OpenImageDenoise

Discussion related to the LuxCore functionality, implementations and API.
User avatar
B.Y.O.B.
Developer
Developer
Posts: 4146
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: OpenImageDenoise

Post by B.Y.O.B. »

lacilaci wrote: Sun Feb 24, 2019 1:51 pm Those builds cannot be redistributed with luxcore?
The OIDN standalone binary is not required by LuxCore, only by the addon, so it would make no sense to put it into the LuxCore archives.
Of course, they will be included in the releases of the addon.
User avatar
B.Y.O.B.
Developer
Developer
Posts: 4146
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: OpenImageDenoise

Post by B.Y.O.B. »

Another demo, this time with Sharlybg's Nakagin Capsule scene, comparing to a 1000 sample reference render at the end:
https://youtu.be/_-6eCi4VX3o
zuljin3d
Posts: 76
Joined: Sun Apr 08, 2018 12:13 pm
Location: Moscow

Re: OpenImageDenoise

Post by zuljin3d »

B.Y.O.B. wrote: Sun Feb 24, 2019 9:54 pm Another demo, this time with Sharlybg's Nakagin Capsule scene, comparing to a 1000 sample reference render at the end:
https://youtu.be/_-6eCi4VX3o
:shock: amazing
Actualy sorry for my google translate english :)
provisory
Posts: 235
Joined: Wed Aug 01, 2018 4:26 pm

Re: OpenImageDenoise

Post by provisory »

I think, something is wrong with Avg. Shading Normal pass, and it reduces the effectiveness of OIDN.
In ASN pass three out of six faces of a cube become black.
I uploaded a simple test scene, and some images (ASN pass, combined, denoiser artifacts enlarged).
Attachments
avg_shading_normal.blend
(487.15 KiB) Downloaded 219 times
avg_shading_normal-asn.jpg
avg_shading_normal-combined.jpg
avg_shading_normal-denoised-artifacts.png
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: OpenImageDenoise

Post by Dade »

provisory wrote: Tue Feb 26, 2019 1:33 pm In ASN pass three out of six faces of a cube become black.
Normals have values between -1.0 and 1.0. If you are mapping the values in a .png file, you will get half of the values as black. It is the expected behavior but Oidn gets the right values (-1.0 to 1.0) in floating point format as it requires.
Support LuxCoreRender project with salts and bounties
provisory
Posts: 235
Joined: Wed Aug 01, 2018 4:26 pm

Re: OpenImageDenoise

Post by provisory »

OK, thanks
User avatar
Dez!
Posts: 368
Joined: Sun Apr 08, 2018 1:09 am
Location: Ekaterinburg
Contact:

Re: OpenImageDenoise

Post by Dez! »

Thank you for such a wonderful tool!

Now I want to provide my tests. This is a comparison between the BCD and OIDN in my animation project. This is a test for the suitability of the OIDN for animation.
In parentheses, I indicated the rendering time of each frame.

I've done a lot of tests on this scene for the best BCD settings for animation. It is configured in the best way possible.

As a result, I would like to note that OIDN does not only provide higher rendering speed.
OIDN also shows more small details.
In my opinion, where a BCD requires rendering to the noise level of 10,OIDN will show the same result at the noise level of 15.

Now I have a chance to finish this project.
The only thing left to do is to solve the noise issue behind the architectural glass.
There are a lot of shots in my project where the glass is close up.

Here's the results archive:
https://mega.nz/#!Ag4j1IKa!TJQFIWyW2do0 ... uVawOjKkS4
Linux Plasma | Ryzen 5, 32Gb, SSD M2, GT 590 RX | BenQ 27 | Wacom One | Microsoft Ergo | Tie Guan Yin tea
http://dezigner.tilda.ws/
User avatar
Sharlybg
Donor
Donor
Posts: 3101
Joined: Mon Dec 04, 2017 10:11 pm
Location: Ivory Coast

Re: OpenImageDenoise

Post by Sharlybg »

E-Cycles switch to OIDN to. The War has begun ! :twisted:

https://blenderartists.org/t/e-cycles-r ... 139717/576
Support LuxCoreRender project with salts and bounties

Portfolio : https://www.behance.net/DRAVIA
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: OpenImageDenoise

Post by Dade »

Dez! wrote: Wed Feb 27, 2019 6:46 am Now I have a chance to finish this project.
The only thing left to do is to solve the noise issue behind the architectural glass.
There are a lot of shots in my project where the glass is close up.
I'm aware of the problem with Oidn and ArchGlass. It is a problem very specific of the way material "transparency" work and interact with ALBEDO AOV.

To alleviate the problem, you can reduce the ArchGlass reflection parameter, more you reduce, less noise you will have in ALBEDO AOV and better Oidn will work.
Support LuxCoreRender project with salts and bounties
User avatar
Dez!
Posts: 368
Joined: Sun Apr 08, 2018 1:09 am
Location: Ekaterinburg
Contact:

Re: OpenImageDenoise

Post by Dez! »

Yeah, it worked. Thank you!
Linux Plasma | Ryzen 5, 32Gb, SSD M2, GT 590 RX | BenQ 27 | Wacom One | Microsoft Ergo | Tie Guan Yin tea
http://dezigner.tilda.ws/
Post Reply