PDA

View Full Version : Render Regions in MR


Dogway
12-05-2006, 09:02 AM
I want to render a 4000 x 3000 about scene in MR, with FG and GI.

The thing is that when rendering in MR with FG and GI, every region has different illumination values. Well that is easy fixable storing previously a GI and FG map, and then calling them while the region render. But FG and GI are both Res. Size and Shader dependent. So the previous "just dgs" render for storing the photons map wont make an accurate approach in the last render.

The solution would be just render the image with its shaders and resolution. (but actually that is what I want to evit!). Can somebody tell me, how to deal with big res. MR images? this is driving me nuts.

And another question, is "GI accuracy 1800 Radius 60" far excessive for a 14 real meters long scene?

fabergambis
12-05-2006, 02:51 PM
You can store your FG or photon map at a very smaller resolution than the final render, something like 400x300 for example; once you stored your maps you can make whatever render regions you want or make an entire 4000x3000 render. Try to use batch render to avoid 'out of memory' problems or check this (http://www.jupiter-jazz.com/wordpress/wp-content/data/tr4kv2/html/index.html) for memory optimization.
Hope it helps.

P.S. I think a value of 512/1024 for GI accuracy it's just enough.

Dogway
12-05-2006, 03:07 PM
well, I should post some tests I did this afternoon but I didnt save thos images. Although I was using GI+FG together, I stored its maps through a 320x240 res render. Then reuse it for a 640x480 res size, and at last a proper 640x480 calculating its maps again. And in these 2 last images there were notable differences, such like banding type or artifacts... in the one with reused maps. Tell me Im wrong and mistook in something, coz i want to think they are not resolution dependent. (That would save my a**) LOL

Anyway, Thanks a lot. Now thats the way Im trying out until I discover something new.

fabergambis
12-05-2006, 04:56 PM
Wich version of Maya are you using? 7.0, 8.0?
You're saying there's a difference between the two maps (FG & Photon) calculated at different resolutions?
and there are differences between the two renderings with frozen FG & Photon maps calculated at different res., right?
Do you freeze both FG & Photon maps?
Because if it's so, it sounds to me really strange...
Once I made some tests with different res. FG maps and I found they were absolutely identical, and it was confirmed by other people who made the same tests.
I'm not sure about photon maps...

Dogway
12-06-2006, 02:24 AM
Ill make some more tests today and post results with both maps.

Dogway
12-06-2006, 05:11 AM
So I just did the tests, and Gi works fine with different resolution. while FG doesnt look to work very nice. Anyway check my settings to see if something might be wrong. Im using Maya 7.0

GI:
Store Reuse New

http://img208.imageshack.us/img208/8452/GIstore.th.jpg (http://img208.imageshack.us/my.php?image=GIstore.jpg) http://img208.imageshack.us/img208/6339/GIReuse.th.jpg (http://img208.imageshack.us/my.php?image=GIReuse.jpg) http://img208.imageshack.us/img208/294/GINew.th.jpg (http://img208.imageshack.us/my.php?image=GINew.jpg)

FG:
Store Reuse New
http://img208.imageshack.us/img208/9880/FGstore.th.jpg (http://img208.imageshack.us/my.php?image=FGstore.jpg) http://img208.imageshack.us/img208/9107/FGreuse.th.jpg (http://img208.imageshack.us/my.php?image=FGreuse.jpg) http://img208.imageshack.us/img208/8360/FGnew.th.jpg (http://img208.imageshack.us/my.php?image=FGnew.jpg)

Settings:
http://img208.imageshack.us/img208/4449/ReuseSettings.th.gif (http://img208.imageshack.us/my.php?image=ReuseSettings.gif)

*Originally Storing Render 320x240, Reuse and New Render 1024x768. But resized for web.

ghostlake114
12-08-2006, 05:33 PM
ya, I dont think FG work fine with different resolution

tfritzsche
12-08-2006, 07:04 PM
Hi Dogway
ticking "view (radii in pixel size)" should fix your problem with FG changing when resolution changes as a starting point use min and max at 25 (http://forums.cgsociety.org/showthread.php?t=365743&page=3&pp=15&highlight=view+radii+pixels), also your GI solution sounds a little to over smooth for a GI/FG render a bit of cloudiness in GI is what you want, then fg will smooth the rest. So tune GI for moderate cloudy solution, then add FG.

good luck
thomas

Dogway
12-09-2006, 07:49 AM
Thanks a bunch tfritzsche.

fabergambis
12-09-2006, 11:15 AM
Forgive me if I insist on that point: I read somewhere that FG maps are something like a 3D map of the scene and this should let you scale this map whatever you want.
Or at least in my experience I store FG maps at a 320 px resolution, then I render the frozen maps at 1280 px and everything works fine, not a pixel changed!
So, what's the problem with your FG map?

Please correct me if I'm wrong.

Dogway
12-09-2006, 02:17 PM
I cant tell you why, by logic it makes sense also to me that if it is a 3D map, its applied in reference to geometry scale and units, and should work, for any res. but anyways it only works with "view radii in pixel" activated, I guess thats the way you are getting it. I tested today an small region, and seems to work, although MR keeps adding more FG points during render :shrug: maybe more FG points density?

Anyone here in the forum can tell you better than me LOL gotta get more reading about it.

Dogway
12-09-2006, 07:06 PM
I have been reading also this thread which is quite related:
http://forums.cgsociety.org/showthread.php?t=365743&page=2&pp=15&highlight=view+radii+pixels
and after a few more tests with view radii pixel, it looks like failing again. I dont understand why while my 320 res render looks clean, reusing its map in the 1024 render causes artifacts, even with radii pixel activated. In the left of the 1024 render I added some more FG points then rerender, and voila! clean. By the way, can I assume that storing FG map through a plain lambert scene would work in the same scene with textures and shaders? (taking to consider color bleeding, etc)

320x240

Diagnostic Render

http://img185.imageshack.us/img185/7360/diagnose320dv4.gif http://img185.imageshack.us/img185/835/render320qx9.gif

1024x768

Diagnostic Render

http://img185.imageshack.us/img185/7575/reuse1024yt8.gif http://img185.imageshack.us/img185/1447/rerender1024freezefp2.jpg

Settings

http://img174.imageshack.us/img174/486/untitled1vi1.gif

*Actually filter was 0. I was just testing it out, and got smooth results but high contrast values between the shadow and highlighted areas. Not too happy :sad:

fabergambis
12-10-2006, 01:03 AM
With both Maya 7 and 8 I had the same results without using pixel for FG radii: lowres FG map>hires rendering ok. But I'm not sure that storing FG map with a Lambert, then assigning new sahders and texturing will work ok. When I was a real beginner with this matter, I made the mistake of storing FG map with a plain white Lambert (or maybe DGS, I can't remember) and when I completed my scene with new shaders, I had some render problems and artifacts so I had to rebuild the FG map with new shaders.
May this help you.

dagon1978
12-10-2006, 02:12 AM
the FG map is a 2d map with a specified resolution
if you freeze it and you render in a different resolution you lose quality

the radius in pixel haven't nothing to do with this

if you use the mray3.4 methods (strict/compatible) a trick could be to use an higher presample density (something like double for a double resolution), freeze GI and FG and make the render region for the raytracing phase
i usually use a lower GI accuracy, but 1800 could be also a good setting, it's strictly scene dependant (not scale dependant)

ciau ;)

mat

fabergambis
12-10-2006, 02:20 PM
Wait! Are we talking about changing Device/Pixel Aspect Ratio with different resolution or am I missing something?
From my experience, I can tell you that computing FG map at 320x240 and rendering the preceiding frozen map at 1024x768 (that is maintaining the same Device/Pixel Aspect Ratio) doesn't make any difference; neither losing quality nor causing artifacts. With or without using FG+GI. How is it?
I made all my latest works this way, and I never had problems.
This thread is challenging my poor knowledge...

P.S. dagon I remember I read some stuff about FG map wich should be some kind of 3D map or spatial map of the scene, but you're telling it's a wrong information. Please can you link something about it? I would like to know...

Dogway
12-10-2006, 08:59 PM
fabergambis, its a FG option found under the AttributeEditor in the miDefaultOptions node, in the Extra Attributes tab. the thread tfritzsche pointed http://forums.cgsociety.org/showthr...ew+radii+pixels was a useful resource.

Im just focusing onto FG which is the problematic in this case. and of course, always using same pixel and device aspect ratio. Im really surprised you can make it work, which is your Maya version? I would really appreciate if you could send me a simple scene where it worked, and the fg map.

Dagon: I have been doing some tests with FG points in that simple scene, where when doubled the res., triple number of points are required (dunno if this works for everything). And in Fg presample density a value of 4 was necessary to get those point numbers. (and in a brief test, took more time, and slight worse results vs 1:1 res. FG store)

something like:


points...........for "" res,............needed density
10000 x3 ......320 x2..................1
30000 x3 ......640 x2..................4
60000 x3 ......1280x960 x2..........16
....
400k~ 4000

Too high Dens. Values tho...


Im a total begginner about FG, but Im starting to think that I need to store the FG map, at output res. with optimizated shaders, or use something like francesca's ctrl_rays, or something about FG+Occlusion that speeds up calculation.

fabergambis
12-10-2006, 11:38 PM
Dogway, the focus of my interest was on which kind of map (3D or 2D) can the FG map be considered...nothing about FG radii in pixel units :) Btw thanks for the link ;)
I'm using Maya 8 but I started to make this 'experiments' (suggested from another maya user) with FG maps , since Maya 7 and never using Presample Density or other tricks...simply it works for me! My process is absolutely simple as I explained before: once I assign the shaders and set the correct lighting, I store my FG map at a very lower resolution (320) than the final render (1280), then freezing the map I can make any kind of experiments on my scene without having to re-calculate my FG map. Only in particular conditions I have to rebuild my FG map, things like lighting or massive shaders replacement, or changing tha Camera position. Recently I make a very intense use of francescaluce's ctrl_rays (God bless her head :love: ) and in a complex scene I'm workin' onto, I had to replace all good old ctrl_shaders with new Arch&Design shaders: I simply replaced the ctrl_shaders in Eye Rays of ctrl_rays with new A&D ones, I rebuilt the map at 320 and I reused it (frozen) for hundreds of tests on the same scene with a resolution of 1280. Everything worked fine as I expected.
Is it possible that I'm the only one who works succesfully this way?

dagon1978
12-11-2006, 02:00 AM
you are not the only one faber, i remeber unhombre at treddi.com using this method many times ago, and i use it too sometimes, but believe me, you lose quality
try two different rendering one with a 320px FG map and one with a 1280px FG map, the final render is not absolutely the same!

and for the 3d map, there is not a simple link, try to render a FG map in a room, freeze it and then try to move the camera in your room, you need to recalculate the map! (at least in the hidden parts of the room)
take a look at the old masterclass (http://www.jupiter-jazz.com/wordpress/wp-content/data/tr4kv2/html/chapter1-FG.html) for more details

fabergambis
12-11-2006, 12:08 PM
From masterclassess: " All this information is cached on disk in a 3D FG map file (temporary held in memory or disk resident) for that specific frame or for bundle of frames."
Maybe that's what was deceiving me.

Yes unhombre was the one who recommended this method, just inviting me to test it to find any difference...I really can't understand! I know if I slightly move my Camera, the map needs to be recalculated (as I mentioned before).
In my last project (still unfinished) I built the same map whether at 320px or 640px (to finally render it at 1280px), and I'm quite sure there wasn't differences between; so you can understand why I decided to use the 320 version...am I blinded?

Exo7
12-11-2006, 02:10 PM
fabergambis,

I think Dagon's point is: do a full 1280 fg render and compare... but at the end dogway is right, it's a matter of density not resolution I believe.

And the fg map is the intersection of a 2d projection (density grid) onto a 3d scene (for computation), back to 2d (camera space) to get an image.

Dogway, have you done more tests regarding density ? Logically when doubling the pic rez you should compute 4 times more rays to reach the same quality, resolution wise. Just curious, as i'm often...

Then I think you can compute a 1000 fg rays map with a res of, say, 4x3 pixels :cool: then use it against a higer rez render... why not ?

Dogway
12-11-2006, 03:10 PM
Exo7: I also tried, adding points through density by regions(FG off), as another enhacing solution, but results are the same as computing regions with FG on, that means different color bleeding and smooth approaching.
You know u also can just render a FG pass to save rendering time, but you will use to want to check results sometimes.

As a saving time factor, Im thinking to turn off bumps (wich use to cause "angle between normal blablabla" message, stated by francesca), and glossy reflections. Add FG points to reflected areas or problematic areas (to avoid "adding 1000 more points.." at render time), and render an Occlusion Pass for bumps and contact areas. I guess that what fabergambis and unhombre do, can be ok sometimes, when working also with GI, a textured scene (not a plain color artifact sensible shader), and the rez. difference is not that much.

You can also Re-add FG points to the entire scene at once, to a previous lower rez. preview stored map. Its useful coz normally we all use to do preview renderings, to check different settings. These FG can be reused as a base to save time, something like filling the lacking new rez. FG points.
well, youve got this way, or just changing density according to desired rez. for add calculation. ( I think that that x4 system should work for any scene, but still not tested)

Im just making my own theorys as far as I know. I hope it can be useful and understandable to the community, thats my apportation.:bounce:


*FGoff can also remove old geometry points? does it make such a big change to leave them so?

dagon1978
12-11-2006, 07:45 PM
From masterclassess: " All this information is cached on disk in a 3D FG map file (temporary held in memory or disk resident) for that specific frame or for bundle of frames."
Maybe that's what was deceiving me.

Yes unhombre was the one who recommended this method, just inviting me to test it to find any difference...I really can't understand! I know if I slightly move my Camera, the map needs to be recalculated (as I mentioned before).
In my last project (still unfinished) I built the same map whether at 320px or 640px (to finally render it at 1280px), and I'm quite sure there wasn't differences between; so you can understand why I decided to use the 320 version...am I blinded?

it's just a matter of words
i mean, the FG is camera dependant (and resolution dependant)
in an adimensional map (like photon map) the time to shoot the rays (photons) is the same for any resolutions, this isn't the case of FG

fabergambis
12-14-2006, 11:54 AM
Sorry for all the fancies I said before:blush:
After some tests I have to agree upon what dagon says :bowdown: : calculating the FG map at the same res. of the final render makes some differences, especially for occlusion shadows when using mia_material.

CGTalk Moderation
12-14-2006, 11:54 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.