ModelExperimental - Default material

[ptrgags]

According to the glTF spec,

If material is undefined, then a default material MUST be used.

The "default material" is described here: https://www.khronos.org/registry/glTF/specs/2.0/glTF-2.0.html#default-material

This should be easy to handle, there's a couple things to check:

• Ensure that the MaterialPipelineStage doesn't throw an error if primitive.material is undefined. A no-op is probably okay
• Ensure that the shader defaults match what is described in the spec
• find a model to test with

[j9liu]

@ptrgags and I found that gltf-pipeline creates a default material, but we should make sure that for other things loaded as ModelExperimental that a default is created, and that the shader defaults are as expected.

[ptrgags]

I confirmed that GeoJsonLoader and PntsLoader do create a default material.

However, I think defaultModelMaterial() in ModelExperimentalFS.glsl isn't quite matching the spec.

According to the spec, default material should have:

• metallic: 1
• roughness: 1
• baseColor: [1, 1, 1, 1]

In other words it should be a metallic but rough material, whereas the shader currently defaults to a smooth non-metal.

[j9liu]

@ptrgags From a glance it doesn't seem like the material struct for ModelExperimental has a metallic factor, and that the other fields (f0, diffuseColor, roughness) are derived from the metallic factor in czm_pbrMetallicRoughnessMaterial. Should this be invoked with the default metallic roughness values in defaultModelMaterial() ?

EDIT: Actually, we can just hard-code the values based on how they would be computed in czm_pbrMetallicRoughnessMaterial, that would be simpler

[ptrgags]

I agree, hardcode the computed values. I think they'll all turn out to be straightforward constants like vec3(1.0) but better safe to check

[lilleyse]

Technically we can use whatever default we want in ModelExperimental but it's good to be consistent. I get these values:

czm_modelMaterial defaultModelMaterial()
{
    czm_modelMaterial material;
    material.diffuse = vec3(0.0);
    material.specular = vec3(1.0);
    material.roughness = 1.0;
    material.occlusion = 1.0;
    material.normalEC = vec3(0.0, 0.0, 1.0);
    material.emissive = vec3(0.0);
    material.alpha = 1.0;
    return material;
}

[jjhembd]

Hi @ptrgags, I'm looking at this and have a couple questions.

I confirmed that GeoJsonLoader and PntsLoader do create a default material.

Is there anywhere else I need to check to make sure we always get a default material?

Also, in PntsLoader, I see this:

  const metallicRoughness = new MetallicRoughness();
  metallicRoughness.metallicFactor = 0;
  metallicRoughness.roughnessFactor = 0.9;

  const material = new Material();
  material.metallicRoughness = metallicRoughness;

Based on the spec, do we need to set metallicFactor = 1.0 and roughnessFactor = 1.0?

This is my first time looking through the code, so let me know if I'm misunderstanding the context here.

[ptrgags]

Hi @jjhembd,

Is there anywhere else I need to check to make sure we always get a default material?

At this point I think it's really just ModelExperimentalFS.glsl that needs to be updated since that's hard-coding defaults in the shader.

As far as the ModelComponents.Material() goes, I think we covered all the main cases in the above discussion, but I guess it doesn't hurt to do a search through the code for ModelComponents.Material ModelComponents.MetallicRoughness and ModelComponents.SpecularGlossiness to double check nothing is hard-coding an unexpected default.

Also, in PntsLoader...

That's a good question... .pnts not technically a glTF file in the first place (unlike .b3dm and .i3dm which wrap a .glb file). So I don't think this particular case has a correct answer, it's more about what we think looks good as a default. Might be good to compare Model vs ModelExperimental rendering the same .pnts model and try to make them match.

[ptrgags]

Oh @jjhembd here's A modified Sandcastle example that sets up the default material. Basically I (ab)used a CustomShader's REPLACE_MATERIAL mode so the material is set to the default rather than what was in the glTF.

The relevant code is this part

// Modified this shader to use the default material
const textureUniformShader = new Cesium.CustomShader({
  lightingModel: Cesium.LightingModel.PBR,
  mode: Cesium.CustomShaderMode.REPLACE_MATERIAL,
  fragmentShaderText: `
    void fragmentMain(FragmentInput fsInput, inout czm_modelMaterial material)
    {
    }
    `,
});

EDIT: Also note that the lighting looks weird for the default material since the normal is set to a constant +z direction. This was mainly because we haven't implemented #6506