NERDGASM ALERT: Detailed Rendering of CG just got infinately better. The polygon is dead

reply to post by Uncinus


Exactly. Advanced texture mapping techniques are simply mind blowing in terms of the level of detail achieved using such (relatively) modest resources. As the inventor admitted, he hasn't kept tabs on developments over the last decade. That's pretty much suicide for any serious R&D team.

Originally posted by Glyph_D
But only rendering line of sight.

Both types of engine only render line of sight. Voxel engines use an Octree, and polygon engines use BSP (Binary Space Partition). It's not exactly the same thing, but it's equivalent from an efficiency perspective.

Worst case polygonal engine have maybe 10% overdraw with hardware clipping, but that's negligible, and vastly outweighed by other differences between the two methods.

Originally posted by Uncinus


Polygonal engines do the exact same thing with LOD (Level of Detail) and MIP mapping (LOD for textures).

Well those are slightly different than what i have in mind. LOD is multiple stages of detail at given distances. The closer you get a new geometric design is replaced over the old one. this produces popup.

With this system "it may be the case" that the refresh rate of you video feed regulates the transition.

have you every seen the pictures that when zoomed in it reveals an internal layer that produces a new picture that is woven into the outer layer and continues endlessly image after image?

Just by watching the video this appears to be the effect that is shown.

reply to post by Glyph_D


Back face culling. View frustum culling. Occlusion culling. Early Z rejection. Visibility determination and rejection is standard procedure for any non-trivial real time application. Again, there is nothing groundbreaking going on here, it's a field that has received a great deal of research.

Originally posted by john_bmth
reply to post by Uncinus

 

Exactly. Advanced texture mapping techniques are simply mind blowing in terms of the level of detail achieved using such (relatively) modest resources. As the inventor admitted, he hasn't kept tabs on developments over the last decade. That's pretty much suicide for any serious R&D team.

Indeed, and of course by "Texture" here we are not just talking about the image mapped onto the surface, but also normal maps, bump maps, displacement maps, parallax maps, gloss maps, opacity maps - all kinds of things that can affect the appearance and even the geometry of the rendered surface.

Originally posted by Glyph_D

Originally posted by Uncinus


Polygonal engines do the exact same thing with LOD (Level of Detail) and MIP mapping (LOD for textures).

Well those are slightly different than what i have in mind. LOD is multiple stages of detail at given distances. The closer you get a new geometric design is replaced over the old one. this produces popup.

Note if done right - LODs are blended together. You get a smooth transition just the same as the voxel system.

Voxel octrees are essentially LOD layers anyway. Just rather simpler. The next LOD up is always a solid block 8x the size of the lower level. Again you have to blend the transition.

With this system "it may be the case" that the refresh rate of you video feed regulates the transition.

No it's not. Refresh rate just regulates how much detail you can draw. Same in both systems.

have you every seen the pictures that when zoomed in it reveals an internal layer that produces a new picture that is woven into the outer layer and continues endlessly image after image?

Just by watching the video this appears to be the effect that is shown.

That's a fractal. It's a trick. It's not useful because you can't control it. You could do the same thing with procedural geometry in a polygonal engine.

Originally posted by john_bmth
reply to post by Glyph_D

 

Back face culling. View frustum culling. Occlusion culling. Early Z rejection. Visibility determination and rejection is standard procedure for any non-trivial real time application. Again, there is nothing groundbreaking going on here, it's a field that has received a great deal of research.

Yeah thats what im getting at with the algorithms, these are put into place to restrict what can be viewed. However, Im asserting(it may not be the case) that geometry is not calculated because it is not there. Which would free up the load, so it can be allocated elsewhere.

Originally posted by Uncinus

Originally posted by john_bmth
reply to post by Uncinus

 

Exactly. Advanced texture mapping techniques are simply mind blowing in terms of the level of detail achieved using such (relatively) modest resources. As the inventor admitted, he hasn't kept tabs on developments over the last decade. That's pretty much suicide for any serious R&D team.

Indeed, and of course by "Texture" here we are not just talking about the image mapped onto the surface, but also normal maps, bump maps, displacement maps, parallax maps, gloss maps, opacity maps - all kinds of things that can affect the appearance and even the geometry of the rendered surface.

Aye. Even a simple plain: 4 verts, 2 tris, a nice high res texture and some funky fragment tomfoolery and you've got yourself a party that this "unlimited detail" tech is never gonna get invited to. The last decade has seen a quantum leap in advances, to have missed out on those developments is a serious own goal.

Originally posted by Uncinus


Note if done right - LODs are blended together. You get a smooth transition just the same as the voxel system.

Voxel octrees are essentially LOD layers anyway. Just rather simpler. The next LOD up is always a solid block 8x the size of the lower level. Again you have to blend the transition.

The advantage is these other layers are not necessary to build and allows the artist to focus on other things.

Can you imagine a bunch of videos getting flooded to Youtube of kids digitally manipulating videos to make things look how they want, just like they do with photoshop?

Sounds like we are in for one heck of a future!

Can't you see it now? Video and audio evidence would have to go through forensic analysis before it would be allowed in court. As with the video manipulation tools you would be able to position things within the scene with pre-recorded audio to make it sound like it is official.

Not saying it's a bad thing, but think about the people that don't like you and what they could make you look like. They could create you in their image. Think about it.

Originally posted by Glyph_D
Yeah thats what im getting at with the algorithms, these are put into place to restrict what can be viewed. However, Im asserting(it may not be the case) that geometry is not calculated because it is not there. Which would free up the load, so it can be allocated elsewhere.

No. With voxels you actually have MORE geometry, it's just stored and rendered differently. IN both cases you just store geometry of the parts of the world that are potentially visible.

Originally posted by Glyph_D

The advantage is these other layers are not necessary to build and allows the artist to focus on other things.

In a polygonal engine, the other layers can be generated automatically. In older, lower resolution engines, they were sometimes done by hand. But now it's mostly automated.

Voxel rendering does free you up to model whatever way you like, which is an advantage.

Looking good so far, id have to think this may benefit online gaming first since they have the storage and power to handle it.

This is my point of view there are 2 major approaches to level design. positive and negative.

Positive- is in the realm of radiant editor(quake) and hammer editor(half-life). Where the engine is a complete empty canvas to add and construct your world.

the final steps when using these engines are to enclose your world tightly as to avoid leaks. Leaks wreck havok on your optimization of a map.

Negative- This is the UnrealEd (GoW, UT, ect) this editor avoids the possibility of leaks by starting in solid environment, you then continue by subtracting volumes to encase your world.

With both examples you are limited by the polygons you crate with in your map.

it is uncertain which direction this elucideon engine goes, but i would gather that it uses neither. Much like a voxel but differs when dealing with geometric shapes.

In a voxel rendering it renders the outer layer of a shape. Im under the impression this El system renders all points on a given resolution regardless of shape or color.



It Must be stated that these are my opinions. I know you know that but still it must be said.

Originally posted by Timing
Can you imagine a bunch of videos getting flooded to Youtube of kids digitally manipulating videos to make things look how they want, just like they do with photoshop?

People already do that, and this is not going to help.

reply to post by Glyph_D


Euclideon use standard polygon modeling techniques. They just start with very high numbers of polygons, and convert to voxels.

Here's the elephant from their demo. It's just a normal polygonal model, but very high resolution.

[atsimg]http://files.abovetopsecret.com/images/member/253379cb0d7c.jpg[/atsimg]

A polygonal engine would use the same approach, but take this model, and convert it to a lower resolution model, with various surface maps to give the detail.

So there's really no difference in modeling.

reply to post by Uncinus

"We made a search algorithm, but it's a search algorithm that that finds points, so it can quickly grab just one atom for every point on the screen." According to Bruce Dell, it's all about efficiency. "So think about the difference," he says. "If you had all of the points you are seeing on the screen, like in our demo, it's going to take forever. You'll be waiting for a long time. But if you're grabbing only one for every pixel on the screen, then you don't have a trillion dots, you have… well, pick a resolution and do the maths!

"That's the difference. In layman's terms that's how we're doing what we're doing. The workload is so small that at the moment we're running software just fine with real time demonstrations and we're still optimising, because we keep finding more efficient ways to do this."

link

heres a convo with the front man of the development.

as above he states that the work load is relative to the resolution.

reply to post by Glyph_D


He's just describing how voxel rendering with ray casting using sparse voxel octrees works. It's nothing new. He did not invent it. There's loads of other examples of the exact same thing.

www.google.com...

He's just a spin artist, trying to get more funding.

Originally posted by Uncinus
reply to post by Glyph_D

 

He's just describing how voxel rendering with ray casting using sparse voxel octrees works. It's nothing new. He did not invent it. There's loads of other examples of the exact same thing.

i agree it is nothing new. The only new thing i think he has accomplished is optimization of that already present tech.

Originally posted by Glyph_D

Originally posted by Uncinus
reply to post by Glyph_D

 

He's just describing how voxel rendering with ray casting using sparse voxel octrees works. It's nothing new. He did not invent it. There's loads of other examples of the exact same thing.

i agree it is nothing new. The only new thing i think he has accomplished is optimization of that already present tech.

Beat me to it, done before, they merely evolved a technology that had become frozen for a while - good for them and makes sense but it's hell of a lot of data to work with either way.