PRMAN Problems with MtorUltraFur and DepthMapBasedAmbientOcclusion

[playmesumch00ns]

Looking at your pictures I think you just want to blur your shadow map. Try turning the blur up to like 0.03 or something, but you may have to increase the number of samples

[luna20]

yes i tryed blurring the shadow maps and increasing samples (and...artifacts are away) but if u do that, u loose so much detail that the occlusion pass seems not usefull anymore because almost any pixel is greyish so that areas which are normally not affected by diffuse lighting are NOW...so,Unfortunately, that cant be the solution :(

[Noxerus]

xcpt 0x00000096 - *coughs slightly* I believe that this is a license problem. :shame:

[luna20]

no this can't be the problem because our filmschool has legal prman licenses that are not limited as i know...


Sylvia

[Anteru]

Did you try increasing the number of lights for the AO pass? We had similar artifacts whit Aqsis, which went away after increasing the light count.

[Mauritius]

Hello Miss junior TD! ;)

First I think st. is wrong if you need 26h per frame for ray traced ao. How many of your 70 PRMan licenses where used when Alfred reported that time?
I rendered very complex scenes with ambient occlusion on my single machine 2GHz Athlon with 3Delight and always managed to keep rendertimes below 2h per frame for 1k res frames.
Second I agree with Anteru -- you'll likely need to use more lights if you take the shadow map approach (which is definitely the way to go for now, anyway).
The rule of thumb is to use more lights and a smaller map size:
512x512 pixels * 64 maps = 16777216 samples
256x256 pixels * 256 maps = 16777216 more evenly distributed samples

There is a Python (http://www.python.org/) script at the Aqsis CVS at sourceforge (http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/aqsis/content/rib/features/occlusion/) that generates the resp. RIB archive with the lighting setup. I debugged this script and modified it to be a lot faster (in particular for render test purposes) by using a different algorithm. Read this thread (http://aqsis.sourceforge.net/phpBB2/viewtopic.php?p=2100#2100) at aqsis.com (http://www.aqsis.com). My new version of the script is here (http://io.plastickitten.net/occlmap_mm.py).
You can then use a RIBBox or an Archiver in SLIM to attach the script's RIB output to an MTOR coordinate system in your scene.

Depending on your camera pov and your geometry, you'll also likely be able to discard a few of the depthmap generating lights on the backside of the sphere. In particular for the pov of the image you posted, this will apply.
Don't forget to constrain the coordinate system, which has the archive containing the lights attached, to the camera's motion to ensure the lightless region stays on the sphere's backside.

If you have a few days, you can also send me the RIBs and I'll take a look. I'm sure though, that it won't be before next thursday.

Regarding the ultraFur issue: ask at the c.g.r.r newsgroup (via groups.google, if you don't have any newsreader installed) and particularly post it to the forums at renderman.pixar.com. Don't forget to mention you're a student -- you might get help from someone from Pixar more likely than if they think you're a plain customer w/o support contract.

Cheers,

Moritz

[luna20]

thank you Mauritius and Anteru!!!!!!!

Mauritius: I rendered the raytraced pass on 4 dual-Pcs = 9ghz (we have no gigabit network here...so i can't use networkrendering as much as i would like to) but for the dmapbased solution we tested renderings on 20 CPUs @ one time that was fun =)

cya
Sylvia

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.