Chaos Corona Forum
Chaos Corona for 3ds Max => [Max] General Discussion => Topic started by: Polymax on 2013-08-27, 08:57:04
-
What about "Metropolis light transport "?
It might be worth it to implement? Instead of PPM in VCM, or ADD.
Or is it not worth the effort? Ondra, can you tell us about this algorithm, how effective is it, or vice versa, is not effective?
-
metropolis just does not work and never had. It might work occasionally in some scenes, but it is very easy to find a scene where each particular metropolis flavor fails. And it is not progressive in the sense you are used to, the image randomly abruptly changes when rendering - you see an image with completely missing portions that appear randomly out of nowhere after a random amount of time... or not at all
Some examples from our experiments:
(http://i.imgur.com/WZBx9ac.png)
(http://i.imgur.com/zBpBGJb.png)
(http://i.imgur.com/VnSQBlf.png)
(http://i.imgur.com/u9YfA4n.png)
(http://i.imgur.com/VS7ZJjh.png)
-
I find that Indigo image surprisingly good actually!
By the way, does the geometry of that scene extend much beyond what's visible from the camera? Cause VCM could suffer in such case, and could actually use some simple form of Metropolis to remedy that.
-
more tests
-
Han do you use indigo with GPU?
-
Yes, but my GTX 580 no advantage over my CPU i7 3930 K.
Indigo is now on the CPU faster.
update In these tests, the Indigo GPU is not used.
-
But the secondary caustics has only Corona ))
-
Indigo always had very good MLT implementation.
-
the metropolis is very important. We want it.
with bidirectional path tracing, caustic effect not cleaned very well.
I think the best method metropolis light transport
nwidia iray very well for example..
-
If you look at the corona image there are coloured caustics on the lower left, where are these in the other images?
-
the metropolis is very important. We want it.
with bidirectional path tracing, caustic effect not cleaned very well.
I think the best method metropolis light transport
nwidia iray very well for example..
MLT usually looks like a good idea, but is not ;). I want to experiment more with more advanced algorithms, but speeding up path tracing and doing stuff like C4D plugin or hair support has priority.
-
iray and corona caustic comp.
I like corona render. Very good. But this caustic firefly problem for me.
-
40 minutes for such simple scene is still way too much for it to be usable ;). Proper VCM implementation could do this in 40 seconds ;)
-
my computer's processor old :)
you can give me an example of VCM practice setting?
-
Render 25.10.2013 Iray with Arch sampler + Caustic
-
I viewed a rendering program on the market.
I've tested. I did not see the good as Corona. You found the right place ...
Corona best. VCM work very good.
-
metropolis just does not work and never had. It might work occasionally in some scenes, but it is very easy to find a scene where each particular metropolis flavor fails. And it is not progressive in the sense you are used to, the image randomly abruptly changes when rendering - you see an image with completely missing portions that appear randomly out of nowhere after a random amount of time... or not at all
Some examples from our experiments:
(http://i.imgur.com/WZBx9ac.png)
(http://i.imgur.com/zBpBGJb.png)
(http://i.imgur.com/VnSQBlf.png)
(http://i.imgur.com/u9YfA4n.png)
(http://i.imgur.com/VS7ZJjh.png)
Hey ! sorry, i'm a bit late to the party! ^^
What metropolis implementation is this ? Veach's one ?
-
indeed a "bit" late. AFAIK it was that energy redistribution stuff... dont remember the exact name, the one with very short chains
-
indeed a "bit" late.
Just a bit :)
Thanks for the answer Ondra.