Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

I cannot open MERL .binary BRDFs in the latest release version windows10 x64 i7 #16

Open
hsdxpro opened this issue May 13, 2017 · 5 comments

Comments

@hsdxpro
Copy link

hsdxpro commented May 13, 2017

I've found the latest release version here:
https://www.disneyanimation.com/technology/brdf.html

Opening analytical BRDFs working well, but I'm unable to open any measured BRDFs from MERL database.

Here is a picture of what's happening:
image

In addition to the fix, It would be great to have a fresh released version.

Keep up the good work.

@brentb
Copy link
Contributor

brentb commented May 15, 2017

Thanks for reporting. I can reproduce this problem, and there seems to be a regression. I will look for the cause as soon as I have time.

@hsdxpro
Copy link
Author

hsdxpro commented May 15, 2017

Hi brentb, that would be awesome for my personal researches if you can fix it.

Please keep in mind, I think for lot of users it would be great to have binary (.exe) releases every 1 - 2 months. I don't know if you are the right people I can tell that.

I really appreciate your time, if it's not possible to globally change the release intervals. At least can I have a binary package after you have fixed this issue? Please it's really important for me.

Thanks in advance !

@hsdxpro
Copy link
Author

hsdxpro commented May 15, 2017

I've just noticed you are Brent Burley, wow I'm impressed by your older work https://disney-animation.s3.amazonaws.com/library/s2012_pbs_disney_brdf_notes_v2.pdf

I'm just curious. The paper said that the Burley diffuse BRDF have been modified because of artists. Is there any publication about the version that is not modified based on artist requirements?

Some notes - And as the result of artistic modifications the energy conserve checking integral can go above 1 in some circumstances for example when NdotL is small and roughness is big.

And I've noticed that the diffuse lobe is decreasing strongly when roughness is small and NdotL is small too. But with high roughness the diffuse lobe is just getting bigger (might be the non energy conserving retro reflection?). What's I've noticed, Is it based on merl data or is it the part of the empirical model?
Becasue I've checked Eric Heitz's paper about raytraced micro geometry & stochastic BSDF: https://eheitzresearch.wordpress.com/240-2/

And within my C++ framework I've noticed that at shallow light angles, so for NdotL small, diffuse is decreasing regardless of roughness, sure at different rate.

I know Eric Heitz's work isn't based on spectral physics, so it could be inaccurate, I'm just askin why is the diffuse lobe differences.
roughness ~= 0, NdotL ~= 0 // diffuse lobe decrease
roughness ~= 1, NdotL ~= 0 // diffuse lobe increase

@brentb
Copy link
Contributor

brentb commented Feb 6, 2018

Hi Richárd,

(Apologies for the very slow response.)
I don't have an easy way to do a windows build, but I still have this on my todo list.

Regarding your BRDF questions:

  • The BRDF was developed with artist input. There was no other version. The main non-physical concessions were in adding specularTint and allowing specular and sheen values to be pushed beyond what might be considered a plausible range. Sheen is also physically motivated (by the Merl cloth samples) but largely artist designed based on how we rendered cloth previously.
  • The albedo can indeed go above 1 due to several factors including diffuse retroreflection, our diffuse fresnel, sheen, and clearcoat. This is more a theoretical than practical problem as plausible baseColor values shouldn't reach 1. But we agree that energy conservation should be addressed.
  • The diffuse retroreflection grows with roughness. Is this what you are referring to? This (along with our diffuse fresnel) is approximating an effect observed in the MERL data, though this hasn't been modeled by any physical process.
  • Eric Heitz, Wenzel Jakob, and Chris Kulla have each created energy conserving microfacet models. None of them produce retroreflection. I would like to see better measured data as the MERL data is inaccurate at grazing.

@hsdxpro
Copy link
Author

hsdxpro commented Feb 6, 2018

Hi Brent !
I'm glad to hear back from you.

It's a bit funny but so much time passed, so I wrote my private framework for researching stuff like that :D including a multithreaded numerical integrator to check energy integral etc.

I referred only to the diffuse equation, not the whole equation with clearcoat, etc. And yes I'm referring to the diffuse retroreflection mainly.

Well I only checked the specular model from Eric Heitz, you might be right !

Yes I've read the paper, agree on "MERL data is inaccurate at grazing".

To summarize, I just wanted to create a similar but more energy conserving diffuse eq. like yours. I was able to correct the integral with a LUT / alu calculation respecting all variables, but it's a bit costly. So I ended up analyzing the MERL. What I've noticed the retro reflection exists, but not that strong like in your eq. That's causing the problem mainly. I might create something in the future for diffuse. But for now I just created a real time approx for Eric Heitz's specular multi scatter, hard to approx, but the results are just fine.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants