Page 14 of 15

Re: Adaptive sampling improvements

Posted: Fri Oct 25, 2019 2:42 pm
by alpistinho
marcatore wrote: Fri Oct 25, 2019 6:51 am Sorry for the dumb question.
With this film pipeline, could we have the same color results between blender output and network rendering output?
At the moment, it seems that Blender applies something that changes colors and the output are different.

Could your develop solve this issue? Or am I completely wrong?
Blender applies its own color management after the render is finished.

I intend to fix that, but I need to use the OpenColorIO library and it has been complicated to do on Windows.

Once that is done, I will be able to create new film pipeline plugins that implements the same kind of color transformations that Blender does and then we'll be able to produce the same results anywhere

Re: Adaptive sampling improvements

Posted: Fri Oct 25, 2019 3:17 pm
by marcatore
Great. I hope you'll reach your goal!!!

Re: Adaptive sampling improvements

Posted: Fri Oct 25, 2019 11:01 pm
by alpistinho
lacilaci wrote: Fri Oct 25, 2019 7:50 am unless there's visible sky... what?
I am having a hard time reproducing that. Everytime I put any visible sky, it works as intended and it gets very few samples.
Would you be able to provide a scene that shows the issue?

Re: Adaptive sampling improvements

Posted: Sat Oct 26, 2019 6:40 am
by lacilaci
alpistinho wrote: Fri Oct 25, 2019 11:01 pm
lacilaci wrote: Fri Oct 25, 2019 7:50 am unless there's visible sky... what?
I am having a hard time reproducing that. Everytime I put any visible sky, it works as intended and it gets very few samples.
Would you be able to provide a scene that shows the issue?
I'm trying to create a small example scene, but there it works as intended... super weird.
I'm trying to figure out what makes this happen

Re: Adaptive sampling improvements

Posted: Sat Oct 26, 2019 6:51 am
by lacilaci
Ok, so what happens is that when I have camera outside the interior then sky gets very little to no sampling and stuff works as it should.
outside.jpg
But when I put camera inside and at the same time I see the outside parts/sky then sampling pretty much just goes for the sky
inside.jpg
Btw, there is no glass it's just an opening so I have no idea why this happens

Re: Adaptive sampling improvements

Posted: Sat Oct 26, 2019 6:53 am
by alpistinho
Very weird.

Does it happen if it is a simple box with a hole?
Happens with both sky and hdri?

Re: Adaptive sampling improvements

Posted: Sat Oct 26, 2019 9:52 am
by lacilaci
alpistinho wrote: Sat Oct 26, 2019 6:53 am Very weird.

Does it happen if it is a simple box with a hole?
Happens with both sky and hdri?
Sorry I'm not at pc anymore. Will check on monday few options. But aside this issue it seem to work really good now.

Re: Adaptive sampling improvements

Posted: Wed Oct 30, 2019 6:33 am
by lacilaci
hm, so in a simple, box and sky scenario this doesn't happen... I'm still trying to find out what is going on.
01.jpg
that weird pattern comes from env.cache (I really don't like env cache)

Re: Adaptive sampling improvements

Posted: Wed Oct 30, 2019 6:46 am
by alpistinho
lacilaci wrote: Wed Oct 30, 2019 6:33 am hm, so in a simple, box and sky scenario this doesn't happen... I'm still trying to find out what is going on.

01.jpg

that weird pattern comes from env.cache (I really don't like env cache)
Yeah, I've tried something like that and it didn't reproduce the issue

The BlendLuxCore you're using had a bug with the denoiser enabled. It was using the denoised image to drive the sampling, maybe it has something to do with that.
I sent a PR to the official repo and it should be fixed now if you want it.

Re: Adaptive sampling improvements

Posted: Wed Oct 30, 2019 7:13 am
by lacilaci
That denoiser bug was definitely the problem, works now
02.jpg