three-gpu-pathtracer icon indicating copy to clipboard operation
three-gpu-pathtracer copied to clipboard

Metalness implementation seems incorrect

Open gkjohnson opened this issue 2 years ago • 3 comments

See straps that should be dark gray and yellow (and for some reason have a high metalness value)

image

gkjohnson avatar Apr 03 '22 20:04 gkjohnson

Is possible this is a normal map / texture wrapping issue? Are UVs within [0, 1]? Is the normal map being interpreted correctly?

gkjohnson avatar May 27 '22 20:05 gkjohnson

This can somewhat be emulated with the material ball when setting the color to be fairly dark, roughness high, and then ramping the metalness up and down. It seems to get really bright in the metalness mid range and entirely dark at the ends. Perhaps theres an issue with the GGX functions?

gkjohnson avatar Jun 06 '22 23:06 gkjohnson

Related: #221

gkjohnson avatar Jul 28 '22 02:07 gkjohnson

See metalness (or maybe just glossiness) issues in the metal spheres comparison:

  • loss of energy on metalness axis
  • loss of specular on smooth axis

image

gkjohnson avatar Oct 02 '22 06:10 gkjohnson

Squaring the roughness improves the falloff and make it look closer to the comparison though the "matte" setting still has low reflection. ~Metal may be the same~. Perhaps the textures are being read in the wrong color space? Or are stored in the wrong color space?

image

gkjohnson avatar Oct 02 '22 16:10 gkjohnson

Glancing fresnel looks incorrect - "khronos-SpecularTest":

image

gkjohnson avatar Oct 02 '22 21:10 gkjohnson

Closing in favor of #264

gkjohnson avatar Oct 04 '22 03:10 gkjohnson

Aviation Helmet after #315:

image

And Specular Test from Model Viewer:

image

gkjohnson avatar Dec 10 '22 06:12 gkjohnson