV Space
  • UDIM workflow
  • an iterative packing algorithm like UVPackmaster
  • A toolset like the super awesome TexTools in a 2.8 version into Blender as a default package.

The algorithms for packing and relaxing individual UV Islands are the most crucial, in my opinion. I am sure TexTools will follow for 2.8 eventually but as fantastic as they are they can't solve missing relax and pack algorithms. Also baking very polyheavy models in 2.79 was still so slow that I tried it only once and then went back to Substance for the time being.

Can the two proposals be merged somehow? I think it's really counterproductive to have it split into two threads.

giakaama Author

So, I come back to this proposition again . I didn't say " make it better", I just pointed out A and B testing between Blender and Maya, in the hope that the devs know of what can improve in Blender's unwrapping/relaxing algorithm.

But it seems that I needed to give more informations about. Don't have more info but @motionblur.of.doom He pointed out more informations in his post on how to make the UV Unwrapping better .

Check it out

I don't know what kind of algorithm Maya uses, but in my oppinion the sugestions that @motionblur.of.doom tells in his post are great .

Maybe the devs should look at those insights, having those inside blender can lead to a great start in unwrapping . We won't have to rely on other software to do unwrapping.

Thanks

giakaama Author

All replies here are good. The thing is that I don't know what kind of algorithm Maya uses. I think they use the unfold from RizomLabs, I think. I think Blender is using LSCM ... The guy from 3D Coat was trying to improve the unfold algorithm in 3DCoat and stumbled across a new one GU ... I don't have any in depth info about that either.

What I was trying to say here is just maybe the devs could try to understand why maya does it better or other application . I really think they'll know where to look and figure out how Maya's unfold works in the backgroud .

I know it's hard work and I don't know how hard it is. But We're just trying to get Blender better and not switch to 3rd party UV unwrapping applications .

camplowell

Any information on the algorithm Blender uses vs. the algorithm Maya uses? The proposal would be significantly more actionable if you provide a specific unwrapping algorithm you would like to see in Blender.

Alex-D

* @camplowell * How standardized are relax algorithms, anyways? Personally I found that Modo's 'adaptive relax' in conjunction with the standards ABF+/LSCM and occasional frontal projection from viewport can unwrap mostly everything. Paired with a 'make square' algorithm for long quad lines is basically my current ideal workflow. Only in very small edge creases (doubled edges for SubDivs) I've encountered really annoying plroblems. I haven't had a achance to really use Rizom or Hudini, yet.

Vyach

Did you try different modes?

Lauloque

yeah make it better ! dunno how ? your problem ! maya does it !