PDA

View Full Version : Stag Beetle



Elmar Moelzer
02-08-2010, 06:46 AM
http://www.mediastudio-graz.com/volumedic/images/pic/vm_beetle.jpg

http://www.mediastudio-graz.com/volumedic/images/pic/vm_beetle2.jpg

This is a volume rendering of a stag beetle dataset provided by the technical university of Vienna. It was rendered using the VoluMedic volume rendering plugin for LightWave.
The treetrunk is made using simple geometry and bump displacement. The whole project took less than a day to do and much of that time was spent on doing lighting, etc.
Rendertimes were rather long for these stills, due to the use of lots of area lights and DOF which required a lot of AA. It is worth noting that the polygonal treetrunk and the volumetric beetle take roughly the same amount of rendertime.
Let me know what you think!

Cougar12dk
02-08-2010, 07:52 AM
Very good Elmar :)

So the beetle is all volumes? It's been quite a while since I looked at volumetric, so I don't quite remember how it works :)

alesxander
02-08-2010, 03:50 PM
wow, amazing work

congratulations

alesxander,
http://alesxander.com

probiner
02-08-2010, 04:20 PM
The second one is really cool. Nice render.

jameswillmott
02-08-2010, 04:45 PM
Wow, that's very cool!

Cageman
02-08-2010, 05:06 PM
Howly cow! Well done! I would never have guessed that the beetle was volumerender!

Pretty damn cool! :thumbsup:

Elmar Moelzer
02-09-2010, 10:39 AM
Thanks for the comments everyone. Yes the beetle is completely volumetric. There is not a single polygon on it(you could count the Null Object as a single point though).
The only polygons in the scene are in the branch thingy that the beetle is standing on.

Lightwolf
02-09-2010, 10:54 AM
Wow, wicked stuff indeed. What is the resolution of the volumetric data?

Just densities?

Cheers,
Mike

Elmar Moelzer
02-09-2010, 12:03 PM
This was a free dataset provided by the TU Vienna (RAW Data). The resolution was 832x832x484 (quoting that from memory right now, so it might be off a bit somewhere, but in that ballpark). It had 8 bit depth greyscale.
It is rather nice resolution actually. It is even possible to see some inner details if you wish. Here is a link to a video that I made. It is in false colors, to better show the inner workings.
http://www.mediastudio-graz.com/volumedic/html/pic/vm_beetle_video.htm

Lightwolf
02-09-2010, 12:14 PM
This was a free dataset provided by the TU Vienna (RAW Data). The resolution was 832x832x484 (quoting that from memory right now, so it might be off a bit somewhere, but in that ballpark). It had 8 bit depth greyscale.

Wow, neat. I would have thought the resolution is higher. Well done on the rendering algorithms, it looks extremely smooth!

Cheers,
Mike

Elmar Moelzer
02-09-2010, 12:33 PM
Thanks for the kind words Mike.
This is actualy on the higher side as much as datasets are concerned. Usually datasets are arround ~512x512x250 or so. This one was nicely high res for a change :)
We do pretty well with low res datasets too.

The embryo here is extremely low res (94x164x101):
http://www.mediastudio-graz.com/volumedic/html/pic/vm_mouse.htm

Philbert
02-09-2010, 02:55 PM
I'm confused. Is it like a voxel sculpture in 3D-Coat? Was is "modeled" in LightWave or somewhere else?

Lightwolf
02-09-2010, 03:02 PM
I'm confused. Is it like a voxel sculpture in 3D-Coat? Was is "modeled" in LightWave or somewhere else?
It wasn't modelled. It's been created using a (presumably) medical MRT scanner. Or at least something similar.

Cheers,
Mike

Lightwolf
02-09-2010, 03:03 PM
We do pretty well with low res datasets too.

The embryo here is extremely low res (94x164x101):
http://www.mediastudio-graz.com/volumedic/html/pic/vm_mouse.htm
*clap clap* Quality work (and I especially mean the code ;) )!

Cheers,
Mike

Bog
02-09-2010, 03:08 PM
A couple of years back I opined (possibly in a bar, just possibly) that holodeck technology would far more likely come from volumetric imaging than from polygonal or spatches.

Thanks for proving me right, Elmar! That's just plain smashing.

mav3rick
02-09-2010, 03:09 PM
holly shiat.... this is wicked... elmar you mastered volumedic how about fluid simulation system for lw now?

Skittixch
02-09-2010, 03:10 PM
*edit* wow, lots of posts between when I started my reply to philbert...carry on!

I'm not familiar with the exact format of volumedic data, but it's not "sculpted" at all. It's probably scanned using tools that I'll never be in the same room with. I think it's the same concept as a grid based volumetric fluid container(think maya fluids, or dynamite for LW) taken to an absurd resolution. The forms produced are based on thresholds of density in any given sample inside the volume container. think of a 3d matrix of density data (as well as other surface attributes) ranging from 0-1.

it's essentially the raster equivalent of 3d technology, where we're all used to vector/poly based models thus far.

Awesome stuff! :D

Lewis
02-09-2010, 03:12 PM
That's excellent stuff Elmar, it looks truly remarkable. Do you have bigger resolution renders and what was the rendertime :)?

Philbert
02-09-2010, 04:31 PM
It wasn't modelled. It's been created using a (presumably) medical MRT scanner. Or at least something similar.


Oh I see. It was confusing since the 3DC voxel mode was also called volumetric sculpting at one point.

Elmar Moelzer
02-10-2010, 01:26 AM
Skittixch, you got it pretty right, but one thing. Never say never when it comes to this equipment. Just like 3d Scanners came down in price over the years, there are now table top CT- scanners that cost only a fraction of what they used to cost.
Give it a few more years and that tech is within every studios reach.
Only MRI scanners will stay more expensive for another while. That is mainly due to the complicated super conducting magnets (which need a fresh supply of liquid helium). They may come down too though once high temperature super conductors become broadly available.

For now you can make use of the fact that quite a few universities have CT- scanners and in my experience they are always quite helpful.
Plus, there are more and more free resources on the web. MRI scanners and scans are harder to come by though.

Btw, VoluMedic 2.0 does not only allow you to segment the dataset by density. We also have a 3d- magic wand like tool now that allows you to create masks for the dataset. Plus our Volume Painting allows you to literally paint things away that you dont like (or paint in stuff that is not there).

On the fluid system: We leave that field to others. There is currently a system in the works called "Bukadan". It has been talked about on various boards. That will be a lot cheaper than VoluMedic too.
We have been thinking about supporting various mathematical models and procedurals as inputs though. So far we have not had the time to work on that though. One reason was that we were part of the development of a "serious" coloscopy product, which just came to the market a couple of months ago (we did the realtime volume rendering part).

Rendertimes for the 720p rendering were about 35 minutes/frame. The main issue was that lazy me did not spend any time on optimizing the lighting (5 area lights) and the branch geometry (tons of polys with displacement). The DOF and motion blur did not help either. A single pass renders in 5 mins, but the adaptive sampling takes a good time. I did not have much time to optimize the scene though. As I said, I spent less than a day on this project, before I had to move on to doing some sales and marketing related stuff (so much less fun, but oh well) ;)

Lewis
02-10-2010, 02:03 AM
35min per frame at 720p is not bad at all (unless it's on soem super computer with 32 cores :)) . I'm dealing with 1,5-2 hours per frame on i7 at 3.8Ghz every day (IES/Night renders, hundreds of lights, lot of trees/shadows...) so 35min per frame would look like breeze for me :) :).

sampei
02-10-2010, 06:19 AM
interesting stuff !

Elmar Moelzer
02-10-2010, 11:58 AM
35min per frame at 720p is not bad at all (unless it's on soem super computer with 32 cores )
It is a fairly standard Q9450 quadcore.


I'm dealing with 1,5-2 hours per frame on i7 at 3.8Ghz every day (IES/Night renders, hundreds of lights, lot of trees/shadows...) so 35min per frame would look like breeze for me .

Nice setup!
The scene would probably fly on that one :)
A i7 920 is already 40% faster than my machine, so that one gotta fly.
Lighttypes and numbers really make a huge difference on rendertimes in LW. The same scene would probably take a few seconds to render with a single distant light.
In fact I have had scenes render at the standard res (no AA) in fractions of seconds with raytraced shadows on even (though all settings optimized for speed instead of quality).

precedia
02-10-2010, 08:08 PM
The treetrunk is made using simple geometry and bump displacement.

I really like the bark geometry and texture. Can you share your textures and treetrunk model? I've failed to create tree bark that looks as good as yours.

Elmar Moelzer
02-11-2010, 12:58 PM
The bark texture was actually from one of NewTeks free texture collections, if I remember correctly:
http://www.newtek.com/freestuff/index.php
The branch geometry is laughably simple. It is basically a slightly tapered and bent Subpatch cylinder. I subdivided it a few times, froze the Subpataches, then ran a small jitter on it to make it more uneven. That object I then Subpatched again and in Layout I applied said bark texture in the bump channel. I then set a Bump Displacement in the Object properties and that was it. I think I also added some procedural in the color channel to give it slightly more variation and to make it a bit less pale (it was very bright otherwise) and added the same bark texture into the Diffuse Channel with a low opacity (again to darken the whole thing a bit).

Edit: Yupp, I just checked, it is texture collection 1. You can sign in there with your Forum ID and Password.