SGI: Computer Graphics

"Unlimited Detail" - new rendering technology - Page 1

While the claims sound somewhat exaggerated (there's definitely a good amount of puffery in here), if there's even a modicum of truth to this video then it could be the biggest leap forward in computer graphics ever conceived.

It claims to be a major advance on voxel-based graphics that somehow sorts massive volumes of point-cloud data in a super-efficient manner. According to the narrator and the web site, it's entirely software-based.
mmh, "entirely software based" sounds slow somehow

_________________
r-a-c.de
We'll see in 16 months.

_________________
-ks

:Onyx: :Onyx: :Crimson: :O2000: :Fuel: :Octane: :Octane2: :PI: :Indigo: :Indigo: :O2: :O2: :Indigo2: :Indigo2: :Indigo2IMP: :Indigo2IMP: :0300: :Indy: :Indy: :320: :540: :1600SW: :1600SW: :hpserv:

See them all >here<
stuffy british accents!

_________________
:Skywriter:

DECUS Member 368596
I can't help but notice the complete lack of animation in his environment. As one would expect. Without polygons, how can you animate anything? Are they going to recalculate the locations for each point between every frame? And once you start moving things, you have to recache your search data. He also talks about having billions of points in a scene, how can you have billions of points, each which need XYZ information for positioning, and RGB values for color, without requiring huge huge quantities of RAM? His 'one point per pixel' system is going to going alias like crazy and any surface with detail is going to look like an unfiltered texture (I'm certain this video was rendered at a higher resolution than is displayed).

It's an interesting technology and I'm eager to see where it goes, but it has a huge amount of critical technical hurdles to overcome. Though spending half the video explaining why polygons are bad and his technical explanation only goes as far as to say that it's "like google" kinda smacks of a charlatan. I would very much to see them publish a paper on this technology.

_________________
- Jim
:Indigo: :Indigo: :Indy: :Indy: <- signed by The Screensavers :) :Indigo2IMP: (230L) (230L) :540: :1600SW: :1600SW: <- touchscreen :PI: :Octane2:
skywriter wrote:
stuffy british accents!


It makes me cringe when I hear accents like that and realise that's what most people outside the British isles think we all sound like.....

_________________
:Indigo2IMP: [ R10k-195/MaxImpact +TRAMs/896Mb/73Gb-15krpm/G160/6.5.22 ]
:Octane2: [ Dual R12k-400/V8/2Gb/73Gb-15krpm/6.5.29 ]
Megatron-UK wrote:
skywriter wrote:
stuffy british accents!


It makes me cringe when I hear accents like that and realise that's what most people outside the British isles think we all sound like.....

He's putting it on, trying to sound 'posh' for the camera...
Megatron-UK wrote:

Makes me cringe when I hear accents like that and realise that's what most people outside the British isles think we all sound like.....


Nah - there's always the circa-1908 Cockney and absurdly heavy Scottish dialects that everyone thinks of :D .

Oh, and lest I forget, the rugby players from Monty Python.

But then don't all US people sound like either New Yorkers or a caricature of the Louisiana or Texas drawls?

_________________
Damn the torpedoes, full speed ahead!

Systems available for remote access on request.

:Indigo: :Octane: :Indigo2: :Indigo2IMP: :Indy: :PI: :O3x0: :ChallengeL: :O2000R: (single-CM)
SAQ wrote:
Megatron-UK wrote:

Makes me cringe when I hear accents like that and realise that's what most people outside the British isles think we all sound like.....


Nah - there's always the circa-1908 Cockney and absurdly heavy Scottish dialects that everyone thinks of :D .

Oh, and lest I forget, the rugby players from Monty Python.
....


... and of course, the number one culprit; Dick Van Dyke :lol:

_________________
:Indigo2IMP: [ R10k-195/MaxImpact +TRAMs/896Mb/73Gb-15krpm/G160/6.5.22 ]
:Octane2: [ Dual R12k-400/V8/2Gb/73Gb-15krpm/6.5.29 ]
Megatron-UK wrote:
... and of course, the number one culprit; Dick Van Dyke :lol:
What? Just a spoon full of sugar didn't make that medicine go down?

_________________
***********************************************************************
Welcome to ARMLand - 0/0x0d00
running...(sherwood-root 0607201829)
* InfiniteReality/Reality Software, IRIX 6.5 Release *
***********************************************************************
Megatron-UK wrote:
... and of course, the number one culprit; Dick Van Dyke :lol:


I remember an original Tom and Jerry set in as Robin Hood in Sherwood Forest, ie, vicinity of Nottingham. On of the main mouse characters had this dreadful broad cockney accent. Gor blimey, luv a duck.

_________________
:Indy: :Indigo2IMP: :Octane: :Indy: 4xRS6K 2xHP9K 6xSUN 1xDEC 14xMAC 7xPC 2xPS2
On a somewhat related note the mine planning software I use (which used to run on SGI machines ~ 12 years ago) chokes when I try and triangulate more than 1.5 million data points. The points are are actually stored as topography (x,y, elevation) contours (ie. strings) and are obtained from an aerial survey company. Many of these flight companies have switched over to using LIDAR and generate billions of points (point clouds) and provide much higher resolution (detail).

The project I'm working on is approximately 24 km x 12 km and consists of about 26 million points residing in 9 rectangular "tiles". Just displaying these 26 million points on my computer screen (dual quad core Dell T5400 with 4gb of ram) is painful as the disk I/O is slow. The vendor is still a year or more away from having a 64bit software version.

To display a large topography area like this the steps recommended by our mining software vendor is to build a 2D grid of the area and then interpolate the Z elevation into each grid cell thereby avoiding the whole DTM (triangles) model. The datasets I routinely work with are large and always break the software. It took my workstation 9 hours just to build my (small grid size) gridded surface and then the software refused to display the grid.

My question is: Does anyone know of any software which can handle between 1.5 and 3.5 million points (in a single "tile") and create a DTM without crashing?

Next question: I can "thin" the points on each topography line to reduce the number of points however, the algorithm is stupid and thinning can cause adjacent lines to cross. Elevation lines can't cross (they only do this in the real world when there are cliffs and overhangs) as this "breaks" the DTM engine (in every mining or AutoCAD related program I've ever seen or used). Does anyone know of a "smart" line thinning algorithm and looks at the uphill and downhill lines to make sure deleting a point doesn't cause a problem?

I've asked 3 of the big mining software vendors and they all give me the deer in the headslight look and say that might be a useful tool. I've thought about the problem but I'm too busy doing design work to tackle this problem. I keep thinking there must be an algorithm out there somewhere. Most of the code (or ideas) that these mining software vendors use was "borrowed" from other industries and academics and they don't really create much us we as a company fund them.

thegoldbug

_________________
thegoldbug
:ChallengeXL: :ChallengeXL: :ChallengeXL: :OnyxR: :ChallengeL: :ChallengeL: :ChallengeL: :ChallengeL: :Onyx: :O200: :Fuel: :Fuel: :Octane2: :Octane: :Octane: :Octane: :Octane: :Indigo2IMP: :Indigo2IMP: :Indigo2IMP: :Indigo2IMP: :O2: :Indigo2: :Indigo2: :Indigo2: :Indigo2: :320: :320: :320: :320: :Indy: x5 :Indy: x2 :1600SW: x5 :Indigo: x7
Vault L, 6-1200s, Fibre Hub, Dials, Boards, spares, 3 Sun Blade 2000s, Sun Fire V480, 2-Hero 2000 Robots, 5-Hero 1 Robots
Presumably you've looked at this:

http://terrain.cs.duke.edu/

...and found it inadequate?

_________________
Choosing stones, big enough to drag me down...
I haven't checked the link out yet, but isn't this something that the standard LightWave 3D renderer claimed to have or to have been based on? (With some kind of awkward meshing algorithm, that didn't count the actual polygons but the visible edges and so on; I forgot the details).

_________________
:Tezro: :Indigo2: :rx2600:
2:20 "There's been some fantastic demonstrations by ATI and SGI and even nVIDIA when they run millions and some times as much as billions points on point cloud data on very very big supercomputers"

_________________
:Octane2: 2xR12000 400MHz, 4GB RAM, V12 GFX
SGI - the legend will never die!!
If this "search-engine-gfx-algorithm" really works and this is not a fake, then the guys have a code that is worth more than millions.

_________________
:Octane2: 2xR12000 400MHz, 4GB RAM, V12 GFX
SGI - the legend will never die!!
...mmmhhh ...interesting. Looking forward to know more about this project.

_________________
Image
Octane / Dual Head
Some new videos about the "Unlimited Detail" tech here . The claims still sound hard to believe, but if the technology doesn't actually work, one does have to wonder how the images are being generated.

Also, when Comanche came out on the PC in 1993, its Voxel graphics pretty much blew away any polygon-based graphics available at the time, yet it still ran very well on a 486. Computers are now several hundred times faster.

But whatever, I'll believe it when I see a packaged product.
ajerimez wrote:
Also, when Comanche came out on the PC in 1993


The Roland GS MIDI background music in the video is quite 1993 as well :)

_________________
私のホバークラフト は鰻が一杯です。
IRIX Release 4.0.5 IP12 Version 06151813 System V
Copyright 1987-1992 Silicon Graphics, Inc.
All Rights Reserved.
I think the concept is very similar to "Unlimited Bandwidth". Wait and see......

_________________
:Indigo2: :O2000: :ChallengeL: :Indy: :PI: