[Solved] Artifacts on bake

[houGenie]

Hi guys,

I have a little issue when normal baking with 3d Coat. There are some artifacts on the mesh which are only visible with some shiny shaders and with an certain angle.

1. Pic one shows low poly and the issue marked in red

2. Pic two shows the low poly mesh in wire mode

3. Pic three shows the high poly mesh

4. Pic four shows normal map on low poly mesh

5. Pic five shows normal map in knald

I have already tried all normal map software presets in the preferences, no difference.

My guess is that when importing the retopo mesh and checking the "lock normal" option, 3d coat ignores this option and calculates all normals as soft. This seems to be a bug. When everything is soft, then the normal map will look like in 4. and thus these artifacts. I have baked with different Software, see pic 5, which take the normals from the obj file and the artifacts are not there because the hard edges are not softened and therefore the normal map is flat in the regions which have the artifacts in 3d Coat.

I think the "lock normal" bug breaks the whole baking process for more complicated meshes, are the developers aware of this?

So how can I solve this problem in 3d Coat? Any one an idea? Thanks in advance.

[digman]

Is it possible to share that section of the model to test.

Question, you said importing for 3DC but you baked in Knald. Are you baking  a voxel / surface model model to the paint room or importing straight into the paint room using per pixel mode.

It appears you are baking in 3DC as well as I see the picture of the model in the sculpt room. Your wording is throwing me off.

Edited December 21, 2017 by digman

[houGenie]

24 minutes ago, digman said:

Is it possible to share that section of the model to test.

Question, you said importing for 3DC but you baked in Knald. Are you baking  a voxel / surface model model to the paint room or importing straight into the paint room using per pixel mode.

It appears you are baking in 3DC as well as I see the picture of the model in the sculpt room. Your wording is throwing me off.

The whole process is in 3d Coat. Load HP in sculpt room, load LP in retopo, bake in retopo to paint room --> artifacts. Knald was only an example that this issues is software dependent. I have inserted you a file. Hope it helps, thank you!

probe.3b

[Carlosan]

Can you share the LP mesh too to check "lock normal" option ?

Thx

[houGenie]

13 minutes ago, Carlosan said:

Can you share the LP mesh too to check "lock normal" option ?

Thx

Sure!

probe.obj

[digman]

Some of the problems could be related to the mesh.

You have texture stretching in the areas that are causing problems in the normal map. I re-worked the uv layout.

Also you have 18 n-gons in the mesh. Retopo Menu---Remove n-gons.  It will triangulate them.

Baked to the paint room with locked normals. First and last picture shows result.

Edited December 22, 2017 by digman
more info

[Michaelgdrs]

Handled like a Boss!!!!

[houGenie]

6 hours ago, digman said:

Some of the problems could be related to the mesh.

You have texture stretching in the areas that are causing problems in the normal map. I re-worked the uv layout.

Also you have 18 n-gons in the mesh. Retopo Menu---Remove n-gons.  It will triangulate them.

Baked to the paint room with locked normals. First and last picture shows result.

Thank you, I will look into this. Can you provide me your modifications with the modified 3d Coat file?

[houGenie]

I have tried to improve the UV map, but it is still visible (weaker). Not with the metal_dark material which you use, but for example with the material gold_rough.

[digman]

I found some spots that were not fitting the mesh well enough causing problems too. In picture.

You had a very long polygon, so I split it to soften the normals. Shown in picture.

Deleted some polygons. The corner ones that had 3 edges. Added back polygons and the corners now match how you did the rest of your corners. The other way was causing tension in the normals.   

Second picture shows the results.

I did my best with your current topology. Am I the most experienced modeler, no not by a long shot. I do know that if a mesh is not fitting close enough with a long polygon as was the case this time., the polygon is not really lying flat on the plane and might just be a non-planar polygon or at the most too much for the normal map to work with.  Look at the 3rd picture and you will see clearly.  The other thing is tension in the normals, too much tension and you get artifacts.  I call it tension, maybe another name others call it.

I have a a friend, a very experienced modeler. He even goes back after baking normal maps and goes I miss the tension in that spot and that spot. Corrects them and rebakes. 

In this case too, alot is asked of the normal map, as no edges are bevel. 3DC does a good job at creating the normal map in my opinion.

 

 

Edited December 22, 2017 by digman
more infor

[houGenie]

Thank you digman. I very appreciate your support.

You are of course right, that the topology is very important for this. The baking process is really a lot of fiddling around. For example non planar faces are really hard to fix, because if you correct one you break the other. What you call tension in normal is the same what I first called as the reason behind the artifacts. Because the normals are smoothed you have another tangent base in comparison to hard edge normals and the "tension" is bigger. That is the reason the normal map is really clean in knald with the same model, because it uses another tangent base because of the hardened normals (another option is to use support edges). That is the reason I wanted to use the "lock normals" option, which at the moment does nothing in 3d Coat

Yes, I do no bevels on the LP model, had them before but removed them because the result was good enough.

I have improved the topology like you advised and now it seems to work fine. At least the topology is now better. And the baking result is pleasant to look at. Again thank you very much, must be hard work to look for errors in other ppl models

Edited December 22, 2017 by houGenie

[houGenie]

How can I mark the thread as solved?