WinOsi Caustic Test Scene

[CTZn]

If that's what I think this is due to a small value of <large_mutation_prob>, wich cause, as far as I understand, bigger zones to be evaluated with a loss in local color accuracy. That is, big colored zones. When it converges, the zones merge together and form a smooth image.

When you set <large_mutation_prob> higher (but always below 1), it tends to the level of a pixel, not an area of the screen. That is, noise frequency is higher.

That's what I found, what do you think about that ?

[manitwo]

You expected this kinda artistic result? Why? Is v0.7 THAT different? Can you "fix" it?

Hi Kram

Thats not really something that had to be fixed - that's because of the way PathTracing works.
In Indigo 0.7 "MLT" and "BiDir" are temporarily disabled. In 0.6 not.
I rendered the 0.6 in the "MLT + BiDir"-mode and the 0.7 in "PathTracing"-mode.
Pathtracing is very good for exteriors and direct lighting but not for indirect light and caustics
(because the emitter is behind a glass-lens EVERTHING is lit by direct and indirect caustics).
But i'm not the one who should explain that - IanT described that much better.
If i find his explanation i'll post a link.

EDIT: http://www2.indigorenderer.com/joomla/f ... .php?t=572

[IanC]

MLT isn't disabled, just BiDir MLT.

I'm not sure if pathtracing is still in indigo, I haven't used it since back in the 0.4 days.

[manitwo]

MLT isn't disabled, just BiDir MLT.

oops youre right anyway, this scene is nearly impossible without BiDir.

I'm not sure if pathtracing is still in indigo,

It is!

I haven't used it since back in the 0.4 days.

You should. As i said - works wonderfull for exterior and for scene with much direct light. Also some things changed since 0.4 - PT is progressive now

[suvakas]

Any reason why BiDir is disabled?

[fused]

nick told us not to use BiDir becuase its buggy...

interesting comparison by the way... im really curious why the indigo render looks so different

[manitwo]

i dont think it looks too different.
Beside that lines (i posted that in the bug-forum) and the reflective parts
(which aren't show the caustics) the indigo image looks really good.
Especially for that time!
To be sure - you should see two images

Redqueen:




Indigo: (3h50min)

[Kram1032]

redqueen seems to be very fast...
Two questions: what are your settings for rendering in 0.6?
(large_mutation_prob; max_change; max_num_consec_rejections; russian_roulette_live_prob and max_depth)

And: did you change the two smooth spheres to be <sphere>s instead of <mesh>es?

I'm not sure, if you got the answer: The "secret Door" can be seen in the sence with Radium (PT): rendered in 3d by "tg_jp", you posted in the very first post. It's the second image.
And now, as I look at it again, I can see another change:
Behind the door, there should be a cylinder on the wall, but here, it's edgy...

[fused]

the redqueen image could use some AA..

and is there a reason why the lightsources look so different? the indigo image has a cross.

i also noticed that the sphrere doesnt really reflect the caustics and the caustics in the indigo image look much more "datailed".

think these two are really different

[Zom-B]

Here my result after 36h and 11k muts per pixel:



I'll try now a 0.7t3 version using specular for reflections, not phong...
strange about your rendering manitwo is the "black cross" inside the lenses
the main difference about Indigo & the rest of the world is this shadow, the stand throws on the left & right wall...

[CTZn]

Another difference seems to be that the mirror material in the Indigo image is not a perfect mirror.

[manitwo]

Two questions: what are your settings for rendering in 0.6?
(large_mutation_prob; max_change; max_num_consec_rejections; russian_roulette_live_prob and max_depth)

"metropolis" - "true"
"large_mutation_prob" - "0.4"
"max_change" - "0.02"
"max_num_consec_rejections" "100"
"russian_roulette_live_prob" "0.99"
"max_depth" "200"
"bidirectional" "true"

And: did you change the two smooth spheres to be <sphere>s instead of <mesh>es?

Till now not - but a good idea - i will

Here my result after 36h and 11k muts per pixel:

why are our renderings different? is your lens normal-smoothed?

strange about your rendering manitwo is the "black cross" inside the lenses

i rendered my pic bigger and resized it after the rendering - perhaps thats why

the main difference about Indigo & the rest of the world is this shadow, the stand throws on the left & right wall...

right - i believe that shadow is produced because of the bug i posted.

Another difference seems to be that the mirror material in the Indigo image is not a perfect mirror.

this might not be a mat-problem, i think they are so dark cause they aren't reflect indirect caustics.

[Zom-B]

why are our renderings different? is your lens normal-smoothed?

yes, my lens got the XML Tag for smoothing, but I recognized that it hasn't any Phong Tag in the 3DS
(perhaps Indigo even ignores this?!?).
I also recognized that I normal smothed the meshes for the light emitters, perhaps an issue...

I'm rendering now a 0.7t3 Scene now, looks better than the one in your first post

My Image is so unsharp, because my focus is set to 100 and not about 0.3,
this results in a small DOF for the whole Scene, changed to white_balance E now to to get this redish tint away

[Kram1032]

The funny thing about recognizing you is, that YOU gave me the hint^^

ZomB wrote:
I'm sure small boys would love to play with your balls

:lol:

@tato: you could use sphere primitives instead of lowpoly sphere meshes.
example (gives you a perfect sphere):
Code:
<sphere>
<center>1 0 0.3</center>
<radius>0.3</radius>
<material_name>mat</material_name>
</sphere>

_________________
cheers,
manitwo

Later, Tato answered:

what i found is that the perfect spheres speeded up the rendering nearly 2 times. from 30k to 55-60k. hm, and gives me different refraction

So, I thought, it would be faster this way^^

[IanT]

Excellent results manitwo ... although I think you might have to wait a few more days for those reflected caustics to converge (there would be the same problem with Maxwell or any other unbiased renderer too I suspect).

Ian.