Learning Specular and Roughness fails and produces a strange phenomenon

Question

Learning Specular and Roughness fails and produces a strange phenomenon

HaldonFu opened this issue 2 years ago · 5 comments

Thanks for your excellent work!
I use the duck data from phySG you provoided, and I used the mask *.png in stage1, mask_weight=100.0, but the result of the stage2 is incorrect.
The logim result as follow:

Can you tell me where the problem might be?

Answer 1 · 2022-07-22T12:30:51.000Z

This work assumes colocated flash lighting, rather than the envmap lighting as in PhySG.

Answer 2 · 2022-07-23T01:56:15.000Z

Thank you for your reply. I also want to ask how the results of duck or kitty in the synthetic_assets are obtained. Can you tell me?

Answer 3 · 2022-07-23T13:26:25.000Z

You might want to take a closer look at the readme regarding synthetic data rendering with Mitsuba.

Answer 4 · 2022-07-25T12:52:09.000Z

Thank you for your reply! I want to know how to convert the envmap_lighting data as in PhySG into flashlight data, and how to get the mesh, diffuse/specular albedo and roughness like in synthetic_assets? Can you tell me?

Answer 5 · 2022-07-29T02:50:20.000Z

As far as I know, there is no way to convert images with envmap lighting into images with flash lighting. You would have to re-render using Mistuba.