Author Topic: Newbie having trouble  (Read 580 times)

Update: 17 April 2017
After running more tests, the problem with the sample cube no longer displays any errors that I can see.
I'll chock it up to a user malfunction.

I would like to go ahead and discuss the actual project that I'm working on.
The company that I work for is developing specifications for GLTF 2.0 along with the KRONOS Group.
For those who are unaware of the format, here is some info.
https://www.khronos.org/gltf

As part of that effort, We've got a whole host of models that will need to be converted to utilize PBR materials.
Most of our models are of mechanical objects and have many individual mesh layers or parts.

Right now we're using the OBJ format and will be computing our own tangents and Bi-tangents from the OBJ.

The first model I chose was of an MRAP Cougar ground vehicle.
I have a high-res version containing over 448,000 polys.
My low-res version has about 12,000.
The model was created in pieces, fenders, main body, windows, wheels etc.

After having gone back to the excellent tutorials that Wes produced, I've structured the two files so that each has its respective _high and _low suffixes per mesh item. Each model has 25 individual mesh items.
I gave each model a default material and uv'd the low-res version.
I've used different UV'ing methods, unwrap and atlas with a lot of island sewing, but not sure which is the best to use.

I know that I need to smooth the normals of the low-res mesh, by either splitting edges or by padding the uv's or edge beveling to accommodate for the smooth surfaces of the high res mesh.

I've tried all these methods without really much success and am really getting frustrated at this point.

Not all of the edges of the low-res MRAP model need to be smoothed, as in the case of the cube.
So I started with the front fender.
I'm selecting the edges that need to be smoothed and then use the "Vertex Toolkit" written by James O'Hare.

A "Vertex Normal" map gets created, but for some reason, all of the edges are being smoothed.

To make matters more confusing, once brought into Substance Painter and an attempt to generate a "Normal" map is made, the edges don't look correct. I understand that this requires a UV split with padding and or an edge split, but even after making these corrections, the results don't clear up.

Randall.
Last Edit: April 18, 2017, 12:31:43 am

Update: 17 April 2017
After running more tests, the problem with the sample cube no longer displays any errors that I can see.
I'll chock it up to a user malfunction.

I would like to go ahead and discuss the actual project that I'm working on.
The company that I work for is developing specifications for GLTF 2.0 along with the KRONOS Group.
For those who are unaware of the format, here is some info.
https://www.khronos.org/gltf

As part of that effort, We've got a whole host of models that will need to be converted to utilize PBR materials.
Most of our models are of mechanical objects and have many individual mesh layers or parts.

Right now we're using the OBJ format and will be computing our own tangents and Bi-tangents from the OBJ.

The first model I chose was of an MRAP Cougar ground vehicle.
I have a high-res version containing over 448,000 polys.
My low-res version has about 12,000.
The model was created in pieces, fenders, main body, windows, wheels etc.

After having gone back to the excellent tutorials that Wes produced, I've structured the two files so that each has its respective _high and _low suffixes per mesh item. Each model has 25 individual mesh items.
I gave each model a default material and uv'd the low-res version.
I've used different UV'ing methods, unwrap and atlas with a lot of island sewing, but not sure which is the best to use.

I know that I need to smooth the normals of the low-res mesh, by either splitting edges or by padding the uv's or edge beveling to accommodate for the smooth surfaces of the high res mesh.

I've tried all these methods without really much success and am really getting frustrated at this point.

Not all of the edges of the low-res MRAP model need to be smoothed, as in the case of the cube.
So I started with the front fender.
I'm selecting the edges that need to be smoothed and then use the "Vertex Toolkit" written by James O'Hare.

A "Vertex Normal" map gets created, but for some reason, all of the edges are being smoothed.

To make matters more confusing, once brought into Substance Painter and an attempt to generate a "Normal" map is made, the edges don't look correct. I understand that this requires a UV split with padding and or an edge split, but even after making these corrections, the results don't clear up.

Randall.

Hi Randall,

Thanks for sharing your project : ) Your update indicates that you got the bake working correctly. Is it working now?

Cheers,
wes

Integrations Product Manager / Training
wes.mcdermott@allegorithmic.com
Twitter: The3DNinja

Hi Wes.

I did get the issue that I was having with the default cubes resolved by adding supporting edge loops to the low poly model.
But I don't think that I couldn't started off with a more complicated model than the MRAP and have bitten off more than I can chew.

I'm unclear about a number of things, first of which is choosing which method I should use the generate the normal map.
We're not baking for Unity or Unreal.
I do know that we're using OBJ files and that we are going to be deriving the tangents and bi-tangents via the OBJ and that we're using OpenGL.

Can I use a single smoothing group?
Do I need to put in supporting edge loops everywhere, or just split the edge of the low poly model where I need to support the smooth shading of the high poly model?
Can I use an Atlas map or must I Unwrap everything?
Should I use the "Compute tangent space per fragment in SP

I've tried just about every combination that I can think of.

Thanks

Randall




 

Export your textures and test them in toolbag or anything else. Substance painter (as i know) does not interpolate normals on UV islands correctly so you get the visible seam.

staughost,

The model setup is being done in modo 10.2 and all the texturing is being generated in SP.
I've seen visible seams, but just thought that it was the cause of my poor workflow.


Export your textures and test them in toolbag or anything else. Substance painter (as i know) does not interpolate normals on UV islands correctly so you get the visible seam.

HI,

Are you talking about an issue in the viewport or a tangent mismatch? I'm not aware of an interpolation issue that causes seams.

Cheers,
wes
Integrations Product Manager / Training
wes.mcdermott@allegorithmic.com
Twitter: The3DNinja

staughost,

The model setup is being done in modo 10.2 and all the texturing is being generated in SP.
I've seen visible seams, but just thought that it was the cause of my poor workflow.

Export your textures and test them in toolbag or anything else. Substance painter (as i know) does not interpolate normals on UV islands correctly so you get the visible seam.

HI,

Are you talking about an issue in the viewport or a tangent mismatch? I'm not aware of an interpolation issue that causes seams.

Cheers,
wes


Issue in viewport. From what I understand the problem with sp creating seams on uv islands is due to fast interpolation (might bee that I am using wrong words) and/or it is "The seam you see is related to the way we sample the environment map to create specular reflections.You can try to increase the sample quality in the shader parameters to reduce the artifact, but it will always be present at some point. However it shouldn't affect other applications." - Froyok.

Cheers :)

staughost,

I'm not sure of what you're referring to.
If you think there's a problem, file a bug report.

Wes, do you have any suggestions or recommendations regarding my questions?
Last Edit: April 19, 2017, 04:37:13 pm

staughost,

I'm not sure of what you're referring to.
If you think there's a problem, file a bug report.

Wes, do you have any suggestions or recommendations regarding my questions?

HI,

Can you send me one of the mesh layers the low and high version? The one you show in the screen shots and I will take a look. You don't need to smooth all edges.

The general rule I follow is that anytime you have a hard edge (smoothing split), then you need to have a corresponding UV split with padding or you will get a seam. It sounds like you are following this as well.

In the example you have above, you may need to add support edges to the low res mesh. You have a hard edge right at a curved edge for the high res mesh. The low res edge might be too sharp and causing this seam.

Cheers,
Wes
Last Edit: April 20, 2017, 07:55:02 pm
Integrations Product Manager / Training
wes.mcdermott@allegorithmic.com
Twitter: The3DNinja

Wes,

Previously I was just smoothing everything and then separated the edges of the uv's where I wanted support for the smoothing of the high poly mesh.

Since then, I've added more supporting geometry, mainly for the headlight.

I've also spoken with the gentleman who wrote the "Vertex Normal Toolkit" and subsequently have modified how I'm generating the normals.

His tool has a function that will "Harden the edge boundaries of the currently select UV map and soften all other edges"

I'm still having a bit of trouble with my original edges that I need to shoulder the smoothing of the high-res mesh.
I think mainly that's because I have those edges on the boundary of the uv island, which is getting hardened by the previously mentioned tool. So now I'm seeing visible edges in SP.

I've attached the requested files.

Thanks
Randall

Wes,

Previously I was just smoothing everything and then separated the edges of the uv's where I wanted support for the smoothing of the high poly mesh.

Since then, I've added more supporting geometry, mainly for the headlight.

I've also spoken with the gentleman who wrote the "Vertex Normal Toolkit" and subsequently have modified how I'm generating the normals.

His tool has a function that will "Harden the edge boundaries of the currently select UV map and soften all other edges"

I'm still having a bit of trouble with my original edges that I need to shoulder the smoothing of the high-res mesh.
I think mainly that's because I have those edges on the boundary of the uv island, which is getting hardened by the previously mentioned tool. So now I'm seeing visible edges in SP.

I've attached the requested files.

Thanks
Randall

Hi Randall,

I think you are doing things correctly. I recommend using FBX export from modo so you can include the smoothing groups. With your UVs, the packing was pretty tight. You need to make sure you have enough of a gap between the UV shells when you have hard edges. You need to split the UVs at the hardedges and make sure that is some padding between them. That will help to produce a cleaner edge.

Another issue that is coming into play is that the texel density for the UVs might be a little low for the areas you see the dark seam line. If the texel density is lower than you get too much linear interpolation from the shader which is why you are seeing this dark edge seam. You can either scale the UVs to take up more of the 0-1 space and thus use more pixels of the texture or you can bake higher res map. If I bake at higher resolution such as 4K with the current UV layout and set the texture set to 4K, I get a cleaner result.

In the second image, I scaled the UVs to be larger and also baked at 4k and get a better result. You can't always do this though as larger texture sizes and filling UVs to the entire 0-1 is not practical. You just need to balance it as much as you can.

 
Cheers,
wes
Last Edit: April 27, 2017, 11:11:36 pm
Integrations Product Manager / Training
wes.mcdermott@allegorithmic.com
Twitter: The3DNinja