CGI tech news box

[Sharlybg]

Decide to open this thread to gather here news and good papers around the internet about CGI tech.Anyone can post here what he found interesting and we can discuss. As the thread goes wider and wider i will update the first post with all news and find from everyone so anyone can be update a bit by reading one post instead of a whole thread . From last development news papers to IA base rendering all are welcome. So i make the first jump.

* GDPT not a news for most of you but really love this one there are aso a github implementation available :

https://github.com/mmanzi/gradientdomain-mitsuba http://drz.disneyresearch.com/~jnovak/p ... index.html https://github.com/githole/gdpt and paper https://mediatech.aalto.fi/publications/graphics/GPT/

PDF about GDPT : https://darioseyb.com/report/gdpt.html

*Vray 4 research: Skynet ( is nvidia skynet Nvidia optix denoiser ) want to render your project :

https://www.chaosgroup.com/blog/experim ... x-denoiser

and advance tech to make portal useless https://www.chaosgroup.com/blog/v-ray-f ... dome-light

For IA denoiser i hope AMD or intel will replicate.

*Nvidia and opencl 2.0 : https://www.khronos.org/news/permalink/ ... on-support https://www.linkedin.com/pulse/nvidia-f ... hindriksen
https://www.phoronix.com/scan.php?page= ... IDIA-Preps

[Dade]

• http://cgg.mff.cuni.cz/~jaroslav/ => "Product Importance Sampling for Light Transport Path Guiding" plus referenced papers (Corona UHD cache finally explained)
• http://cgg.mff.cuni.cz/~jaroslav/papers ... /index.htm => to finally improve Russian Roulette

[patrickawalz]

Unbiased Photon Gathering for Light Transport Simulation



A Practical Extension to Microfacet Theory for the Modeling of Varying Iridescence


There have actually been numerous papers surrounding the use of the GGX Microfacet Distribution. A lot of new papers are using this as the primary microfacet distribution. I would love to see this research work its way into luxcore.

[Sharlybg]

I found this blog very interesting, the author also mention Luxrender :

https://agraphicsguy.wordpress.com/2016 ... h-tracing/

[Dade]

I found this blog very interesting, the author also mention Luxrender :

https://agraphicsguy.wordpress.com/2016 ... h-tracing/

From the link:

"I searched a lot of materials about MIS in BDPT, however there are only quite limited materials on the internet. Although some open source renderers, like luxrender, give detailed implementation, most of them doesn’t give any insight in the math behind it, without which one can be quickly confused by its code. SmallVCM expends the algorithm further, offering a better solution over MIS BDPT and it has detailed paper on the math."

For the record, old Lux uses a classic MIS solution (it was written before some new paper). LuxCore use a SmallVCM-like solution.

BTW, we are going to need a noise reduction post-processing filter for the future. At the moment, my best candidate is https://benedikt-bitterli.me/nfor/ (I talking of a practical and realistic candidate, not of some "thing" full of AI here and there ).

[B.Y.O.B.]

Benedikt Bitterli really creates high-quality papers.
He also implements some of the stuff you see in there in his Tungsten renderer, so you can actually look at the code which I really like.

I think it would be great to have this portal implementation in LuxCore, it could speed up indoor archviz.
A few months ago I started looking into it, but I never got something useful working because of a lack of time.

[Sharlybg]

LuxCore use a SmallVCM-like solution.

You want to say the BidirVCM we had under 1.7 ?

BTW, we are going to need a noise reduction post-processing filter for the future. At the moment, my best candidate is https://benedikt-bitterli.me/nfor/ (I talking of a practical and realistic candidate, not of some "thing" full of AI here and there ).

Improving raw noise output with tech like GDPT ; Guided Bidir/path tracing and Portals will provide better quality with less sacrifice. Without enough speed/efficiency denoiser produce such Ugly sploches....

I think it would be great to have this portal implementation in LuxCore, it could speed up indoor archviz.
A few months ago I started looking into it, but I never got something useful working because of a lack of time.

Powerfull Portals can really play a big role in high performance (especially in indoor archviz).This is why Fstorm GPU renderer is so appreciate in Indoor Archviz community. Bertrand Benoit Test make it obvious to me:

BERTRAND :

I initially found FStorm to be slow for a GPU renderer when rendering interiors, but this was before I placed portals in my windows. This sped things up considerably. It’s been years since I had to work with portals in any renderer, so it felt strange to go back to them. But why not. They certainly work well here.

Altogether, I found FStorm very pleasant to use. It gave me very fast feedback on this interior (perhaps the fastest of all real-time renderers I’ve tried so far) and I could generate final 2K images in about an hour each on a single 1080ti.

[Harvester]

Hello. I am not sure if this could be useful or of interest to the developers but I found interesting and with references to technical papers the "Two Minute Papers" by Kàroly Zsolnai-Fehér, who appers to have been or still is involved in the Luxrender development. In the contrary, feel free to remove this post.

https://www.youtube.com/channel/UCbfYPy ... upoX8nvctg

https://users.cg.tuwien.ac.at/zsolnai/

Thank you for your attention.

[Dade]

LuxCore use a SmallVCM-like solution.

You want to say the BidirVCM we had under 1.7 ?

No, BIDIRCPU in latest LuxCore, the SmallVCM-like MIS can work for everything from Photon Mapping to BiDir and up to BiDirVM.

[Sharlybg]

No, BIDIRCPU in latest LuxCore, the SmallVCM-like MIS can work for everything from Photon Mapping to BiDir and up to BiDirVM.

Ah Ok good know .

by the way do you think it is possible to merge Our Bidir with GDPT + Portal and obtain a sort of combined raw speed benefit from all theses tech ?