LuxCoreRender v2.1alpha3 released

News & Announcements regarding releases, features, exporters and project coordination.
User avatar
Sharlybg
Donor
Donor
Posts: 3101
Joined: Mon Dec 04, 2017 10:11 pm
Location: Ivory Coast

Re: LuxCoreRender v2.1alpha3 released

Post by Sharlybg »

Because joinig several hundreds of selected objects takes time and makes not much sense.
Even high quality Pro trees model are single mesh ! try to create your trees with dedicated addon or software like The grove.
Support LuxCoreRender project with salts and bounties

Portfolio : https://www.behance.net/DRAVIA
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: LuxCoreRender v2.1alpha3 released

Post by lacilaci »

Can I ask why does random sampler tooltip say it's recommended only when used with denoiser?
Is there some problem with it? I haven't noticed much differences against sobol for now, except a little better noise performance here and there.
User avatar
B.Y.O.B.
Developer
Developer
Posts: 4146
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: LuxCoreRender v2.1alpha3 released

Post by B.Y.O.B. »

Sobol has a better noise pattern that converges more quickly, but it introduces some correlation between samples which makes the denoiser result a bit worse.
By the way, Metropolis produces a lot of correlation and should not be used with the denoiser.
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: LuxCoreRender v2.1alpha3 released

Post by lacilaci »

B.Y.O.B. wrote: Thu Aug 09, 2018 12:25 pm Sobol has a better noise pattern that converges more quickly, but it introduces some correlation between samples which makes the denoiser result a bit worse.
By the way, Metropolis produces a lot of correlation and should not be used with the denoiser.
Is the difference in convergence performance big? Should I just avoid using sobol if I want to use denoiser in any way?
Could it for example ruin high frequency details in textures?
User avatar
B.Y.O.B.
Developer
Developer
Posts: 4146
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: LuxCoreRender v2.1alpha3 released

Post by B.Y.O.B. »

In my tests, the difference between Random and Sobol was not that big. Random delivers a slightly better result, but in complex scenes this might be countered by its worse convergence.
lacilaci wrote: Thu Aug 09, 2018 12:32 pm Could it for example ruin high frequency details in textures?
Yes, a bit.
Attached is a comparison I made in the early stages of the denoiser integration.
Note that this was rendered with the bare minimum of samples (20, the denoiser is supposed to work only with 64 samples or more) back when the implementation only allowed a single thread.
So it would be interesting to make some new comparisons (and also compare with a noise-free ground truth).
Attachments
Random
Random
Sobol
Sobol
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: LuxCoreRender v2.1alpha3 released

Post by lacilaci »

I don't know if it shows on the crappy jpeg screenshot.
On top is random, bottom is sobol.
I'm seeing quite a bit better noise performance in the same amount of time on random sampler in some areas, like the O hole in the air purifier.

In the beginning of the rendering it looks like random sampler does worse job but at the end of 5min. rendering it gave somewhat better results. Not a big difference though.

I'm gonna test in more interior-like scenario.
Attachments
random_sobol.jpg
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: LuxCoreRender v2.1alpha3 released

Post by lacilaci »

For interior situations I'm getting somewhat inconsistent results.
Random sampler gets more samples in the same time and is also a little bit better looking especialy in rough reflections.
But then I change some camera angle and while still random sampler got more samples in the same given time, sobol got better results.
Both seem to give similar results with denoiser.(it was a tight crop area though)

I'm gonna do some more tests tomorrow and post results with more samples/time so that I don't compare noise and noise, but practical usable results hopefuly I'll have some time to do that cause I'm really curious about this.
Racleborg
Posts: 621
Joined: Sat Apr 07, 2018 10:31 am
Location: UK

Re: LuxCoreRender v2.1alpha3 released

Post by Racleborg »

Luxcore_Statistics(compare slots). I've not noticed this feature before; but whether it is new or not, it is cool!

You've fixed the Light Group issue-thank you.

I've wasted a lot of time trying to work out how to render sofas in Luxcore (a long story); it's time for me to start being more creative with such an awesome piece of software!

Thank you for all the hard work done on the project
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: LuxCoreRender v2.1alpha3 released

Post by lacilaci »

Racleborg wrote: Thu Aug 09, 2018 8:27 pm Luxcore_Statistics(compare slots). I've not noticed this feature before; but whether it is new or not, it is cool!

You've fixed the Light Group issue-thank you.

I've wasted a lot of time trying to work out how to render sofas in Luxcore (a long story); it's time for me to start being more creative with such an awesome piece of software!

Thank you for all the hard work done on the project
Wow, I haven't seen that one too!
User avatar
B.Y.O.B.
Developer
Developer
Posts: 4146
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: LuxCoreRender v2.1alpha3 released

Post by B.Y.O.B. »

Racleborg wrote: Thu Aug 09, 2018 8:27 pm Luxcore_Statistics(compare slots). I've not noticed this feature before; but whether it is new or not, it is cool!
It's new, from the first post:
new render statistics panel in the image editor toolshelf (LuxCore category),
provides in-depth render information with option to compare two render slots
I would like to make another panel with more in-depth information about the export process, so you can easily spot the objects/particles that slow down the export the most.
Post Reply