Texture baker updates

[kwokcb]

Issue

• Writing documents separately means that each documents elements are uniquely named but when combined it is possible to get clashes. As there is no API to merge and resolve duplicates, a new option is added to just create a single document and use the built in texture baking element uniqueness resolve things. This also means that texture instances that used by more than one shader are not baked many times (e.g. chess set).
• Update MaterialXView: baking can miss image size detection when using include files #1139, Texture baking output access for multiple material documents #1140

Changes

• Add in option to write output to single document: writeDocumentPerMaterial

  • Add texturebaker utility option
  • Add translateshader utility option

• Update MTLXView to have a minimum bake size set to 1K to match texturebaker and allow for single document option:

Results

• Example translating shader to N (default)

python translateshader.py resources\Materials\Examples\StandardSurface\standard_surface_look_brass_tiled.mtlx x.mtlx gltf_pbr --writeDocumentPerMaterial True
Processing material: Tiled_Brass
Wrote baked image: Tiled_Brass_gltf_pbr_base_color.png
Wrote baked image: Tiled_Brass_gltf_pbr_roughness.png
Wrote baked image: Tiled_Brass_gltf_pbr_clearcoat.png
Wrote baked image: Tiled_Brass_gltf_pbr_clearcoat_roughness.png

Processing material: Greysphere_Calibration
Wrote baked document: x_SR_brass1.mtlx
Wrote baked document: x_SR_Greysphere_Calibration.mtlx

vs 1 document:

python translateshader.py resources\Materials\Examples\StandardSurface\standard_surface_look_brass_tiled.mtlx single_doc.mtlx gltf_pbr --writeDocumentPerMaterial False
Processing material: Tiled_Brass
Wrote baked image: Tiled_Brass_gltf_pbr_base_color.png
Wrote baked image: Tiled_Brass_gltf_pbr_roughness.png
Wrote baked image: Tiled_Brass_gltf_pbr_clearcoat.png
Wrote baked image: Tiled_Brass_gltf_pbr_clearcoat_roughness.png

Processing material: Greysphere_Calibration
Wrote baked document: single_doc.mtlx

with the following result:

<?xml version="1.0"?>
<materialx version="1.38">
  <nodegraph name="NG_baked" colorspace="srgb_texture">
    <image name="base_color_baked" type="color3">
      <input name="file" type="filename" value="Tiled_Brass_gltf_pbr_base_color.png" />
    </image>
    <output name="base_color_output" type="color3" nodename="base_color_baked" />
    <image name="roughness_baked" type="float">
      <input name="file" type="filename" value="Tiled_Brass_gltf_pbr_roughness.png" />
    </image>
    <output name="roughness_output" type="float" nodename="roughness_baked" />
    <image name="clearcoat_baked" type="float">
      <input name="file" type="filename" value="Tiled_Brass_gltf_pbr_clearcoat.png" />
    </image>
    <output name="clearcoat_output" type="float" nodename="clearcoat_baked" />
    <image name="clearcoat_roughness_baked" type="float">
      <input name="file" type="filename" value="Tiled_Brass_gltf_pbr_clearcoat_roughness.png" />
    </image>
    <output name="clearcoat_roughness_output" type="float" nodename="clearcoat_roughness_baked" />
  </nodegraph>
  <gltf_pbr name="SR_brass1_baked" type="surfaceshader">
    <input name="base_color" type="color3" output="base_color_output" nodegraph="NG_baked" />
    <input name="roughness" type="float" output="roughness_output" nodegraph="NG_baked" />
    <input name="clearcoat" type="float" output="clearcoat_output" nodegraph="NG_baked" />
    <input name="clearcoat_roughness" type="float" output="clearcoat_roughness_output" nodegraph="NG_baked" />
  </gltf_pbr>
  <surfacematerial name="Tiled_Brass_baked" type="material">
    <input name="surfaceshader" type="surfaceshader" nodename="SR_brass1_baked" />
  </surfacematerial>
  <gltf_pbr name="SR_Greysphere_Calibration_baked" type="surfaceshader">
    <input name="base_color" type="color3" value="0.478431, 0.47451, 0.47451" colorspace="srgb_texture" />
    <input name="metallic" type="float" value="0" />
    <input name="roughness" type="float" value="0.698039" />
    <input name="clearcoat_roughness" type="float" value="0.0980392" />
  </gltf_pbr>
  <surfacematerial name="Greysphere_Calibration_baked" type="material">
    <input name="surfaceshader" type="surfaceshader" nodename="SR_Greysphere_Calibration_baked" />
  </surfacematerial>
</materialx>

• Note that an invalid arg will now throw an exception error: argument --writeDocumentPerMaterial: Boolean value expected.
  which is performed in a new stringToBoolean() utility added to MaterialX datatypes.py

• Chess set bake result:
  standard_surface_chess_set_baked.mtlx.txt

[jstone-lucasfilm]

This looks really promising, thanks @kwokcb.

I would recommend that we harmonize on a single name for the flag that controls this new behavior, rather than inverting the name between our Python scripts and C++ classes (e.g. "writeSingleDocument" vs "writeSeparateDocuments").

One suggestion for a unified naming convention would be "writeDocumentPerMaterial", which seems clear enough for new readers to follow. We would then provide this same flag name in both Python scripts and C++ classes, and follow your existing convention of defaulting its value to "true" for backwards compatibility.

[kwokcb]

This looks really promising, thanks @kwokcb.

I would recommend that we harmonize on a single name for the flag that controls this new behavior, rather than inverting the name between our Python scripts and C++ classes (e.g. "writeSingleDocument" vs "writeSeparateDocuments").

One suggestion for a unified naming convention would be "writeDocumentPerMaterial", which seems clear enough for new readers to follow. We would then provide this same flag name in both Python scripts and C++ classes, and follow your existing convention of defaulting its value to "true" for backwards compatibility.

Sorry, I missed this. I will clean up the naming issues. Thanks.

[kwokcb]

Fixed to use consistent option naming and added in example.

[jstone-lucasfilm]

This looks good to me, thanks @kwokcb! I'd be interested in thoughts from @lfl-eholthouser as well, since she has done development work in this area.

[jstone-lucasfilm]

@kwokcb With the recent contribution of Metal support, you may need to update this changelist to accommodate the refactoring of GLSL/MSL rendering functionality.

[lfl-eholthouser]

I think this look great @bernardkwok. Thanks!

[kwokcb]

Merge is done. The only change is to add the Python wrapper for Metal version of the TextureBaker interface.