SBMCD (Sample-based Monte Carlo Denoising using a Kernel-Splatting Network)

Discussion related to the LuxCore functionality, implementations and API.
User avatar
FarbigeWelt
Donor
Donor
Posts: 1046
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: SBMCD (Sample-based Monte Carlo Denoising using a Kernel-Splatting Network)

Post by FarbigeWelt »

Odilkhan Yakubov wrote: Fri Jun 14, 2019 9:33 am It will be cool if someone implement this feature to the engine.
Although I am not a software developer and also do not know LuxCoreRender code I assert if there is not any gratis, free, open source software with a smart interface available the method described in the paper takes several months full time work to be implemented, not to mention all the details missing for a proper implementation.

The method is really cool and its results very promising but sorry implementation in LuxCoreRender is currently and in near future
impossible (personal opinion).
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
MacBook Air with M1
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: SBMCD (Sample-based Monte Carlo Denoising using a Kernel-Splatting Network)

Post by lacilaci »

FarbigeWelt wrote: Fri Jun 14, 2019 10:48 am
lacilaci wrote: Fri Jun 14, 2019 9:48 am Results aren't that impressive
Are you sure? Did you look at the pictures on p. 125:11 of the paper linked on main post?
I don't see anything I wouldn't see with OIDN from my experience, so I doubt it is worth investing the time on this.

Maybe it could be better in some cases, and if we still had only BCD it would be definitely worth it. What is there to gain now? We have a good denoiser and we would have 2? While still missing key features like displacement for example...?
User avatar
FarbigeWelt
Donor
Donor
Posts: 1046
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: SBMCD (Sample-based Monte Carlo Denoising using a Kernel-Splatting Network)

Post by FarbigeWelt »

lacilaci wrote: Fri Jun 14, 2019 11:25 am We have a good denoiser and we would have 2?
While still missing key features like displacement for example...?

This right, there is no need for a second one. And it is not sure that the one of the paper deals better than OIDN in scenes where PIDN fails currently.

I have the same opinion. Implementation of more useful missing features is more important.

By the way this sample based monte carlo denoiser takes at least 600 MB RAM for a 1920*1080 picture and takes estimated at least 3 s. (Estimation is based on paper‘s note that 75 floating point numbers are required per sample, guessing 16 operations per number.)
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
MacBook Air with M1
Post Reply