Jump to content
Search In
  • More options...
Find results that contain...
Find results in...
DooM_RO

Unllimited point cloud data

Recommended Posts

Some of you might be asking themselves "wtf is unlimited point cloud data?" while others will say "yeah, I know about it but it's a scam" (I'll get to that later)

"Unlimited Point Cloud Data" is a new technology that promises to bring the next big step in 3D graphics - in short, if developers adopt this new technology "games will look 100,000 times better" as Bruce Dell, the one who invented the engine states.Obviously this is a very big number and he needs to back his statements with solid facts.

So how does it work?The answer is that instead of 3D models, they import them or scan real life objects, smash them into billions of atoms and display only the amount you need(what you see on your screen) using a complex search algorithm(works a bit like google)!Sounds absolutely ludicrous, doesn't it?

Of course, such claims cause a great amount of controversy.Big names like Notch say that it's impossible to do such a thing and it's most likely a scam, but can you really blame him?

A year after his company, Euclideon posted their first presentation on youtube http://www.youtube.com/watch?v=00gAbgBu8R4 they uploaded their first interview in which he proves that the technology is in fact REAL http://www.youtube.com/watch?v=JVB1ayT6Fdc

What are your thoughts on this?I think it's bloody fantastic!

Share this post


Link to post

The last thread about it didn't go too well.

That being said...it's not exactly a scam. Just good old marketing something mundane as novel (see e.g. Creative Labs, Apple, etc.). Infinite points? Had this in 1992, with Fractal compression.

I wonder how none noticed the simple fact (well, Notch from Minecraft did) that if you reuse and reuse the same data over multiple scales, you do create in fact the illusion of infinite precision. Also, if make just one big-ass ultra-detailed tree and repeat it a million times, it may also distract from the fact that, well, it's still ONE fucking tree.

Share this post


Link to post

I'll believe it's real when it can replicate the look of, say, Skyrim. With all its variety of objects and textures. Rather than just a tree, an elephant statue and a rock.

Go on, Euclideon guy. You can do it. Instead of scanning a real-world tree or whatever, just extract all the data files from the game, grab NifTools, and make yourself a nif-mesh-to-point-cloud converter. You'll get a wealth of assets. Then make a video of you wandering around Skyrim in your engine. Don't need to implement gameplay and all that stuff, just, you know, displaying the world with all the objects in it. Maybe having animated models loop their animations endlessly.

Let's see how it fares.

Share this post


Link to post
Gez said:

I'll believe it's real when it can replicate the look of, say, Skyrim. With all its variety of objects and textures. Rather than just a tree, an elephant statue and a rock.

Go on, Euclideon guy. You can do it. Instead of scanning a real-world tree or whatever, just extract all the data files from the game, grab NifTools, and make yourself a nif-mesh-to-point-cloud converter. You'll get a wealth of assets. Then make a video of you wandering around Skyrim in your engine. Don't need to implement gameplay and all that stuff, just, you know, displaying the world with all the objects in it. Maybe having animated models loop their animations endlessly.

Let's see how it fares.


I was actually thinking about an Elder Scrolls game with this tech, it would be immensely epic!In the interview he says that they plan to release a tech demo way before they actually release the SDK(which is in 16 months, he says)

Share this post


Link to post
Gez said:

Let's see how it fares.


He'll probably convert a brick, a branch or somesuch from Skyrim and replicate it a million times with maybe some minor variation like color or light, looking like a Minecraft on steroids ;-)

Also, before lapping on the "infinite point cloud"'s doubtfully tasty dick, have you considered watching any 64KB or 4KB demos? Those too are capable of generating seemingly "infinite" detail, and have been doing that for years.



Share this post


Link to post

VOXATRON IS THE FIRST STEP OF THE NEW FUTURE!




Seriously: Voxatron is a promising voxel-based game. It has a lot of potential. Sadly, there are a lot of people who expect realistic graphic and physics in the same time, which is hard to make with voxels...or at least, it is not worth for the biggest companies (Electronic Arts, Valve, etc.)

Share this post


Link to post
DooM_RO said:

So how does it work?The answer is that instead of 3D models, they import them or scan real life objects, smash them into billions of atoms and display only the amount you need


Had this in 3DMark01. I think they were called "point sprites"



The problem isn't displaying them, nor optimizing their display (who would use a display algorithm that doesn't optimize away what you don't see, anyway?)

DooM_RO said:

Of course, such claims cause a great amount of controversy.Big names like Notch say that it's impossible to do such a thing and it's most likely a scam, but can you really blame him?


What Notch and other critics actually say is that it's not a memory-efficient way for representing arbitrary 3D objects, and that Euclideon tech conveniently handwaved this incovenient detail all along.

The fact that their demos show the same piece of terrain and the same, single non-animated tree replicated again and again suggests that they either:

  1. Were too lazy to add more (so that makes them look bad)
  2. The technology is too early
  3. Adding objects is too labor and memory intensive
  4. Animating objects is too labor and memory/CPU intensive
The fact that they haven't addressed those points in a YEAR doesn't help their case. And the burden of proof rests not upon Notch or anyone else but upon Euclideon tech alone.

Share this post


Link to post

Seems like this is actually using that octree stuff Carmack was talking about. I also wonder how they would implement lighting.

Also, the video of Elevated that Maes posted is amazing. One of my favorite demos.

Share this post


Link to post

@Maes I don't know about the other stuff but in the interview he shows that animation was in for 7 years, you should watch it, it will clarify a lot of things.Oh yeah and carmack said it could be useful in a few years https://twitter.com/#!/ID_AA_Carmack/statuses/98127398683422720

Share this post


Link to post

Don't waste your energy cheerleading for this. If it's useful technology, then it will eventually be adopted. If it continues to look like ass, then it's useless and not worth paying attention to.

Share this post


Link to post
DooM_RO said:

I don't know about the other stuff but in the interview he shows that animation was in for 7 years, you should watch it, it will clarify a lot of things.

Alright, so since you skipped the previous thread: I've spoken to the guy about it several years back, and a few people I've worked with for many years are now at Euclideon working on the tech.

You know what that animation is that's been in for 7 years? It's the equivalent of sprite animation. As in, they've created their data in keyframes, and then display those keyframes without any interpolation. Which gives thoroughly jerky results. Animating such data efficiently is best approached in the same sense as streaming a movie efficiently. Of course, being a voxel-esque tech, it's always going to be behind what an optimised poly pipeline will produce in terms of visual fidelity on consumer hardware.

And yeah, the guy is just as much a dick in real life as he is in the videos, but if you remove all his hyperbole and get to what the tech actually entails the idea is sound enough for next-gen hardware when used for the specific purpose of terrain. I hope it'll actually get to the point of useful now that it's receiving Australian government funding (especially since it means it'll keep my friends in a job).

Share this post


Link to post

Not sure if this is legit or not but if it is, I have some serious doubts about this unlimited detail shit

Share this post


Link to post
Maes said:

Also, before lapping on the "infinite point cloud"'s doubtfully tasty dick, have you considered watching any 64KB or 4KB demos? Those too are capable of generating seemingly "infinite" detail, and have been doing that for years.


These guys apparently have a nifty tool for creating demos and games with realtime algorithm-generated data:
http://www.theprodukkt.com/theprodukkt

Here's one of their 64k demos (from a decade ago):
http://www.pouet.net/prod.php?which=1221

Share this post


Link to post

@GooberMan

Really?Well then can't devs just combine this tech with polys?for instance, they make a tree using this tech, with branches and roots and the rest(leaves etc) do it the traditional way.Characters could be made with polygons so that they can be easily animated.This would be awesome for static environments!

In the interview, he also said that the demo only uses software mode at the moment and that they haven't touched the video card.Can't they just use the CPU for rendering and the GPU to handle stuff such as physics, animation and particles?(or at least animation)

Of course, my knowledge about programming etc. is rather limited so don't jump at me if I say something stupid :), I'm just saying that if the engine uses only the CPU in its current state and runs at reasonable frame rates, the GPU could be used for something else.

I imagine that this is going to really strain current hardware but they said that they will finish the engine in about two years, which means that if a game dev licenses the tech it will take about four years to develop a game.Wouldn't hardware six years from now be capable to do this?

Share this post


Link to post

Traditional triangulation will continue to be the preferred method of rendering 3D simulations until voxels, point clouds, or other rendering methods can solve the animation problem. A bunch of faceted triangles can be easily manipulated to allow for complex and subtle animation, whether it's motion capture or a response to physics. Trying to do the same thing with a series of points is extremely difficult and the results are usually ugly, unnatural and unrealistic.

Share this post


Link to post
hex11 said:

These guys apparently have a nifty tool for creating demos and games with realtime algorithm-generated data:
http://www.theprodukkt.com/theprodukkt

Here's one of their 64k demos (from a decade ago):
http://www.pouet.net/prod.php?which=1221


I remember seeing "ultra-detailed" generated worlds as early as in 1994 4KB demos. The problem is NOT performing display culling with octrees or whatever: it would be like bragging that Doom is so cool because it uses "binary space partition trees" so it doesn't have to overdraw stuff.

The problem is a) where the fuck do you store all the Goddamn data and b) how the hell do you manipulate it in real-time in a way that doesn't suck.

Those intros at least provide an answer to a) with techniques such as splines, fractal compression etc. and in general, ad-hoc compression and texture/object generation/replication depending on what they wish to show. This is labor-intensive on the programmer's side, and unless something can be efficiently compressed/represented with mathematical formulas, it's not a good candidate for such demos.

Euclideon's tech provides an answer to neither: they are essentially praying that one day PCs will have enough memory to store all of their billions arbitrary data points they would need to even approximate the visual variety of a modern game. It's pointless to be able to draw them very efficiently if you can't store enough of them to begin with, and if you sacrifice resolution with "point based data", then you'll end up with a fugly voxel world made of Lego bricks rather than "unlimited detail".

Share this post


Link to post
DooM_RO said:

Well then can't devs just combine this tech with polys?

They could, but that wouldn't work out well for the total revolution Bruce is preaching, wouldn't it?

DooM_RO said:

Can't they just use the CPU for rendering and the GPU to handle stuff such as physics, animation and particles?(or at least animation)

Or they could implement it in OpenCL/CUDA and let the video card do all the rendering work again. I'm not in the console game at the moment so I don't know what kind of pre-release devkits are floating out there (if any) and have no idea what the hardware actually entails, but I'd reckon they're waiting till they get some next gen hardware to play with before they attempt that.

Share this post


Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×