Bedroom with flies

delle · Post by **delle** » Fri Sep 26, 2008 7:54 pm

Hi, this is my second render with Indigo. It's an old Easy Architect 3D project exported to Indigo.

The room is full of flies

Is the same case of the fireflies? (even if they are dark?)

Is there a way to limit or eliminate them (don't say DDT

). Leaving the image to cook longer does't help...

Thank you

Delle

rgigante · Post by **rgigante** » Fri Sep 26, 2008 9:25 pm

Hi Delle, how long did the rendering ran?
Is there any "Specular" material in your scene (i mean glass on windows?)?
Is there any material exposing high albedo (for example a white with 255,255,255 values?)

Regards, Riccardo.

CTZn · Post by **CTZn** » Fri Sep 26, 2008 9:30 pm

Filter ! Overbright pixels are a pain with mitchell-netralvi pixel filtering, it underlines strong contrasts.

PureSpider · Post by **PureSpider** » Fri Sep 26, 2008 9:33 pm

*OT* Would you mind sharing that curtain material?

Post by **OnoSendai** » Fri Sep 26, 2008 10:21 pm

The dark spots look like a bug.
Can you try and isolate them in a simpler testscene?
thanks.

suvakas · Post by **suvakas** » Sat Sep 27, 2008 12:56 am

Could it be a scale issue?

delle · Post by **delle** » Sat Sep 27, 2008 2:16 am

Hi, I've eliminated the glass from the window and the dark flies went away (see new scene version that is still cooking).

From what I've understood, the problem is present only if the Sun directly crosses the specular material (glass).

P.S. Here is the curtain material (very simple, is Indigo that it's a great engine

)


<material>
<name>Material_56_Diffuse</name>
<diffuse>
<albedo>
<constant>
<rgb>
<rgb>0.800000 0.800000 0.800000</rgb>
<gamma>2.2</gamma>
</rgb>
</constant>
</albedo>
</diffuse>
</material>
<material>
<name>Material_56_DiffuseTransmitter</name>
<diffuse_transmitter>
<albedo>
<constant>
<rgb>
<rgb>0.800000 0.800000 0.800000</rgb>
<gamma>2.2</gamma>
</rgb>
</constant>
</albedo>
</diffuse_transmitter>
</material>
<material>
<name>Material_56</name>
<blend>
<a_name>Material_56_Diffuse</a_name>
<b_name>Material_56_DiffuseTransmitter</b_name>
<blend>
<constant>0.300000</constant>
</blend>
</blend>
</material>

rgigante · Post by **rgigante** » Sat Sep 27, 2008 9:38 pm

as expected... Specular mats causes fflies... (as on dispersive prism ono)

delle · Post by **delle** » Fri Oct 24, 2008 12:50 am

Night version.

Delle

delle · Post by **delle** » Fri Oct 24, 2008 8:47 pm

Hi, I've done a test with "Specular material" and transparent=false to simulate a mirror.

Even in this case fireflies are present (over 2750 samples per pixel)

Note that they are more present in the direction of the sun reflection (on the wall and onto floor).

I think that some ray ramdomly shot toward the sun (reflection) get a "so high" value that is impossible to reduce it with other rays when they are averaged with some linear filter.

I don't know how Indigo is coded, but maybe it is possible to eliminate the brightest (and the darkest) rays (the amount of rays eliminated may be configurable, in percentage) from the calculus (for specular material).

More ore less as it does a "salt and pepper filter" with the pixels of an image.

This is my 2 cents contribute...

Bye

Delle

CTZn · Post by **CTZn** » Fri Oct 24, 2008 9:16 pm

removing the fireflies would look nicer but would induce bias too; it's up to Ono to find the balance between. Because if you look carefully you can see that in your image, ff appear in the reflect of the light bouncing on the wall: it is meaningfull, but also a complex sampling situation for bidir. Using mlt only would lead to slower, but ff free renders.

I would distribute their energy over a larger area rather than delete them, or raise sampling around them (just in case) but I don't know if that's even possible.

The Unknown · Post by **The Unknown** » Fri Oct 24, 2008 11:42 pm

Perhaps the same problem:

Image 1, rendered with Indigo 1.0.6
6000 spp
BiDir=true
Filter=mn_cubic
MNCR=5000

Image 2, rendered with Indigo 1.0.6
5000 spp
BiDir=true
Filter=mn_cubic
MNCR=1000
Objects with diffuse material and some plants with an alpha texture.

(I don't know if this is already fixed in the current release because the last version i tried had problems with smoothing normals.)

CTZn · Post by **CTZn** » Sat Oct 25, 2008 2:14 am

In my opinion it is related, yes, apart for the lion wich looks fine to me; the sun is sampled directly, I think you used camera diffraction wich can take long times to finish (camera diffraction can be another candidate for "fireflies" methinks).

Advanced settings (depth, maxNumConsecRejections) may have an impact on these issues, impact wich is still to evaluate.

Normals issues on external meshes were fixed in a recent release

The Unknown · Post by **The Unknown** » Sat Oct 25, 2008 5:41 am

Both images above with:
aperture_diffraction = true
max_depth = 10000
large_mutation_prob = 0.4
max_change = 0.01

Diffraction was true all the time.
I will do some tests without aperture diffraction, thanks for the hint!

The Unknown · Post by **The Unknown** » Sat Oct 25, 2008 7:37 am

Image 3, BiDir=true

Image 4, Metropolis=true

Both images with aperture diffraction false, all other settings are the same.
I hope it helps.

Bedroom with flies

Bedroom with flies

Who is online