Unlimited detail anyone

[Devin Johnston]

Has anyone seen this, it's a company who claims they have developed a system of generating unlimited geometric detail using point cloud data and that it can render in almost real time. I've attached their web site and a YouTube video they've made, if it's true it will revolutionize how visualizations are made. It's almost too good to be true though, granted the lighting leaves something to be desired but they say they are working on that. I'd like to hear what people have to say about this.

 

http://www.euclideon.com/home.html

 

[luthias42]

Give it a few months, if we're still hearing something from them then there might be something to it. Looking at what they can do, I'm hoping they can do what they claim.

 

Thanks for finding it.

[AubreyM]

It sounds like a hoax to me. I did watch the YouTube video, that convinced me more than anything else it was BS.

IS the technology possible? Well anything is possible sure, but even if it was real you would need a Cray supercomputer to run it.

 

Your tag line made me chuckle

[Devin Johnston]

These guy's seem to be focused on games right now because that market is so big. The tools they are using to create the models are Max and Maya and most people running those programs aren't using super computers. I remember seeing this about a year ago and people were saying it was a hoax then too, if it is what would be the point?

[AubreyM]

Well whats the point to all the other hoaxes that are out there?

If this were real it would be worth billions to the company that developed it. The company that is developing it as far as I know is not a huge corporation with lots of money behind it. Therefore if I was the one developing it I wouldn't breath a word of it until I had it to the point where I could sell the technology. Otherwise some other billion dollar company would "look into it" and develop it on their own and bring it to market faster.

At the very least if the technology existed and they were the only ones that knew how to do it they would have been bought out already by a larger corporation.

Maybe they are angling to get bought out as the hits on Youtube are going up like a rocket. There were a few hundred more hits in the time it took me to watch the video. So someone is spreading the word.

 

I could be proven wrong and if I am great. It would be cool to do all that. Time will tell one way or another.

[cgrant3d]

I have to admit I'm amazed and enthralled at the possibilities! However - there *has* to be a weakness to this technology (ie animation, reflection, refraction, transparency).

[Brodie Geers]

I haven't had a chance to watch the video yet because it's loading very slowly. However, it seems to me that, since we don't work much with point clouds, revolutionary it will not be. Maybe it'd allow for super high detail trees and grass perhaps but that's not much of an issue now (the issue is typically having a tree that's modeled or textured poorly or render engine limitations rather than not having enough detail).

 

I'm not sure the problem now is that we can't incorporate enough detail (have you seen Tora's ridiculously detailed models?), so I'm not quite clear on what this would add. More likely it'd simply be another option to throw into the mix.

 

-Brodie

[beestee]

Well whats the point to all the other hoaxes that are out there?

If this were real it would be worth billions to the company that developed it. The company that is developing it as far as I know is not a huge corporation with lots of money behind it. Therefore if I was the one developing it I wouldn't breath a word of it until I had it to the point where I could sell the technology. Otherwise some other billion dollar company would "look into it" and develop it on their own and bring it to market faster.

At the very least if the technology existed and they were the only ones that knew how to do it they would have been bought out already by a larger corporation.

Maybe they are angling to get bought out as the hits on Youtube are going up like a rocket. There were a few hundred more hits in the time it took me to watch the video. So someone is spreading the word.

 

I could be proven wrong and if I am great. It would be cool to do all that. Time will tell one way or another.

 

Putting out material like this and keeping the details behind it very cryptic tells me that they are looking for financial support to bring it to a marketable state. There is some competition in this area it seems, atomontage is a similar technology with a lot of information avaialable about themselves and their aim. Some of their videos make note that the rendering is done on a modest laptop, but captured using a camcorder due to the stress it puts on the system affecting framerate.

[Devin Johnston]

I haven't had a chance to watch the video yet because it's loading very slowly. However, it seems to me that, since we don't work much with point clouds, revolutionary it will not be. Maybe it'd allow for super high detail trees and grass perhaps but that's not much of an issue now (the issue is typically having a tree that's modeled or textured poorly or render engine limitations rather than not having enough detail).

 

I'm not sure the problem now is that we can't incorporate enough detail (have you seen Tora's ridiculously detailed models?), so I'm not quite clear on what this would add. More likely it'd simply be another option to throw into the mix.

 

-Brodie

 

We don't work with point clouds now because it's too difficult to do so and model quality is directly linked to how "good" an image looks. This technology would have more of an impact on animations than anything else, anyone who's tried to do a high quality animation quickly finds themselves pushing the limits of what they can fit into a scene. If geometric limits are removed that's going to remove one of the biggest hurtles that exists in the 3d world.

[Brodie Geers]

I guess, I could see how removing geometric restrictions would allow us to place an unlimited amount of high quality 3d trees, grass, bushes, cars, and people which would all be nice. But those assets would need to be created (a small hurdle once the technology is there).

 

I suppose it might also be able to aid in replicating the existing site and buildings as well so long as you could get high quality point cloud data from it which I'd think would be difficult (perhaps taking more time and technology than reconstructing it yourself).

 

I'm not sure what it could do to help add quality to the actual architecture that we're promoting though.

 

-Brodie

[Devin Johnston]

Definitely new content would have to be created because the existing stuff is just too simple in comparison. As for buildings there are lots of tricks used to fake detail like displacement and bump maps that would no longer be necessary. It would be possible to model every brick or strand of carpet in a building but the question is how long would that take and would it look better than what we're already doing? Trees, cars, plants, clouds, people are no brainers they would obviously benefit but the buildings themselves I'm just not sure about.

[DanGrover]

Honestly, the issues I have with rendering our visualisations and other animations is never the volume of geometry, because we already have tricks to keep the footprints down (Proxies, instances, plugins etc). My problem is always things like AA settings, glossy reflection samples, frosted glass - basically shading. It's shading that "holds me back" insomuch as that's really where I have to make compromises now, not geometry. Maybe that's all because those things do, in part, help mask the fact that the geometry is polygonal, but it's still a problem.

 

I think for games this could (Assuming they sort out the lighting and whatnot) have massive implications. For what we do, I'm less sure.

[Amen]

Here is some nice reading.

 

http://www.rockpapershotgun.com/2011/08/02/notch-vs-unlimited-detail/

 

If you wonder who Notch is, then google minecraft.

[Wrender]

I have a very dumb question about material mapping something like this.

 

Obviously, with a polygon, there is a surface to apply a map to. However, with "atoms", there really is no surface. How do you map something like that? For some reason I'm just not getting my head around it...

[ctk111]

I have a very dumb question about material mapping something like this.

 

Obviously, with a polygon, there is a surface to apply a map to. However, with "atoms", there really is no surface. How do you map something like that? For some reason I'm just not getting my head around it...

 

Thanks for sharing. According to the video, the modeling and texturing occurs before the object is converted to a point cloud. Once converted, I believe each point would be assigned a color value. Like the traditional pointillism style of painting.

 

Still not sure how having that many points on screen would be possible. Can any of the available graphics cards handle rendering that many points in real time and with that many frames per second? It seems to run pretty smoothly on the video.

Edited August 8, 2011 by ctk111

[Wrender]

Thanks for sharing. According to the video, the modeling and texturing occurs before the object is converted to a point cloud. Once converted, I believe each point would be assigned a color value. Like the traditional pointillism style of painting.

 

Still not sure how having that many points on screen would be possible. Can any of the available graphics cards handle rendering that many points in real time and with that many frames per second? It seems to run pretty smoothly on the video.

 

This still doesn't make sense to me. How are transparency, reflections, refractions, etc handled? What if you want to change a material post "point-illization"? Beyond that, if the modeling has to be "complete" before breaking down into points, as you suggest, then your model isn't getting refined or detailed, just more infinitesimal geometry describing the same shape. Like, if you took a 72 dpi image, and just changed the resolution to 300 dpi, you still have the exact same image - just with more pixels....

[jophus14]

Wow. I too believe this will be more geared toward the gaming industry, but it will be interesting to see if it ever sees the light of day with larger companies.

[Jonathan Sanchez]

actually, the visualization industry would probably benefit more from this tech as our scenes tend to be more static. What's preventing this tech from the gaming scene, is that it's almost impossible to animate such a huge amount of info. We would need terrabites of memory. But this does work on a simple static scene like the one on the video.