Does NIE work well for Inverse Rendering with calibrated RGB images?

[seekever]

I have read the paper, the theory is so wonderful. In "6.2 Inverse Rendering of Geometry", a single-bounce forward rendering is done with known pointlights and a known diffuse-Phong BRDF. It means that the lights and BRDF are known. And the rendered result are only 1-channel values of meshes without RGB colors.

IDR works well for calibrated RGB images without known lights and BRDF, then does NIE work well in the same case?

[seekever]

I have read the example of joint recovery of geometry, material and lighting from multi-view images in Section 7 "Discussion", but the related codes could not be found. Is it possible to share these codes?