Page 3 of 3

Re: question

Posted: Thu Oct 22, 2020 2:30 pm
by Sharlybg
daros wrote: Thu Oct 22, 2020 2:16 pm Thanks a lod Dade.
Light distribution ipmroved a lot!
This is a 70 seconds test using oidn.
It is Bidir engine ?
what is your System ?

Re: question

Posted: Thu Oct 22, 2020 7:01 pm
by daros
yes, bidir+metropolis. My computer is a old lenovo x1 5th gen. I'm using a cloud rendering tool, Felix, that's why it's fast. I'm helping to implement Luxcore in to Felix.

Re: question

Posted: Thu Oct 22, 2020 9:31 pm
by Sharlybg
daros wrote: Thu Oct 22, 2020 7:01 pm yes, bidir+metropolis. My computer is a old lenovo x1 5th gen. I'm using a cloud rendering tool, Felix, that's why it's fast. I'm helping to implement Luxcore in to Felix.
That is really nice to see felix here. The rendering on felix website are really very high quality like the one you do with extremelly high setting.you are swiming in my dream :D

As Felix is interested in the usage of Bidirectionnal engine do you think a 4 to 8x speed up provided by Bidirectionnal path guiding can interest them even if their have gigantic computing power ? :mrgreen:
Image

Re: question

Posted: Fri Oct 23, 2020 10:19 am
by daros
i think yes, every speed improvement which do not has a cost in quality is welcome. Is there already an implementation in lux of the guided bidir?

Re: question

Posted: Fri Oct 23, 2020 10:33 am
by daros
interestingly this sample video shows some problems of this approach. https://cgg.mff.cuni.cz/~jirka/papers/2014/olpm/
It is true that it cleans up quickly direct and diffuse indirect illumination however caustics seem to suffer compared to standard bidir. I believe that caustics are important for delivering rich images, not only in interior spaces but even more in exterior urban spaces where often 50% of light is caused by caustics. But i imagine this are choices the user should take in the last moment depending from the client, budget, deadlines and so on: important is that engines are aligned as much as possible in order to not make the choice biased by other factors.

Re: question

Posted: Fri Oct 23, 2020 10:53 am
by Sharlybg
daros wrote: Fri Oct 23, 2020 10:33 am interestingly this sample video shows some problems of this approach. https://cgg.mff.cuni.cz/~jirka/papers/2014/olpm/
It is true that it cleans up quickly direct and diffuse indirect illumination however caustics seem to suffer compared to standard bidir. I believe that caustics are important for delivering rich images, not only in interior spaces but even more in exterior urban spaces where often 50% of light is caused by caustics. But i imagine this are choices the user should take in the last moment depending from the client, budget, deadlines and so on: important is that engines are aligned as much as possible in order to not make the choice biased by other factors.
There is an improved version of path guiding better than this one especially for caustics and probably more convenient to implement in Luxcore.
Here is the Video about that :
https://youtu.be/pCqqiwt3JZ4

BTW I don't know if this one is compatible with GPU.The best scenario would be a GPU compatible path guiding algo. BUT as we don't have something like that on the table i think starting with Bidir is a good step and this can also speed up simple CPU path tracing.
People with weaker PC will appreciate and it can unlock lot of animation scenario.

But I prefer to Trust DADe he is in this field for much longer and have a better understanding of the code even if it is hard from an artist point of view to see so much locked performance left. but who know what the futur is made of :idea:

Re: question

Posted: Fri Oct 23, 2020 11:13 am
by daros
Cool, thanks!