Path Space Regularization (aka the solution to SDS paths)

Discussion related to the LuxCore functionality, implementations and API.
User avatar
B.Y.O.B.
Developer
Developer
Posts: 4146
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

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

Post by B.Y.O.B. »

Dade wrote: Tue Sep 03, 2019 2:39 pm (and there start to be very little, if no source at all, of fireflies, if you use the full "arsenal" of features).
I thought the same, it will probably make clamping obsolete in most scenes because the "fireflies" will resolve to noise-free caustics.
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

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

Post by lacilaci »

B.Y.O.B. wrote: Tue Sep 03, 2019 3:06 pm
Dade wrote: Tue Sep 03, 2019 2:39 pm (and there start to be very little, if no source at all, of fireflies, if you use the full "arsenal" of features).
I thought the same, it will probably make clamping obsolete in most scenes because the "fireflies" will resolve to noise-free caustics.
No way. The performance hit and uselesness of caustics in many indoor lighting situations will definitely keep clamping as important optimization in luxcore. But it would be cool if it actually was so fast and problem free that clamping isn't needed.
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

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

Post by lacilaci »

So I guess this is not in the latest daily builds, or disabled by default?
acasta69
Developer
Developer
Posts: 472
Joined: Tue Jan 09, 2018 3:45 pm
Location: Italy

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

Post by acasta69 »

Only builds originated from the master branch are published as "latest", but probably you can get it anyway from here:
https://dev.azure.com/LuxCoreRender/Lux ... uildId=687
You should see an "Artifacts" button in the upper right corner allowing to download the build.
Support LuxCoreRender project with salts and bounties

Windows 10 64 bits, i7-4770 3.4 GHz, RAM 16 GB, GTX 970 4GB v445.87
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

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

Post by lacilaci »

acasta69 wrote: Wed Sep 04, 2019 5:29 am Only builds originated from the master branch are published as "latest", but probably you can get it anyway from here:
https://dev.azure.com/LuxCoreRender/Lux ... uildId=687
You should see an "Artifacts" button in the upper right corner allowing to download the build.
Thanks, doesn't work there either, I guess it has to be enabled somehow... Oh well, ok I wait then...
provisory
Posts: 235
Joined: Wed Aug 01, 2018 4:26 pm

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

Post by provisory »

Edit:
You can find the settings on GitHub, but I rather not write them here, I'm sure Dade will tell if/when he wants us to know.
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

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

Post by lacilaci »

Ok, I "hacked" it into blendluxcore... :D
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

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

Post by lacilaci »

so I know where my super saturated colors on photongi come from...
It seems lighttracing saturates volumes a LOT so a thin glass with weak absorption looks ok on pathtracing, but is super dark and saturated in lighttracing (I've seen those crazy saturated colors in a very first pass where lighttracing samples were shown)

It likely affects PSR too, cause volumes are still pretty dark even though sds caustics show through
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

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

Post by lacilaci »

And yes, caustics don't seem to converge at all.. 3000 passes of pool water is just as blurry as ~600... some fireflies do appear though...
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

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

Post by lacilaci »

I know it's early to some testing but I had a while so I played around a bit...

This is 300 samples with psr scale at 1.5:
visiblewater.png
This is 1300samples at scale of 2:
1300scale2.png
you can see that it never reaches the same contrast and brightness of lower scale, it's just getting a bit cleaner in terms of noise

and here's how caustics look not refracted (water invisible to camera):
invisiblewater.png
What I find interesting is that lower values than 1-1.5 also makes caustics very weak and instead of seeing a lot of fireflies I see just hint of caustics,
this is 1700samples with scale 0.4:
scale04_1700samp.png
So maybe caustics are getting sharper, but also weaker and so we cannot really see them?
Post Reply