PDA

View Full Version : Alternatives to MR Occlusion?


EricLyman
11-08-2006, 06:40 AM
Hey render know-it-alls:

I'm on a freelance job right now that's going to require rendering about three minutes of animation at HD (1280x720). I'm interested in turning out something looking great, and part of that will require the implementation of Occlusion (of some sort). I have the know-how in that area, but unfortunatly I don't have the computer power to make it happen fast. I'm on a single CPU machine that was new about three years ago. This is all in Maya by the way.

Previously I had used a light dome in my work to get an occlusion-like effect. I did this because I had access to a smallish renderfarm that would render with 'Maya Software' but not Mental Ray. The other reason was simply not knowing the in's of MR :) To clear any confusion up front, my definetion of a light dome is basically a sphere with roughly 80 lights casting shadows around a scene. This sort simulates occlusion... but not really.

I'm basically trying to find a method that will yeild good results without the massive rendertimes. I only have access to either Maya Software or Mental Ray for this work... I'll use MR's occlusion if it comes down to it, but with all the rendering I have, I won't have time for too many revisions. Any suggestions, tips or tricks?

the light dome in question:
http://img174.imageshack.us/img174/4168/domegb8.jpg (http://imageshack.us)

Here's the result of light dome usage with a quick test scene I made up. Rendered a 1280x720 frame in roughly 3-4 minutes. (the object in the scene has been subdivided quite heavily to simulate a heavy scene file):
http://img171.imageshack.us/img171/3515/domerenderya2.jpg (http://imageshack.us)

Here's the same scene, same frame as above, but rendered with MR's Occlusion. As you can see it puts the light dome to shame, but one frame took roughly 7-8 minutes (ouch!):
http://img218.imageshack.us/img218/6678/occrenderxf6.jpg (http://imageshack.us)

I've used final gather before as well, but from what I recall the render times are about the same as occlusion. Thanks in advance for any help.

mer
11-08-2006, 06:39 PM
Hey!

Final gathering is generally slower than occlution..BUT you get light bounces.

In xsiblog website there is and addaptive ambient occlution in xsi for faster results.Do a search.Perhaps you could implement it in other software too(maya max).

Ian Jones
11-09-2006, 12:04 AM
Try searching for MR dirtmap shaders... a similar result can be achieved. I'm not sure about rendertimes though, but I suspect a small increase is possible.

EricLyman
11-10-2006, 01:38 AM
I've been looking more in to this. It appears that Maya 7 has a new Final Gather based occlusion shader... similar results but supposedly faster rendertimes. Anyone try this? I'm going to do some testing either tonight or tomorrow, I'll post my results.

Here's my source:
http://www.eatbug.net/tutorial/fo/tutorial_fo.htm

Also, maybe baking the occlusion would help? More to come on this as well.

EricLyman
11-12-2006, 02:17 AM
Doing some tests with FG. Looks good, but the render isn't significantly faster. About as high quality as standard MR AO.

test scene: somewhat heavy geometry @ 1280x720:

fg_occlusion = 5:36
amb_occlusion (128 samples) 5:55

I ran another test with 64 samples just to see how good or bad it is. Rendered the same scene above in 3:33 minutes. It's somewhat more grainy of course, but I thought perhaps this could still be used with some blur in post. Anyone have experience with this technique?

amb_occ @ 64 samples:
http://img89.imageshack.us/img89/8951/ball64cd1.jpg (http://imageshack.us)

amb_occ @ 128 samples:
http://img299.imageshack.us/img299/9057/ball128fu8.jpg (http://imageshack.us)

amb_occ @ 64 samples w/ a gaussian blur of 2:
http://img141.imageshack.us/img141/590/ball64blurvg0.jpg (http://imageshack.us)

So.. results could be good. But perhaps it could be problematic where the blur actually moves occlusion pixels off of the geometry borders.

Interested in hearing your thoughts.

cpan
11-12-2006, 10:30 AM
the fg occlusion is faster but take care that the global FG settings influence the fg occlusion. With a 0.1 presample density and 100 rays you should get that sphere in seconds and not minutes, at ~the same quality.

cya

CGTalk Moderation
11-12-2006, 10:30 AM
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.