Jump to content

  • Log In with Google      Sign In   
  • Create Account

[Theory] Unraveling the Unlimited Detail plausibility


Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.

  • You cannot reply to this topic
168 replies to this topic

#41 szecs   Members   -  Reputation: 2185

Like
2Likes
Like

Posted 09 August 2011 - 11:52 PM

Way2lazy2care you are simply trolling now.

I went to my bathroom this morning. Wooden furniture with marmoreal counter. Smooth as ice. Tiles: flowers and a noise like pattern. Smooth as ice. Surface-plate: also marmoreal. Not as smooth as ice, but the pattern has nothing to do with the surface topography.

Maybe you are color-blind, but the cliff you've shown has many more colors than one or two. I see shades of yellowish-brown, somewhere more yellow, somewhere more red, even some green too. With transitions. How would you deal with those transitions? Lots of color data?

Even the building example has shades of the brick color, which has nothing to do with lighting.

Or are you suggesting that with should really go back to atomic scale with accurate physical simulation of the atoms? I remember a thread about it....

Sponsor:

#42 Syranide   Members   -  Reputation: 375

Like
1Likes
Like

Posted 10 August 2011 - 03:49 AM

Firstly, I was talking at the sub-meter level as we were talking about not being able to use a parent node's colors. There's no reason the majority of dirt nodes have to have a unique color. The majority of them are just brownish orange. You can still have children that are their own unique color, but most of them can just be the same orangish brown with the majority of the interest coming from shadow and light differences over the surface.

...

How many of the voxels in a model of this bank would just use the same salmon color? Sure there are places like what I am guessing is bird poo over the sign, but those are easily stored in voxels containing color data while all their salmon neighbors just have to sit there and exist.


It seems like you don't really appreciate the difference between shades of a single color, and small variations of a single color. To demonstrate, I took your mountain and approximated it with a single color.

http://imageshack.us/photo/my-images/52/mountxr.jpg/

First image is the reference, the second is the same but with a single color applied... however, I would be seriously impressed if you manage to get shadows that look anyway as good as that, in realtime, in UD.
Does it look like a mountain, sure it does, does it look like a good mountain, no it does not, the lack of nuance and variation makes it look dull. And you are forgetting that, while at a distance, things may look rather even in color, but up-close there is a lot more and important color variations going on. Additionally, if you do note bake lighting into the voxels, then you need to store MORE DATA, unless you want everything shaded using the same method and same parameters, which rarely looks very neat when trying to render realistic scenes.

Quite simply, I don't buy your argument unless you show me that it actually works.



#43 way2lazy2care   Members   -  Reputation: 782

Like
0Likes
Like

Posted 10 August 2011 - 06:25 AM

cut


Did you read what I said? I didn't say it was all one color. I said there was one color that could easily be reused throughout the entire SVO. That does not mean that every voxel is the same color, just that the majority of them could be the same color. I even went into detail specifically calling out areas that are different colors and how you'd handle those as well. It doesnt even mean a voxel is the same color as it's neighbors. All it means is that many of the voxels can inherit their color from their parent.

I mean it's pretty obvious I saw the colors as I address them specifically in my post. Is it that hard to actually read what I said before going, "OHPICTURES! I KNOW EVERYTHING HE SAID NOW!

edit: This is really just like applying a sort of volumetric RLE using the existing SVO.

#44 A Brain in a Vat   Members   -  Reputation: 313

Like
0Likes
Like

Posted 10 August 2011 - 08:17 AM

*facepalm*

Anyway, moving on. Guys, I found some tech that's related to what we're talking about. Don't know if anyone's seen it, check it out.

#45 way2lazy2care   Members   -  Reputation: 782

Like
-1Likes
Like

Posted 10 August 2011 - 08:26 AM

*facepalm*

Anyway, moving on. Guys, I found some tech that's related to what we're talking about. Don't know if anyone's seen it, check it out.


that's the same link in the op.

#46 Tachikoma   Members   -  Reputation: 552

Like
2Likes
Like

Posted 10 August 2011 - 10:56 AM

Most surfaces are monochromatic. At the level of detail we are talking about almost everything is entirely monochromatic. You don't need to change the color of ground that severely when you can add visual interest by just adding a pothole or tire tracks to the actual geometry. All you have to do is walk around. All the walls in my apartment are exactly the same color but get all their different values from the light they take in. Same with most of the chairs and other fabric. Even the wood doors have relatively huge bands of monochrome when you're comparing them to the voxel. The coat hangers are all the same matte metal. The knives and kitchen utencils are 3/4s metal and the other quarter is all black. The lamp behind me is monochrome matte metal even though it has details in the metal with a monochrome lamp shade. The TV is entirely the same glossy black. The speakers around the room are all solid matte black. The vents are all single color matte white.

I disagree. There is a reason why Gouraud and Phong shading models look horribly fake. If you do spectral analysis of materials, you will find that various parts of materials exhibit different spectral reflectivity. Those differences might be subtle in some cases, but they can be quite striking as far as realism is concerned. Not only that, but there is anisotropy as well. Try rendering velvet with your proposal. Brushed metal. Diffraction grating. Sub surface scattering. Even grains of wood on your door will look different at various viewing angles. To achieve those effects, you will need a hell of a lot more information than just a solid colour.
Latest project: Sideways Racing on the iPad

#47 Sirisian   Crossbones+   -  Reputation: 1793

Like
2Likes
Like

Posted 10 August 2011 - 11:26 AM

I disagree. There is a reason why Gouraud and Phong shading models look horribly fake. If you do spectral analysis of materials, you will find that various parts of materials exhibit different spectral reflectivity. Those differences might be subtle in some cases, but they can be quite striking as far as realism is concerned. Not only that, but there is anisotropy as well. Try rendering velvet with your proposal. Brushed metal. Diffraction grating. Sub surface scattering. Even grains of wood on your door will look different at various viewing angles. To achieve those effects, you will need a hell of a lot more information than just a solid colour.

You realize rendering what you described through a volumetric object is much easier than triangles right? The concept though is pretty easy. way2lazy2care already explained the parent inheritance system. This applies for all metadata in the data structure. If you define translucency for instance on a skin voxel the parent can define that and then be overridden on lower levels to turn it on or tweak its parameters. Also brushed metal and other such surfaces inherit their specular and diffusion attributes in the same way. The detail and subtle rendering effects can be achieved by recursive linking of children nodes to higher parents. (This allows a grainy surface to appear grainy even as you zoom in). Not to mention noise perturbation techniques and then thousands of possible procedural routes to generate detail where none is defined. (I wish more people studied noise functions).

I agree that you do need a lot of data. It's surprisingly similar to the amount of data needed for a regular triangle based object in fact. Most people overlook that though and assume that a different data encoding would need to store more. :unsure: In any case there's still a lot of research to be done on the subject and a lot of bias toward non-triangle based graphics since we've been using them for years.

#48 D_Tr   Members   -  Reputation: 362

Like
0Likes
Like

Posted 10 August 2011 - 01:09 PM

The chances that Euclideon is not a scam are very very very slim (I just don't like being 100% absolute). I find the tone of the narration in the videos and the claims totally ridiculous.... Other projects such as the Atomontage engine are actually serious and realistic.

#49 Sirisian   Crossbones+   -  Reputation: 1793

Like
0Likes
Like

Posted 10 August 2011 - 09:36 PM



#50 Hodgman   Moderators   -  Reputation: 31964

Like
3Likes
Like

Posted 11 August 2011 - 01:12 AM

You say the technology has unlimited power?
Umm, yes, yes we do.
We have a search algorithm [which] grabs one atom for every pixel on the screen. So if you do it that way, you end up being able to have unlimited geometry.

When he says "search algorithm", he's obviously referring to their spatial acceleration structure, such as an SVO.
What he's implying here, is that their data structure has a computational complexity of O(P), where P is the number of pixels rendered.
So, if you're rendering a single pixel, that's a structure with complexity of O(1). Not O(N) (where N is the amount of geometry), or O(K.N), or O(log(N)), or O(sqrt(N)), no... O(1).

They've got an acceleration data-structure where the search time is unrelated to the amount of data being searched, whatsoever. A search within 10KB of data is the exact same complexity as a search within 10PB of data.

They've apparently proved that there is such a thing as a free lunch, which isn't just a revolution for computer graphics, but for computer science in general. Google should be buying his company and patenting this discovery. Seriously, it's that much of a big deal.

People were claiming that you must have some sort of memory limitations?
Umm. No. The simple answer is: no. Our memory compaction is going remarkably well.

So not only do they have O(1) search on unlimited data, they've also got infinite compression ratios on unlimited data.
There are no memory limitations at all; they'll just compress infinity into finite space.

And he wonders why people are having a bad reaction to his presentation? He wonders why he's being called a liar when he's saying things that can't be true?

The video JPEG'ed the poor thing.
We know what LOD'ing is; level of distance.
Most people don't really know what tesselation is ... it means the polygons .... have information about how high they are, [which] is used to break them up into little polygons and create bumps. Tesselation bumpy map.
...if you want to put it on a Game Boy DS or something like that, you've got to rebuild all the graphics.
Hello my name is John, I do *raises eyebrows* data compaction. I take *raises eyebrows* atoms, and I smash them *raises eyebrows* with a sledgehammer, until they submit to me.

W.T.F.


He goes on a lot about how great it is to import graphics from the real world -- but this is a red herring. He mentions their elephant a lot, how it was scanned into a 500k poly model, and then converted the polygonal model to 'atoms' for rendering... which means the scanning stuff is in no way related to their rendering tech - it's just as applicable to polygon renderers.

He also completely misinterprets Carmack and Notch's objections, and creates a false dichotomy between their statements. That was just painful to watch.

#51 Sirisian   Crossbones+   -  Reputation: 1793

Like
0Likes
Like

Posted 11 August 2011 - 01:30 AM

I don't see why you find this hard to believe. I mean they said it themselves using the CPU they only get 15-25 FPS. He also mentioned the algorithm requires sorting. I'm not sure if they is because of the separate geometry or what. In SVO rendering with multiple objects you do end up using something similar to DX11 linked list for each pixel to sort the order of things. However they clearly state they don't fire rays and that they're not using raycasting.

Only a few people are having a bad reaction. If you check 85% of people liked the video. I'd imagine the rest are extremely skeptical as they should be. I mean I myself can't think of an algorithm other than raycasting to get the data back to the screen. :mellow: No wonder they're staying so secret. I mean it must be something no one has figured out before or something someone figured out and threw away because it had some problem that only this guy solved.

He also completely misinterprets Carmack and Notch's objections, and creates a false dichotomy between their statements.

I believe Carmack's reasoning is probably because they have people at ID doing SVO research. If he thought Euclideon was using an SVO renderer then he had reason to make the claims he did. The fact is it's probably not an SVO renderer so Carmack's assumption regarding his own technology is probably off-target. The same can be said for Notch who was seriously off on a lot of points in his posts regarding memory limitations. (Assuming they're storing the inside of the model and the surface in a naive format is kind of sad).

#52 Hodgman   Moderators   -  Reputation: 31964

Like
3Likes
Like

Posted 11 August 2011 - 01:43 AM

I don't see why you find this hard to believe. I mean they said it themselves using the CPU they only get 15-25 FPS.

What?

I believe that they can render their stuff at 15 FPS as they demonstrated.

I don't believe in the literal interpretation of his statements, which are nonsense. Do you beleive that infinity:1 compression ratios are possible, or that O(1) search on unlimited data is possible?
If their island demo contained unique geometry, instead of the same instanced pieces, would it still run on that laptop? Dell implies that there are no restrictions here...

The complexity of the search will turn out to be related to the amount of data in some way.
There will turn out to be memory limitations -- you can't fit unlimited data in finite space.

I believe Carmack's reasoning is probably because they have people at ID doing SVO research. If he thought Euclideon was using an SVO renderer then he had reason to make the claims he did. The fact is it's probably not an SVO renderer so Carmack's assumption regarding his own technology is probably off-target.

Carmack said it wouldn't be ready for a game for a few years.
If it's running at 15FPS, then it's not ready for a game. If it's using 100% of the CPU for rendering, then it's not ready for a game. Carmack was simply saying it will take them a few years to fix those issues, which is probably true.
However, Dell interpreted this observation of facts as if Carmack was saying that it's impossible to make a game with their tech...Posted Image

There's no way to know if their acceleration structure is a SVO or not, but he does say that they're using voxels... while also admitting he has no idea what a voxel is.

As for notch, he's right to call Dell a 'snake oil salesman', even if his product is real. As above, he's making claims that contradict computer science (e.g. ∞:1), and making many other correct-but-misleading claims, which very fairly makes it snake-oil - a "product with exaggerated marketing but unverifiable quality or benefit".

In the video, Dell interpret's Notches view as one of "it's not special, everyone is doing it", and then dismisses this view by comparing it with Carmack's *supposed* view that "it's impossible". He also dismissed atomontage by saying it always demonstrates small, constrained scenes (which btw, contain unique and modifiable data instead of instanced data), and at the same time reveals that UD doesn't support real-time destruction/modification of the scene.

If he were honest about it's strengths/weaknesses, didn't try to confuse the issue by merging unrelated ideas with his tech, and honestly assessed competing approaches, they'd be fine. If he did all that, and also didn't speak like a condescending douche, and actually described his approach using real terminology, he'd be someone to look up to and respect.
But as is, he's just a snake oil salesman.

Looking on LinkedIn, they do actually employ at least 3 real games programmers (people with backgrounds in the industry) and a professional director (the kind of guy that directs 3 companies at once and has a work history of senior management roles) on their pay-roll, so they are acting like a real company trying to commercialize a real product. ...but that doesn't change the fact that he's currently made their real product into snake oil.

#53 Digitalfragment   Members   -  Reputation: 875

Like
2Likes
Like

Posted 11 August 2011 - 01:46 AM

I don't see why you find this hard to believe. I mean they said it themselves using the CPU they only get 15-25 FPS. He also mentioned the algorithm requires sorting.


If this sorting has to happen based on where the camera is, where the objects are or whether something is animated, then that cost in reality is part of the search algorithm. So it can't be O(1) as Hodgman pointed out.

#54 _moagstar_   Members   -  Reputation: 465

Like
2Likes
Like

Posted 11 August 2011 - 03:56 AM

Is it just me or does anyone else find that Dell guy to be a complete douche-bag? Obviously I don't know him or his background (apart from the fact he worked in a supermarket), but from hearing his explanation of 'level of distance' :-\ I would say he is a charlatan who has no clue what he is talking about. Anyway....

He goes on a lot about how great it is to import graphics from the real world.


I found this particularly strange too, I love his idea that artists 'would go back to more traditional mediums such as clay' :-\ Someone should show him ZBrush! Does he write all his emails by hand too and then scan them in? Because it's so much easier that way! This just demonstrates that he has no idea about a real world content pipeline.

I have to admit though that the demo does look good, there's lots of repetition but the demo does look nicely polished. If it was here on image of the day I would be very impressed. What is less impressive is the claim that they can handle 'unlimited' detail. Anyone with half a brain knows that this claim is impossible, there are always limitations, and it's trying to get the best effect while working within these limitations that makes programming so much fun! He could save himself a lot of criticism by dropping the 'unlimited' part.

#55 D_Tr   Members   -  Reputation: 362

Like
2Likes
Like

Posted 11 August 2011 - 04:33 AM

Is it just me or does anyone else find that Dell guy to be a complete douche-bag? Obviously I don't know him or his background (apart from the fact he worked in a supermarket), but from hearing his explanation of 'level of distance' :-\ I would say he is a charlatan who has no clue what he is talking about. Anyway....


Oh yes he is. His arrogance is beyond belief... He is probably betting on getting someone to buy his startup and then get away with the money. He makes it impossible for me to believe that he may have something of value, even of slightly more value than other research voxel renderers. The way he looks down on polygons and the serious, real work done by other non full-of-BS people just makes me mad...

I would also be VERY impressed if I saw this last demo (or even the older, unpolished ones) on image of the day but the poster would probably not make the claims Bruce Dell does.

#56 teutonicus   Members   -  Reputation: 518

Like
5Likes
Like

Posted 11 August 2011 - 04:36 AM

Is it just me or does anyone else find that Dell guy to be a complete douche-bag? Obviously I don't know him or his background (apart from the fact he worked in a supermarket), but from hearing his explanation of 'level of distance' :-\ I would say he is a charlatan who has no clue what he is talking about. Anyway....


But he's a vegetarian and sponsors a whole orphanage in India.

#57 _moagstar_   Members   -  Reputation: 465

Like
0Likes
Like

Posted 11 August 2011 - 05:03 AM

But he's a vegetarian and sponsors a whole orphanage in India.


Posted Image Shit yeah, maybe his claims are right after all!!

#58 Syranide   Members   -  Reputation: 375

Like
2Likes
Like

Posted 11 August 2011 - 06:28 AM

There is serious research too, without the bullshit.
Which is where I would bet my money, sometime in the future...

http://research.nvidia.com/publication/efficient-sparse-voxel-octrees



#59 way2lazy2care   Members   -  Reputation: 782

Like
0Likes
Like

Posted 11 August 2011 - 06:37 AM

I found this particularly strange too, I love his idea that artists 'would go back to more traditional mediums such as clay' :-\ Someone should show him ZBrush! Does he write all his emails by hand too and then scan them in? Because it's so much easier that way! This just demonstrates that he has no idea about a real world content pipeline.


I'm not 100% sure I agree with this. I was a 3D animator for a while before I switched to programming and I'm still a potter. Modeling in clay is a good amount easier and more natural to do. There's a lot to be said for the tactile feedback of the clay and interacting with a model in 3D instead of interacting with a model through a variety of 2D interfaces which interact with the 3D model.

Not that I think it's realistic that everybody do that, but it's not quite so crazy as it sounds. For things like character busts I could see it being tremendously useful, but the place I'd think it was most useful would be importing architecture.

You can get the whole interior of a building in a relatively short amount of time completely textured and to scale; exterior is a bit trickier, but can still be done in under a day depending on the size and complexity of the building.

#60 __Homer__   Members   -  Reputation: 58

Like
0Likes
Like

Posted 11 August 2011 - 06:49 AM

A voxel at pixel granularity, in screenspace, is a pixel.
Fractal image compression is just that.
Let's not mince terms.
Let's say you can process an image and compress it into a tree of pixels, good for you, now apply that to a 3D space with pixel granularity and you have applied a volumetric texture to your world at one to one granularity, to fit this in memory, that is a very small world, which if we scale up, looks blocky :)
Perhaps cloud tech can help us around the physical limitations imposed by our current generation architecture.
I'm not sold, but I see potential.
Anyway, on a desktop machine, unlimited detail is as impossible as it sounds implausible.
You cannot fit infinity into a finite space.
Sorry.
You can however map the empty and non empty space in a better way than we do perhaps.
I would accept that.
But that is not unlimited detail.
It is a data structure.
In C++, friends have access to your privates.
In ObjAsm, your members are exposed!




Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.



PARTNERS