Chaos Corona Forum

General Category => Gallery => Topic started by: Polymax on 2013-06-25, 08:55:30

Title: VCM test
Post by: Polymax on 2013-06-25, 08:55:30
Test 1:

.... other tests soon!
Title: Re: VCM test
Post by: Alessandro on 2013-06-25, 09:19:23
Water feels really like water ;) !!!

Curious for the next test!
Title: Re: VCM test
Post by: ondrike on 2013-06-25, 09:44:54
Water has full geometry, or plane only? that looks absolutely perfect !!!!
Title: Re: VCM test
Post by: Polymax on 2013-06-25, 09:48:19
Water is only plane!
Title: Re: VCM test
Post by: ondrike on 2013-06-25, 09:54:39
this is superb, but only one sad think is, that VCM is very slow :/
Title: Re: VCM test
Post by: Polymax on 2013-06-25, 09:55:24
But Only VCM can this!
Title: Re: VCM test
Post by: ondrike on 2013-06-25, 09:57:14
sure, I´m pretty amazed, really, but you cannot use it for exterior render, cos it´ll take ages to get rid of noise
Title: Re: VCM test
Post by: Polymax on 2013-06-25, 10:05:03
I think for some of the exteriors is possible to use VCM. Exteriors with pools, example.
Title: Re: VCM test
Post by: Richy on 2013-06-25, 10:11:25
VFB filters off? I think some filter combination could get rid of noise.
Title: Re: VCM test
Post by: ondrike on 2013-06-25, 10:14:00
Polymax could you post the scene to learn from it?
Title: Re: VCM test
Post by: Polymax on 2013-06-25, 10:49:42
please, but this simple :)
Title: Re: VCM test
Post by: Ondra on 2013-06-25, 14:10:31
this is superb, but only one sad think is, that VCM is very slow :/

VCM is not slow - path tracing would take days to get this quality. There is nothing faster in general for this type of scene
Title: Re: VCM test
Post by: michaltimko on 2013-06-25, 17:03:08
this is superb, but only one sad think is, that VCM is very slow :/

VCM is not slow - path tracing would take days to get this quality. There is nothing faster in general for this type of scene

BiDir MLT can do the trick. I was using this method in indigo. It was hella fast for scenes like this comparing to PT BiDir.
Title: Re: VCM test
Post by: Polymax on 2013-06-25, 19:10:40
4 h 26 min - very good, but next be overnight render
Title: Re: VCM test
Post by: abdsf on 2013-06-25, 19:59:49
Thank you for sharing the file but i have problem i press render in the same setting and i get noise after 2000 pass!
its not like your image , i used A5
Title: Re: VCM test
Post by: Polymax on 2013-06-25, 21:41:05
Strange! I also use A5 version!
Title: Re: VCM test
Post by: cecofuli on 2013-06-25, 23:47:36
Same here. If I open your scene and click render, the render is very different than your. No caustic.
It looks like some radius is too big.
Title: Re: VCM test
Post by: iliyang on 2013-06-26, 04:43:46
Polymax, if you render one image with path tracing and one with bidirectional path tracing modes, these could serve as a good reference for people to understand the differences between the algorithms. Path tracing should get both the underwater and the wall caustics noisy and bidirectional should only have problems with the underwater ones.
Title: Re: VCM test
Post by: Polymax on 2013-06-26, 09:33:41
Chamfer for glass - gives fireflies, nothing can remove them. Even a total shutdown visibility of the sun! It's a bug!
Only tomorrow I can get a good result ((.
Title: Re: VCM test
Post by: Polymax on 2013-06-26, 09:41:15
2 iliyang:
PT + PT can not cope with this task in a reasonable time!
Bidir - give similar results with the VCM, but will not be seen underwater caustics.

The purpose of this test - solve the complex task in the exterior to understand the capabilities of the algorithm. At the moment, none of the popular renderer is not able to!
Title: Re: VCM test
Post by: Alessandro on 2013-06-26, 09:42:22
Chamfer for glass - gives fireflies, nothing can remove them. Even a total shutdown visibility of the sun! It's a bug!
Only tomorrow I can get a good result ((.
Have you tried to use (or change) AA filter? Some time it helps to remove these dots...
Title: Re: VCM test
Post by: Polymax on 2013-06-26, 09:58:48
Quote
Have you tried to use (or change) AA filter? Some time it helps to remove these dots...
I'll  try, thanx!
-------------------------
I tried but it did not work (!
Title: Re: VCM test
Post by: iliyang on 2013-06-26, 10:36:04
2 iliyang:
PT + PT can not cope with this task in a reasonable time!
Bidir - give similar results with the VCR, but will not be seen underwater caustics.

The purpose of this test - solve the complex task in the exterior to understand the capabilities of the algorithm. At the moment, none of the popular renderer is not able to!

The point I was trying to make is to compare the quality achieved by PT, BPT and VCM in the same time (say 3 hours). I understand very well that neither PT nor BPT will be able to produce an acceptable image, but that's the point of the test. I'll be giving a talk about VCM at Siggraph in a month, and if we can get this comparison on an awesome scene that highlights the strengths of VCM, I can show the images in my talk, which will give Corona some publicity.
Title: Re: VCM test
Post by: Polymax on 2013-06-26, 11:57:39
Quote
The point I was trying to make is to compare the quality achieved by PT, BPT and VCM in the same time (say 3 hours). I understand very well that neither PT nor BPT will be able to produce an acceptable image, but that's the point of the test. I'll be giving a talk about VCM at Siggraph in a month, and if we can get this comparison on an awesome scene that highlights the strengths of VCM, I can show the images in my talk, which will give Corona some publicity.
Okay, I'll do it :)
Title: Re: VCM test
Post by: iliyang on 2013-06-26, 12:08:59
Okay, I'll do it :)

Take your time - better looking images means more fame for (https://plus.google.com/_/favicon?domain=corona-renderer.com&alt=feed) :)
Title: Re: VCM test
Post by: Ondra on 2013-06-26, 13:48:49
If you do the comparison for iliyan, use path tracing mode of VCM renderer, not progressive with PT/PT, as you'll probably get different results.


Also upload the problematic scene with fireflies as a bug
Title: Re: VCM test
Post by: Polymax on 2013-06-26, 14:15:01
Intermediate tests.
The purpose of the test - to achieve optimum noise in the minimum time, ie it is more practical test.
Title: Re: VCM test
Post by: cecofuli on 2013-06-26, 22:01:24
Nice test!! So, for now, the best solution (quality/speed) is obviously RT+HD
Title: Re: VCM test
Post by: Ondra on 2013-06-26, 23:41:33
yes, in this scene. But for example in a pool scene VCM is the winner
Title: Re: VCM test
Post by: iliyang on 2013-07-17, 11:49:10
Guys, does anyone have a nice comparison between bidirectional path tracing and VCM? I'd love to put some Corona image on my Siggraph slides.
Title: Re: VCM test
Post by: Polymax on 2013-07-18, 18:34:10
1. PT+PT
2.VCM
3.Bidir
Title: Re: VCM test
Post by: Captain Obvious on 2013-07-18, 19:34:22
That's a pretty substantial difference, and a pretty substantial render time too. Do you have Maxwell, by any chance? If not, any chance you could send the scene and I could give it a try.
Title: Re: VCM test
Post by: maru on 2013-07-18, 20:44:11
found a bug (I think so) : when rendering a region with VCM it behaves strange
Title: Re: VCM test
Post by: iliyang on 2013-07-19, 01:32:07
Thanks a lot, Polymax!
Title: Re: VCM test
Post by: Ondra on 2013-07-19, 06:21:08
found a bug (I think so) : when rendering a region with VCM it behaves strange

Small region, I assume?
Title: Re: VCM test
Post by: maru on 2013-07-19, 10:25:09
Yes, a small region. Here's a render. Second one has a small region rendered at the top corner of the water tank. It's after more than 3000 passes and more than 1 minute of rendering. Also, it looks like VCM isn't very good for glossy refraction.
Title: Re: VCM test
Post by: Bzdashekzapadlovski on 2013-07-21, 19:53:04
After opening your scene, found some changes in the Render setup.
What's this?
Title: Re: VCM test
Post by: maru on 2013-07-21, 21:39:43
It's a secret feature no one but trusted alpha testers should have seen. :/





But really it's what appears after you select bidir/vcm from "renderer" dropdown.
Title: Re: VCM test
Post by: quizzy on 2013-07-23, 10:45:19
some scene I did for a MR book written by a friend of mine now coronized......

both rendered for 23 minutes.....
Title: Re: VCM test
Post by: maru on 2013-07-23, 15:34:52
Looks like your water has no refraction.
Title: Re: VCM test
Post by: quizzy on 2013-07-23, 16:05:12
it does have refraction, but with an IOR of 1, because I did not want to distort the concrete beneath the water..
Title: Re: VCM test
Post by: maru on 2013-07-23, 22:07:23
That's what I meant. Used wrong words. :)