what is causing these yellow normal maps?

Travis C

It seems to happen mostly on meshes that are duplicated.

in maya  I've checked face normals, and they are pointing in the right direction. I've softened all edges and then hardened all the UV seams. 
I've frozen transformations and deleted history. 
All the meshes are triangulated.
I've also made a cage.
this was baked in xnormal.  Does anyone have a clue what is causing this? no amount of googling is bringing me any answers.

vertex_

Can we see your models? You might need to explode your mesh further.

Travis C

vertex_ said:

Can we see your models? You might need to explode your mesh further.

Here is a pic of just one of the baked problem mesh. Since the normal map of this still came out yellow, with no other objects in the scene, perhaps the problem is unrelated to exploding. 



cage:


hipoly:


baked in xNormal. Just this 1 part baked, and nothing else:


my xNormal settings: 


I've also attached this particular problem mesh as a .ma file as a .txt. in case you want to look at it. just change the extention from txt to .ma.  I'm super stumped. Normally this would seem like something easy to fix, but the underlying issue continues to evade me.

Travis C

ok i think i might have figured it out. hallelujah.

the problem is in the highpoly mesh and something funky going on when exporting to obj. When I checked the hipoly before exporting, it seemed all is well. then in a fresh maya scene i imported it and i see that some of the face normals were reversed. however even when i reversed it, the problem persists due to something weird going on that causes the mesh to turn black. so you have to mesh display-> reverse AND set to face to address that issue. 

So then that brings up another issue. how do I export without it sometimes reversing face  normals on me?

ActionDawg

Don't use obj honestly, go with fbx as it is far more predictable and interoperable with most every program. And use it always unless using certain programs that don't support fbx (lookin at you zbrush).

musashidan

somedoggy said:

Don't use obj honestly, go with fbx as it is far more predictable and interoperable with most every program. And use it always unless using certain programs that don't support fbx (lookin at you zbrush).

As of 4r7 ZB now supports. Fbx

ActionDawg

oh well that's awesome and much needed

polygunz

I believe your normals are inverted. Reset Xform and flip the normals if you are using Max.
Nowadays I use substance painter to bake without cage, much easier. Would give that a try.

musashidan

polygunz said:

I believe your normals are inverted. Reset Xform and flip the normals if you are using Max.
Nowadays I use substance painter to bake without cage, much easier. Would give that a try.

I would have thought inverted normals too, but OP states Maya and that the transforms are frozen and history deleted; which I believe is the same as resetting xform in Max.

And I second baking in SP.  

ActionDawg

I third baking in SD/SP, but would recommend using a cage for anything more complex than a plane.

musashidan

somedoggy said:

I third baking in SD/SP, but would recommend using a cage for anything more complex than a plane.

I've baked some complex (ish) meshes cage less in SP. Never really a problem as I usually bake synced.

Also, I've seen an Interesting  technique recently of setting up a graph in SD to blend averaged/non-averaged bakes to get the best of hard edges and non-skewed details from the respective maps.

ActionDawg

musashidan said:

somedoggy said:

I third baking in SD/SP, but would recommend using a cage for anything more complex than a plane.

I've baked some complex (ish) meshes cage less in SP. Never really a problem as I usually bake synced.

Also, I've seen an Interesting  technique recently of setting up a graph in SD to blend averaged/non-averaged bakes to get the best of hard edges and non-skewed details from the respective maps.

Sure, it's not totally necessary. Just speaking my opinion
The way I see it is that if I can get more control for the trivial amount of effort needed to use a cage, that's a huge plus for me.

I've seen the technique you mention. Honestly? I let skewing happen and I'm happy with the results. I don't waste time with control loops or blended bakes. My take is that if you need to do these things to control skewing then it's a sign your low poly should probably be adjusted to match your high poly model better. I feel like these techniques are mostly shown on "stress test" meshes that aren't very representative of real world work.

With that said, I've given some thought to designing a weighting scheme that would control skewing by making assumptions off surface curvature. It's based on the idea of the blended technique you mention. But it hasn't gotten further than a notepad doc on my messy desktop and I'm not even sure if it's feasible.