Path Space Regularization (aka the solution to SDS paths)

Discussion related to the Engine functionality, implementations and API.
User avatar
lacilaci
Donor
Donor
Posts: 1537
Joined: Fri May 04, 2018 5:16 am

Re: Path Space Regularization (aka the solution to SDS paths)

Post by lacilaci » Tue Sep 10, 2019 1:02 pm

I see that clamping also seem to affect how fast the caustics are rendered, but too low or too high values seem to damage the performance.. Is there a preferred value?

look at the difference on the side wall... lower clamping showed some lighting on the wall...

clamping 10K
10K.png
clamping 1 milion
milion.png
but then again going to 10 milion we get some light back... wtf
10mega.png
Last edited by lacilaci on Tue Sep 10, 2019 1:08 pm, edited 1 time in total.

User avatar
Dade
Developer
Developer
Posts: 2925
Joined: Mon Dec 04, 2017 8:36 pm

Re: Path Space Regularization (aka the solution to SDS paths)

Post by Dade » Tue Sep 10, 2019 1:04 pm

lacilaci wrote:
Tue Sep 10, 2019 11:39 am
1.btw, why do the caustics avoid these steps/edges? Is there some angular limitation in place?
No but it depends on where the sun is: sun light is refracted by water surface and hits the pool floor, to hit also a side wall, the sun must be in the right position.
lacilaci wrote:
Tue Sep 10, 2019 11:39 am
2.will there be some settings exposed to control sharpness and brightness? Right now it still seems very blurry and dark in water, nowhere close to my references
Aren't you setting the "path.pathspaceregularization.scale" parameter ? It is the parameter controlling the initial blur/sharpness :?:
Support LuxCoreRender project with salts and bounties

User avatar
Dade
Developer
Developer
Posts: 2925
Joined: Mon Dec 04, 2017 8:36 pm

Re: Path Space Regularization (aka the solution to SDS paths)

Post by Dade » Tue Sep 10, 2019 1:05 pm

lacilaci wrote:
Tue Sep 10, 2019 1:02 pm
I see that clamping also seem to affect how fast the caustics are rendered, but too low or too high values seem to damage the performance.. Is there a preferred value?
The latest code doesn't do clamping on light tracing values, I'm wondering what code are you actually using and how :?:
Support LuxCoreRender project with salts and bounties

User avatar
Dade
Developer
Developer
Posts: 2925
Joined: Mon Dec 04, 2017 8:36 pm

Re: Path Space Regularization (aka the solution to SDS paths)

Post by Dade » Tue Sep 10, 2019 1:07 pm

B.Y.O.B. wrote:
Tue Sep 10, 2019 12:10 pm
lacilaci wrote:
Tue Sep 10, 2019 12:04 pm
ven in this your example everything underwater seem waaaay too dark... There's no way this is how it would look in reality.
It looks to me like it was rendered without a sky (only with sunlight).
Yes, it was intended to make the caustics more visible (it is about like doing an indoor pool rendering).
Support LuxCoreRender project with salts and bounties

User avatar
lacilaci
Donor
Donor
Posts: 1537
Joined: Fri May 04, 2018 5:16 am

Re: Path Space Regularization (aka the solution to SDS paths)

Post by lacilaci » Tue Sep 10, 2019 1:10 pm

Dade wrote:
Tue Sep 10, 2019 1:04 pm
lacilaci wrote:
Tue Sep 10, 2019 11:39 am
1.btw, why do the caustics avoid these steps/edges? Is there some angular limitation in place?
No but it depends on where the sun is: sun light is refracted by water surface and hits the pool floor, to hit also a side wall, the sun must be in the right position.
lacilaci wrote:
Tue Sep 10, 2019 11:39 am
2.will there be some settings exposed to control sharpness and brightness? Right now it still seems very blurry and dark in water, nowhere close to my references
Aren't you setting the "path.pathspaceregularization.scale" parameter ? It is the parameter controlling the initial blur/sharpness :?:
I tried rendering with the scale parameter (different values) and also without
I'm using this build: https://dev.azure.com/LuxCoreRender/Lux ... uildId=709

User avatar
lacilaci
Donor
Donor
Posts: 1537
Joined: Fri May 04, 2018 5:16 am

Re: Path Space Regularization (aka the solution to SDS paths)

Post by lacilaci » Tue Sep 10, 2019 1:17 pm

Dade wrote:
Tue Sep 10, 2019 1:04 pm
lacilaci wrote:
Tue Sep 10, 2019 11:39 am
1.btw, why do the caustics avoid these steps/edges? Is there some angular limitation in place?
No but it depends on where the sun is: sun light is refracted by water surface and hits the pool floor, to hit also a side wall, the sun must be in the right position.
you mean to tell me no sun hits these balls?
balls.png
Something's wrong, I'm telling you...
Also clamping doesn't disable caustics, but it messes something up... But maybe it's just performance hit with large clamping values I don't know..

User avatar
lacilaci
Donor
Donor
Posts: 1537
Joined: Fri May 04, 2018 5:16 am

Re: Path Space Regularization (aka the solution to SDS paths)

Post by lacilaci » Tue Sep 10, 2019 1:19 pm

here's the scene I'm working with... Check it out pls, maybe I fucked up something somewhere..
poolwater_TESTING.blend
(2.77 MiB) Downloaded 8 times

User avatar
lacilaci
Donor
Donor
Posts: 1537
Joined: Fri May 04, 2018 5:16 am

Re: Path Space Regularization (aka the solution to SDS paths)

Post by lacilaci » Tue Sep 10, 2019 2:02 pm

this is how it should look like...
waterbodyvolume.png
there should only be a bit of refraction and bit of reflection and that's it... No heavy darkening and no black balls..

User avatar
lacilaci
Donor
Donor
Posts: 1537
Joined: Fri May 04, 2018 5:16 am

Re: Path Space Regularization (aka the solution to SDS paths)

Post by lacilaci » Tue Sep 10, 2019 2:27 pm

Enabling scattering completely kills sds so no caustics in pool will happen (Although I'm not sure if psr is supposed to work with volume scattering anyways)

User avatar
Dade
Developer
Developer
Posts: 2925
Joined: Mon Dec 04, 2017 8:36 pm

Re: Path Space Regularization (aka the solution to SDS paths)

Post by Dade » Wed Sep 11, 2019 9:31 am

lacilaci wrote:
Tue Sep 10, 2019 1:19 pm
here's the scene I'm working with... Check it out pls, maybe I fucked up something somewhere..
It is a strange problems, it is like if Metropolis is "trapped" on flat surfaces. A smooth sphere seems hard to render and flat sphere produces this very strange result:

face.jpg

It is like Metropolis struggle to find all faces, it finds one once in a while. I'm trying to figure out why.
Support LuxCoreRender project with salts and bounties

Post Reply