Point-based shaders for mray

[MaxTarpini]

You may go here to see tha initial development:
http://forums.cgsociety.org/showthread.php?f=87&t=893152

ciao, max

[kiryha]

This looks like miracle:)
Thanks, Max!

[SGIFreak]

very interesting

[MikeBracken]

Thanks for this. I have been following the dev thread. Great work.


Regards,
Mike

[jasonhuang1115]

Thank you, Max. Look forward to its development. :thumbsup:

[Bitter]

Excellent! I suppose this thread can be used for official feedback?

[swag]

very nice - thanks :thumbsup:

[ytsejam1976]

Grazie max. :bowdown:

[Gabba]

Can I pray some gods to see a XSI port or is a waste of time? :p

[haggi]

Great work! Thanks for sharing.

[MaxTarpini]

Can I pray some gods to see a XSI port or is a waste of time? :pI don't know, - that's up to you. Frankly I'd just try to port the .mi to .spdl and then eventually pray whatever you want that it's working ;)

ciao, max

[ristopuukko]

Thank you Max !

/risto

[Sorath]

awesome, thanks max!

[cavemen]

Finally great work! max ... will definitely test it ..
thanks a lot max...

[jb-alvarado]

AO Baking in Max and Softimage runs perfect :).
Thanks for this nice shader!

[Daniel-B]

Awesome. How would one impliment these techniques in 3ds max?

[cavemen]

Where can I find mix8Layer Shader dont seem to find it on my hdd had it a couple of months ago.. can someone reupload it ...
is there a mix8layer shader for maya 2011 64bit anywhere

[jb-alvarado]

Awesome. How would one impliment these techniques in 3ds max?

Use a mental ray material, connect the PCloud_Shade shader in the surface:
edit the Filepath (for example c:\pcloud.mrmap)
edit the attr string (for example ao)
set the Lockup and the multipler to 1

In the lightshader slot connect the PCloud_Bake shader:

enable bake
use the same path string from the shade shader
use the same string from the shade attr parameter
connect the ao shader in der sample shader slot
set the pre-bake fg parameter to 1

Now render the image and after turn off the bake to PCloud parameter. Render again...

Edit: the mesh must have fine details/smal subdiv ploys

[MaxTarpini]

cool, you mean you're testing that on 3dsmax, and it works ?!

By baking AO you mean baking raytraced AO to pclouds, right ?
I'm just refining right now the point-based AO
edit: I mean to be able to do color bleeding. :)


cheers, max

[jb-alvarado]

Yes I test it in 3ds max, and its work :). At the moment I only can render AO Cloud. I also test FG with an mia_material, but I only become a white image.

Yes I use the normal ambient occlusion shader (mib_occlusion).


edit: I mean to be able to do color bleeding. :)


Oh that's very nice!

[MaxTarpini]

Yes I test it in 3ds max, and its work :). At the moment I only can render AO Cloud. I also test FG with an mia_material, but I only become a white image.
Yes I use the normal ambient occlusion shader (mib_occlusion). OK, great I'll support 3dsmax as soon as the second part of shader pack is out.

To not go for extreme tesselations, you may try the 'interpolation' parameter on PCloud_Shader, set it to 4-12 (based on how many poly you have) and see if results are not too heavily blurred out.

edit: for 3dsmax need still to check how object are splitted out at preprocess and if that may require some custom stuff to support mray approximations like in maya. but I have an internal build of ctrl_approx that works with 3dsmax so that won't be a problem eventually, plus in 3dsmax there's better support for mat_phen to wrap up the whole in a single material without the need for explicit connection to lighmap port and so on. however this is a Maya thread.. :)

ciao, max

[jb-alvarado]

This is a old mi shader. You find it in the base.mi. I don't no how you can use it in Maya. In 3ds max you must delete the hidden string.

[MaxTarpini]

@caveman: you may try here:
http://www.box.net/shared/i6hd2271qk#

btw, is the mix8layer a dependancy in the scene I upped ? Thought I removed any other custom shaders..

ciao, max

[cavemen]

where can I find mix8layer shader for maya 2011 64bit ? I tried to download version 1.1 from creative crash but when I open the example files theres no shader connection between the mix8layer and the SG node.. is there any other version of mix8layer out there...

edit @Max

Thanks .. i'll download it asap ...
yah i they are there in ur examples ...

[sevver]

hello
please linux 64 version :)

[SirRender]

This is fantastic Max. I still have a lot to learn about point cloud rendering so please excuse the dumb question.

With your shaders, you no longer need to use a renderman compliant renderer like 3delight? You can do everything within Mental Ray in Maya, including the generation of the point cloud data?

Thanks!

[SebKaine]

Thanks for this Max, this is gold, hope to see a 2010 release !

[gramulho]

This probably is a dumb question, but will you be releasing source code so we can compile on other platforms ? We want to test it out so bad on our macs 64 bits.

Cheers
www.illusiongraphics.com.br

[Bitter]

There are still other reasons to have a Renderman compliant renderer, however, brickmaps has always been one of the major selling points of such a system.

And yeah Max, there are some mix8layers in there. I tend to stick to mib_colormix myself. Helps on a renderfarm with less craziness.

So far the speed is really great. However, I am curious that it computes FG for the cloud and then I seem to get FG again before the frame renders. Is this truly the case of does it cache the original pass for the lightmap generation?

[ndeboar]

I'm a stickler for details, brickmaps and point clouds are two different beasts.

[Bitter]

Implementation

[Walter]

Thanks for this Shader that seem very powerfull !

But I have some dificulties,
I can bake SSS clouds easily to simple scene as Torus with a Area Light,
In a scene with a Bust exported from ZBrush, 3 Area Lights, the result is nothing.
I apply the same workflow as your exemple scenes and my simple test scene, I tweak the parameters, and the PCloud is definitively nothing.
I don't understand.

[MaxTarpini]

Take a look at verbosity, and compare it with the one you have on docs, to see if something went wrong with your pcloud generation or sss computation.

ciao, max

[MaxTarpini]

So far the speed is really great. However, I am curious that it computes FG for the cloud and then I seem to get FG again before the frame renders. Is this truly the case of does it cache the original pass for the lightmap generation?
The FG pre-process before rendering does just lookup the shader you have plugged into 'indirect' to support pseudo color bleeding from the pcloud, there isn't any FG computation on the pcloud itself.

ciao, max

[MaxTarpini]

Ehy kids, what a poor feedback ... you seem more a kind of obsessive-compulsive plugin/shader collectors than anything else ... ain't you ? :)

B2 is out:

- trivial bug on mel templates was causing bad UI for materials with notes attached (e.g. foliage scene).
- threads handles were not properly deleted before their pointer were released, causing a crash on user-abort (e.g. 'esc' key during rendering).
- data framework has been consolidated for upcoming ao and colorbleeding.

max

[Bitter]

Hey Max,

Sorry. :sad: But I'm stuck with a lot of hard surface work right now. I can't really pound the SSS implementation like a character artist. (I'm dealing with vehicles and lighting right now) As soon as the AO/FG/Reflection part is done I can beat the Hell out of them. :twisted:

[Walter]

Take a look at verbosity, and compare it with the one you have on docs, to see if something went wrong with your pcloud generation or sss computation.

ciao, max

OK Max,
The result verbosity wasn't rich as your documentation:
The verbose to yours pcloud nodes was set to Full and Basic.
The result to the Output Window seems light compared to your screenshot,
No additionnal information. Here in the simple torus scene test.
PHEN 0.3 info : --- thread<M>, diffusion at point 55296 = (0.003409 0.000109 0.000068) (p. eval: 289)
PHEN 0.3 info : --- thread<1>, diffusion at point 18432 = (0.016762 0.001538 0.001393) (p. eval: 211)
PHEN 0.3 info : --- thread<2>, diffusion at point 36864 = (0.003578 0.000108 0.000078) (p. eval: 16)
PHEN 0.3 info : --- thread<0>, diffusion at point 0 = (0.006396 0.000166 0.000162) (p. eval: 84)
PHEN 0.3 info : --- thread<0>, diffusion at point 4608 = (0.008981 0.001476 0.001124) (p. eval: 275)
PHEN 0.3 info : --- thread<2>, diffusion at point 41472 = (0.054119 0.014364 0.008373) (p. eval: 11)
PHEN 0.3 info : --- thread<M>, diffusion at point 59904 = (0.068747 0.021736 0.009368) (p. eval: 828)
PHEN 0.3 info : --- thread<1>, diffusion at point 23040 = (0.070747 0.031076 0.011742) (p. eval: 382)
PHEN 0.3 info : --- thread<M>, diffusion at point 64512 = (0.248260 0.124205 0.061052) (p. eval: 148)
PHEN 0.3 info : --- thread<2>, diffusion at point 46080 = (0.003556 0.000352 0.000238) (p. eval: 599)
PHEN 0.3 info : --- thread<0>, diffusion at point 9216 = (0.007599 0.000281 0.000272) (p. eval: 54)
PHEN 0.3 info : --- thread<1>, diffusion at point 27648 = (0.072015 0.027875 0.012286) (p. eval: 723)
PHEN 0.3 info : --- thread<0>, diffusion at point 13824 = (0.182317 0.089684 0.032063) (p. eval: 108)
PHEN 0.3 info : --- thread<M>, diffusion at point 69120 = (0.142745 0.072013 0.030391) (p. eval: 603)
PHEN 0.3 info : --- thread<2>, diffusion at point 50688 = (0.008955 0.001677 0.000756) (p. eval: 6)
PHEN 0.3 info : --- thread<1>, diffusion at point 32256 = (0.066037 0.023059 0.013021) (p. eval: 161)

Thanks for your disponibility !

[MaxTarpini]

You need to set the mray verbosity first. Go to Rendering->Render->RenderCurrentFrame and click the little square to get the option out. You'll see Messages->VerbosityLevel... set it to InfoMessages. Now you have your verbosity.

Isn't this the first time ever you take a look at verbosity right ? :)
That's all right, but 'point clouding' requires some understanding of the whole render process.

From the verbosity chunk you posted there's nothing wrong, you've 4 threads computing effectively SSS. However your object seems not so much tesselated. Do you know how to apply mray approximations to have objects poly count increased at render time ?

Thanks for testing.

ciao, max

[Walter]

I check the Render Verbosity to the mentalrayGlobals tab to 2011,
but the changement isn't effective during the rendering process.
Your method works fine.

For the Torus test scene, I have a mentalray node approximation set to 2.
For the Bust, isn't the higher subdivision level, But I think it's reasonable for rendering time testing.

With checking the verbosity, I saw that the point cloud isn't read during the rendering process, though the Filepath are good.


PHEN 0.6 info : -----------------------------------------------
PHEN 0.6 info : --- Creating SSS point cloud
PHEN 0.6 info : -----------------------------------------------
PHEN 0.6 info : Pass 0: Pre-baking FinalGather
RCFG 0.6 warn 542002: angle between normal and geometry normal > 90 degrees
PHEN 0.6 info : Pass 1: Begin baking ...
PHEN 0.6 info : --- baking point: 0
PHEN 0.6 info : --- baking point: 2048
PHEN 0.6 info : --- baking point: 4096
PHEN 0.6 info : --- baking point: 6144
PHEN 0.9 info : --- baking point: 10560
PHEN 0.9 info : --- baking point: 14080
PHEN 0.9 info : --- baking point: 17600
PHEN 0.9 info : --- baking point: 21120
PHEN 0.3 info : --- baking point: 23808
PHEN 0.3 info : --- baking point: 31744
PHEN 0.3 info : --- baking point: 39680
PHEN 0.3 info : --- baking point: 47616
RCFG 0.2 warn 542002: angle between normal and geometry normal > 90 degrees
PHEN 0.2 info : --- baking point: 107904
PHEN 0.2 info : --- baking point: 161856
PHEN 0.2 info : --- baking point: 215808
PHEN 0.2 info : --- baking point: 269760
PHEN 0.7 info : --- baking point: 287616
PHEN 0.7 info : --- baking point: 328704
PHEN 0.7 info : --- baking point: 369792
PHEN 0.7 info : --- baking point: 410880
RCLM 0.2 info : lightmap rendering statistics
RCLM 0.2 info : type number per eye ray
RCLM 0.2 info : reflection rays 20505384
RCLM 0.2 info : shadow rays 7106
RCLM 0.2 info : environment rays 39875030
RCLM 0.2 info : finalgather rays 44321425
RCLM 0.2 info : probe rays 12807690
RCLM 0.2 info : fg points computed 230657
RCLM 0.2 info : fg points interpolated 437965
RCLM 0.2 info : on average 14.07 finalgather points used per interpolation
PHEN 0.6 info : Pass 2: Cooking data from cache
PHEN 0.6 info : Pass 3: Creating octree
PHEN 0.6 info : Pass 4: Computing scattering at all points
PHEN 0.3 info : --- thread<M>, diffusion at point 325632 = (0.601158 0.285906 0.125426) (p. eval: 311)
PHEN 0.3 info : --- thread<0>, diffusion at point 0 = (0.030414 0.013936 0.006223) (p. eval: 2)
PHEN 0.3 info : --- thread<1>, diffusion at point 108544 = (1.673332 0.940253 0.396880) (p. eval: 21)
PHEN 0.3 info : --- thread<2>, diffusion at point 217088 = (3.235262 1.902902 0.718892) (p. eval: 353)
PHEN 0.3 info : --- thread<0>, diffusion at point 27136 = (1.459230 0.854226 0.431057) (p. eval: 192)
PHEN 0.3 info : --- thread<0>, diffusion at point 54272 = (0.249933 0.089998 0.066874) (p. eval: 1076)
PHEN 0.3 info : --- thread<1>, diffusion at point 135680 = (1.675941 0.996812 0.426218) (p. eval: 23)
PHEN 0.3 info : --- thread<M>, diffusion at point 352768 = (1.572395 0.583500 0.252603) (p. eval: 117)
PHEN 0.3 info : --- thread<0>, diffusion at point 81408 = (1.778651 0.963534 0.391681) (p. eval: 42)
PHEN 0.3 info : --- thread<2>, diffusion at point 244224 = (3.000457 1.782315 0.659664) (p. eval: 756)
PHEN 0.3 info : --- thread<M>, diffusion at point 379904 = (1.718971 1.051748 0.578249) (p. eval: 48)
PHEN 0.3 info : --- thread<1>, diffusion at point 162816 = (1.967723 1.046445 0.443978) (p. eval: 301)
PHEN 0.3 info : --- thread<M>, diffusion at point 407040 = (3.042052 1.852072 0.765453) (p. eval: 754)
PHEN 0.3 info : --- thread<2>, diffusion at point 271360 = (1.460535 0.904687 0.420879) (p. eval: 346)
PHEN 0.3 info : --- thread<1>, diffusion at point 189952 = (0.510898 0.248526 0.151999) (p. eval: 900)
PHEN 0.3 info : --- thread<2>, diffusion at point 298496 = (3.000889 1.728771 0.647825) (p. eval: 152)
PHEN 0.6 info : Pass 5: Writing pcloud with SSS data
RCMP 0.6 info : wallclock 0:00:00.67 for building the BVH for map 2
RCMP 0.6 info : allocated 126 MB, max resident 139 MB
RCMP 0.6 info : writing map file "D:/fighter_sss.mrmap"
PHEN 0.6 info : -----------------------------------------------
RCMP 0.6 info : map 2 stats:
RCMP 0.6 info : global page accesses : 5106 (0.39% of total accesses)
RCMP 0.6 info : page faults : 4348 (0.33% of total accesses)
RCMP 0.6 info : page faults / global accesses : 85.15%
PHEN 0.6 info : -----------------------------------------------
PHEN 0.6 info : mayabase version 10.8, compiled on Jul 13 2010.
PHEN 0.6 info : -----------------------------------------------
RCFG 0.6 info : depth #finalgather points
RCFG 0.6 info : 0 2854

[MaxTarpini]

That's all fine for pcloud generation. While it seems the pcloud reader is not really engaged by mray, because you should have at least the banner in the verbosity... "--- Reading SSS pcloud". What are you using for shading.. pcloud_shade or pcloud_ssscattering ? For the former you should specify 'sss' as the attr map.

Btw, the scenes that comes with the plugin are rendered correctly, right ?

ciao, max

[Walter]

I use pcloud_SSScatering.
I made a test with the pcloud_Shade and it's the same result, the cloud wasn't loaded.
The scenes that comes with the plugin are rendered correctly...

[MaxTarpini]

ok, can you just up somewhere a simplified scene with the same pcloud rig so I'll take a look asap ?

[Walter]

ok, can you just up somewhere a simplified scene with the same pcloud rig so I'll take a look asap ?

I send you a private message with my email adresse.

[MaxTarpini]

Walter I sent you a msg with my email too.

In the while got some time to fix a main problem with former point-based AO implementation.. e.g. the lack of a triangle visibility check which was returning 'white line' artifacts.. that was the same on first point-based AO for Prman and 3delight (other like Pixie may still have that problem).

http://img52.imageshack.us/img52/1773/aocity005notrivis.png
'White line' artifact on cube base.

http://img269.imageshack.us/img269/7554/aocity005yestrivis.png
No artifacts due to triangle visibility check and formfactor implementation. As that's not always necessary it's an optional parameter in the AO bake material.

max

[Bitter]

I might be on crack. . .the glacier scene renders nothing but the sky even after going through the trouble of baking itself.

// Error: (Mayatomr.Shader) : pcloud_SSScattering1SG: no shader assigned in material //

I have installed mix8layers. The setup is a bit. . .complicated. Trying to decide how/why the shaders are setup like this. Which material should be in the above shading group?

And which switch do I throw to reuse the point cloud over and over? Looks like in pcloud_SSScattering1 I keep it set to "evaluate SSS from SSS pcloud" However, I don't want it to bake over and over, so I deselect "bake to cloud" in pcloud_bake_SSScattering1

That correct?

[rasamaya]

awesome work. props.

[MaxTarpini]

Bitter, you can simply use a mib_color_mix and layer reflection on top of diffuse in add mode.
SSS, read notes in the foliage scene, can be used in two flavours.

You may just do all SSS computation on the pcloud and bake SSS data there too. Or you may first bake irradiance (either with pcloud_bake_SSS or pcloud_bake with an 'irrad' channel) and then use the irradiance pcloud as an irradiance cache for SSS lookups at render time (eval SSS at Render Time).

If you wanna do all on pcloud, yep, once you baked up the pcloud, de-check the BakeToPCloud and leave the pcloud SSS reader in default mode (eval SSS from PCloud). Implementation is transparent, you may also use a pcloud_Shade (with as 'sss' channel to just read the SSS). The workflow is outlined in the docs.

Rasamaya, :)

ciao, max

[cavemen]

Hi
All

@Max

I was trying to use this point cloud techinque for rendering occlusion..
Though I can create a ptc file and connect it and it renders fast but still I cant get the right image out...
I have attached three images and the scene file ( the zip also consits of the renderlog and the ptc) ... dono whats going wrong.. is there something that I am missing ?

Please have a look at the attachments...

01 Actual Mentalray Render without ptc baking
http://img259.imageshack.us/img259/3401/mrrenderimagewithoutbak.th.jpg (http://img259.imageshack.us/i/mrrenderimagewithoutbak.jpg/)

02 Baked Ptc with point in lookup in the p_cloud Shader set to 1

http://img101.imageshack.us/img101/9266/pointinlookup01.th.jpg (http://img101.imageshack.us/i/pointinlookup01.jpg/)

03 Baked Ptc with point in lookup in the p_cloud Shader set to 8
http://img713.imageshack.us/img713/1025/pointinlookup08.th.jpg (http://img713.imageshack.us/i/pointinlookup08.jpg/)

04 Baked Ptc with point in lookup in the p_cloud Shader set to 24

http://img695.imageshack.us/img695/3324/pointinlookup24.th.jpg (http://img695.imageshack.us/i/pointinlookup24.jpg/)

This is the link to the scene (http://rapidshare.com/files/418111518/point_cloud_baking.zip)



Regards

[wizzackr]

Hi
All

@Max

I was trying to use this point cloud techinque for rendering occlusion..
Though I can create a ptc file and connect it and it renders fast but still I cant get the right image out...
I have attached three images and the scene file ( the zip also consits of the renderlog and the ptc) ... dono whats going wrong.. is there something that I am missing ?

Please have a look at the attachments...

01 Actual Mentalray Render without ptc baking
http://img259.imageshack.us/img259/3401/mrrenderimagewithoutbak.th.jpg (http://img259.imageshack.us/i/mrrenderimagewithoutbak.jpg/)

02 Baked Ptc with point in lookup in the p_cloud Shader set to 1

http://img101.imageshack.us/img101/9266/pointinlookup01.th.jpg (http://img101.imageshack.us/i/pointinlookup01.jpg/)

03 Baked Ptc with point in lookup in the p_cloud Shader set to 8
http://img713.imageshack.us/img713/1025/pointinlookup08.th.jpg (http://img713.imageshack.us/i/pointinlookup08.jpg/)

04 Baked Ptc with point in lookup in the p_cloud Shader set to 24

http://img695.imageshack.us/img695/3324/pointinlookup24.th.jpg (http://img695.imageshack.us/i/pointinlookup24.jpg/)

This is the link to the scene (http://rapidshare.com/files/418111518/point_cloud_baking.zip)



Regards

Hey cavenman, looks to me like your mesh is not sufficiently subdivided when baking out the pc. Try to attach an mr displacement approximation node to your geometry and hook up an empty displacement node to the SG, leaving the input of the displacement node empty. You will then have a straight forward way of telling mr how to subdivide the mesh at rendertime.

Note that, as far as I know, cranking up the point lookup will not give you any benefit, as that will just interpolate across your - unsufficiently dense - pc. I may be wrong there, though... max?

...speaking of which: Thanks a boatload for your work, max! I'm still testing the shaders and already love them. In some cases it is just fantastic!

[Bitter]

Please look here for future development on the shaders:

Point-based Shaders (http://forum.mentalimages.com/showthread.php?t=6925)

Max will be keeping up from the mental images forum directly.

Ciao!

[lighto]

This is very interesting indeed. I have been using point clouds for renderman so this should be pretty interesting for mentalray.

[cavemen]

@wizzackr

I'll definitely try what you have posted...

@Bitter

thanks a lot for sharing this info with us...

Also I did a speed fan test on my i7 930 proc and during the baking process my cpu core temperatures went up to nasty 75 deg Celcius ... I guess I should buy those gaming coolers for the baking process.. have never seen my core temp go up to that ... Max I remember one of your posts where you said that you feared that you may fry your processor lol I had the same feeling during this test .. however for those faint hearted my processor is fine just if you all are wondering...

[zerogee]

Hi Max,
can you give as any update about the your progress? I am eagerly waiting for a 2009x64 final release.

thanks

G

[Bitter]

Max isn't updating this forum anymore, look a few posts up to see where he is following now.

He has started on ambient occlusion now, there's a new beta on the mental images forum.

[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.