Cornell-box test scene need someone to render it ...

[psor]

Heya folks!

I was reading some threads over at the maxwell forum again and was surprised by Torolf's post.
I'm posting the pictures here cause as far as I do remember you've to be registered to read
posts over there.



Torolf's Indigo v2.2.2 Renderings


Sampler: MLT - Integrator: PT




Sampler: QMC - Integrator: bidirectional PT



It looks like he is suffering of some kinda energy loss or errors through normals/low geometry.
I haven't contacted him yet so I'm not sure how his scene looks like.




Torolf's Maxwell Renderings


Maxwell v1.7.1




Maxwell v2.0






Tora2097's Maxwell v2.0 Rendering






micheloupatrick's Maxwell v2.0 Rendering ("OK I got it: denoising simply comes from resizing the file from 4k to 700 x 700.")






Q2's Maxwell v2.0 Rendering






As you can see there is some fishy thing going on and of course "they" are bitching about the
quality of Indigo, the dark edges on the spheres and the splotches the on the image cause
Torolf also used MLT to render it. I don't mind them bitching but I do mind the way things are being
handled over there anyway. So I quickly did set up a cornell-box scene and rendered it for 35min
on a P4 3Ghz and this is what I get with Indigo v2.2.5.

Sampler: QMC - Integrator: bidirectional PT




Sampler: MLT - Integrator: bidirectional PT




I don't see any dark edges around my spheres so I'm starting to wonder about his scene tho.


So I would like to find somebody with a faster rig to get the image cleaner in a shorter time.



The Scene file I've used is attached to this post and is set up like this:

resolution: 700x700
render method: biPT
MNCR: 1000
supersampling: 4 (I did choose this to avoid small fireflies popping up)
tonemapping: linear (to be fair and not introduce Indigo's camera tonemapping)
Aperture diffraction: disabled




If someone feels invited go for it.







take care
psor

[fused]

yes, 'bitching' is definitely the right word...

edit: the dark edges in torolf's rendering are weird. rendering now...

[fused]

hey, are pigm/pigs extensions still not allowed?

sorry for double posting....

tonemapping: linear (to be fair and not introduce Indigo's camera tonemapping)

[psor]

Yup pigs and pigm are still not allowed. Cows only!

Thanks for taking the challenge mate. ;o))





take care
psor

[fused]

pigs and and pigms are allowed now

no cows

[pixie]

Could you post your scene in 3ds?

[psor]

@fused

Thanks!


@pixie

I could, but I'll post it as .obj and .blend..





take care
psor

[fused]

bidir MLT, 1h, 1500spp:

edit: the reason for the blurryness might be this:

Code: Select all

        <splat_filter>
            <gaussian />
        </splat_filter>

and iirc a good mn_cubic default is:

Code: Select all

        <downsize_filter>
            <mn_cubic>
                <ring>0.333</ring>
                <blur>0.333</blur>
            </mn_cubic>
        </downsize_filter>

and not:

Code: Select all

        <downsize_filter>
            <mn_cubic>
                <ring>0.2</ring>
                <blur>0.6</blur>
            </mn_cubic>
        </downsize_filter>

2009_10_21_Cornell_Box_01.jpg (86.02 KiB) Viewed 23141 times

[psor]

Thanks fused, looks good! Maybe I play Stinkie and it's a bit out of focus.


And the funny thing is, cause I'm experimenting with luxrender too, that jeanphi
also said that a mitchell filter with 0.3333 | 0.3333 would be best. Anyway the
default in the Indigo inifile.xml is 0.2 | 0.6. And since I'm using 3dsmax I take
a guess that suvakas has used those values as default for maxigo.


I could be wrong tho.






take care
psor

[fused]

yeah, Suv probably just used the settings from the inifile...

maybe ask him to change the default value

[suvakas]

Yes, 0.2 and 0.6 are the defaults in Maxigo. Taken from the Indigo manual.

[fused]

heres the (sort of) finished image:

3900 spp

2009_10_21_Cornell_Box_01.jpg (74.35 KiB) Viewed 23108 times

[Stromberg]

Why do you care?

There will always be bitching about other render engines.
If we showed a indigo render that was clear in 5 mins that took hours in maxwell they will still find something to complain on, and visa verca.
Thats why i try to stay away from these kinds of topics, cause it's so frustrating to read, so many wrong statements.
Like i was reading a topic like this on the luxrender forum, and since luxrender did use longer time to clear and the rendering looked worse(from my eyes), they started complaining about indigo's portals that they are fake and biased..
Just want to say that it will never change, same as the mac vs pc discussion.

[lycium]

there are so many things wrong with these tests; why are people pretending not to notice that the scene settings are very very different? most of the images have a different phong exponent, and some aren't even using phong but perfect specular.

it's not even a case of apples vs oranges, it's more like apples and manatees.

final point: if the aim is to have a well-rendered cornell box, a renderer written in 1 day will actually perform better than a full professional renderer (using a simple list of objects instead of a k-d tree, path tracing instead of bidir, etc). how about a real scene, with real materials... real complexity! when you're watching an olympic sprint, you don't judge the winner by how quickly they can tie their shoelaces...

[psor]

@Stromberg

You are right. I care because I do know Torolf and I am aware that his English is bad. I saw him
posting over at Maxwell's and I've seen people shooting posts at him he couldn't answer properly.

I'm not sure why he did post those pictures at all. And it's not that important. The point is that
it is sad to show pictures of Indigo that are very questionable and represent Indigo in an improper
way. And it's sad to see a very talented guy posting a test while lacking the ability to reply
to those kinda posts. *sigh* But yeah ... fan boys stay fan boys and are blinded anyway.
From my point of view it's unimportant which tool to use it's more important if it suits your needs.
The name doesn't matter nor where it comes from.




@lyc

I don't wanna argue with you. There is no point in doing this. Even if all settings and all things
are the same it will not be a good comparison anyway. Nobody but NL knows what algorithms
Maxwell is using. And to the artists it doesn't matter. They just see the result and that's all that
matters nothing more. Words like PT/MLT/QMC/... are for tech guys not for artists.






take care
psor