Page 2 of 3
Re: DLCS Explained
Posted: Thu Oct 25, 2018 6:50 am
by marcatore
B.Y.O.B. wrote: ↑Wed Oct 24, 2018 2:38 pm
Sharlybg wrote: ↑Wed Oct 24, 2018 11:19 am
Cache entry radius : The bigger the scene the bigger the number
Could we automatically compute this? (Maybe from scene bounding box)
If yes, we could show a multiplier to the user for those cases where the auto-computed value has to be adjusted.
I think that some tooltips and a remark on the documentation could be a good start helping users understanding how to setup the renderer.
I prefer documented freedom instead hardcoded tricks... or a button that switch from the automatic mode to an "advanced" one where you have all the options visible.
Re: DLCS Explained
Posted: Thu Oct 25, 2018 8:39 am
by B.Y.O.B.
I could use some help with the descriptions for the following parameters:
https://github.com/LuxCoreRender/BlendL ... ig.py#L146
- entry normal angle
- entry max. passes
- entry convergence threshold
- light threshold
- target cache hit ratio
- max. depth
- max. samples count
Re: DLCS Explained
Posted: Thu Oct 25, 2018 8:09 pm
by kintuX
Are there any specific papers which were used as a basis?
Did the fast run again... (a short overview):
Since i know
appleseed 1.7 has (had?) Light Tree sampler (WIP) ~ similar to DLSC, i Started with Petra's
GSoC 2017 with AppleseedHQ which lead me to
Efficient sampling of many lights (@ompf2.com). I then also checked the
psychopath renderer...
nothing here
So i headed back to
Dade's original thread & then read the
vray adaptive lights... did took a look at few test threads
here and
here... also, did some myself...
and now i'm like 'overwhelmed'
... got to
grok for some time - let thoughts settle down, sink in... but i also need a confirmation... i have to feel clear-minded of all the complexities to
'explain it in simple terms.'
ain't easy to follow R&D in CGI field
loosing it a bit
- Am i
tracing the right path?
Re: DLCS Explained
Posted: Thu Oct 25, 2018 9:42 pm
by Dade
kintuX wrote: ↑Thu Oct 25, 2018 8:09 pm
Are there any specific papers which were used as a basis?
Nope, it is original work.
It is loosely inspired by this but they are (probably) using the photon map to extract the data while I have to build everything from scratch for obvious reasons (I don't have a photon map).
Re: DLCS Explained
Posted: Sat Oct 27, 2018 12:04 pm
by kintuX
Dade wrote: ↑Thu Oct 25, 2018 9:42 pm
kintuX wrote: ↑Thu Oct 25, 2018 8:09 pm
Are there any specific papers which were used as a basis?
Nope, it is original work.
It is loosely inspired by this but they are (probably) using the photon map to extract the data while I have to build everything from scratch for obvious reasons (I don't have a photon map).
Wow, respect for the work done and thank you for this 'photon VS RT' info, totally missed that...
I assume then, since you followed the principle, behavior is the same?
(It feels pretty much so)
Re: DLCS Explained
Posted: Sat Oct 27, 2018 2:30 pm
by Dade
kintuX wrote: ↑Sat Oct 27, 2018 12:04 pm
I assume then, since you followed the principle, behavior is the same?
(It feels pretty much so)
I guess yes ... but I don't have the V-Ray sources so I'm not sure
Re: DLSC Explained
Posted: Sun Dec 02, 2018 9:33 pm
by Sharlybg
This one seem to be a hard scenario case for DLSC. Don't know wich settings will fix theses artifact. I only play with warmup samples and entry radius. Fom my perspective DLSC should be simply avoid in case of heavy indirect lit 3D scene. But i'm curious to see wich settings could work for this one :
i use last BETA 3.
Ref Log power render :
Resulting DLSC :
Blend file:
Re: DLSC Explained
Posted: Mon Dec 03, 2018 8:33 am
by Dade
Sharlybg wrote: ↑Sun Dec 02, 2018 9:33 pm
This one seem to be a hard scenario case for DLSC. Don't know wich settings will fix theses artifact. I only play with warmup samples and entry radius. Fom my perspective DLSC should be simply avoid in case of heavy indirect lit 3D scene. But i'm curious to see wich settings could work for this one :
Do I have to change something in the attached scene ? If I run the scene as it is, it seems to work fine in LuxCoreUI:
and Blender:
Re: DLSC Explained
Posted: Mon Dec 03, 2018 10:27 am
by Sharlybg
You have to let it run an hour or more to see artifact. Anyway yesterday on my home rig artifact were more strong easy to spot. And today after updating on the firm rig i've to render for an hour to easily spot every splotches. But i remenber i had done the update before doing test yesterday ....
will try when back home.
here the result one hour :
artifact underlined:
Re: DLSC Explained
Posted: Mon Dec 03, 2018 11:54 am
by Sharlybg
And here is the same with higher settings : see the splotches up left.
Entry radius = 15cm
Warmup samples = 4096