pouët.net

Euclideon (aka. Unlimited Detail) is at it again

category: general [glöplog]
then i don't speculate with the fact that they have 'industry profs' in-house when it comes to ex-game industry folks. cos i know at least one! ;) and obviously you need realtime 3d visualisation knowledge for rendering this, comparable to game tech. But making interactive scenery (thus games) out of scanned point clouds seems very unfeasible. e.g. how would you animate a car driving through a street? scan the car separately? scan the rotating wheels separately? how do you scan the light, reflection, see-through from different angle, etc effects on the windows? etc etc etc... the data is as static as the Opera house!
Don't know what they want to present here but it looks like a (bad) voxel engine to me.
added on the 2015-05-15 18:32:45 by movAX13h movAX13h
presenting? their product. ofc it's "just" a voxel engine. i can't tell if it's good or bad. i dunno the details of the solution. some chunks of voxels being transferred. i dunno where that decision happens what to transfer when. and the demo is obviously more about transferring the selective data?!? i get there data they get mine. they can have that lil bits. *shrugs*

and there's nothing to change the fact that the snapshots actually have limits in looking good. number of samples/snapshots. the resolution. or voxolution? i dunno what procedural fillers would do with that. or i'd think completely procedurally filled data doesn't have that problems. and... generally it's not my cup of tea... anyway. i'd prefer it triangulated and textured rather. but... that's another game. whatever... it works for what it's worth.
added on the 2015-05-15 19:26:09 by yumeji yumeji
I got two words for you: "This sucks".
added on the 2015-05-15 20:13:22 by xTr1m xTr1m
at last they show something that you can play with it!!!!!!1!!
added on the 2015-05-15 22:33:32 by swapd0 swapd0
Not a great experience:

BB Image
added on the 2015-05-16 05:58:11 by bloodnok bloodnok
It's pretty much exactly what we've been saying it would be since it first appeared. Reminds me of old 90s barely working webpages where you could "visit stonehenge" etc.
The streaming part sorta works though, so that's cool.
added on the 2015-05-16 08:13:53 by BoyC BoyC
The resolution is pretty underwhelming. Good framerate though. Doesn't seem to be anything special really. Figures.
added on the 2015-05-16 09:50:31 by visy visy
It's nice that they have released something. Their voxel rendering looks decent, but yeah, nothing special. Also streaming is nothing special in SVO world, I've seen many better examples out there.
I just hope they will stop with this annoying overselling. It reminds of this silly slogans "Sell Dreams, Not Products" - my blood is boiling whenever I hear somebody is using this humbug (thanks Jobs!). And there is no unlimited detail in this thing either (even in their earlier videos), as obviously high detail != unlimited, so it's all misleading at best.
added on the 2015-05-16 15:01:59 by tomkh tomkh
The detail level pretty much sucks, especially given that they somewhat promised to be on "atom level". These are voxels, nothing special, not really impressed.

Either you present your awesome algorithm in detail and get the credit for it - our you just keep sucking. That's how euclideon does. Other researchers show their stuff with some proper scale (and that one is two years old already).
added on the 2015-05-16 15:27:33 by las las
Quote:
It's pretty much exactly what we've been saying it would be since it first appeared. Reminds me of old 90s barely working webpages where you could "visit stonehenge" etc.

I miss VRML :/
added on the 2015-05-16 15:27:44 by Gargaj Gargaj
*our=or
added on the 2015-05-16 15:27:48 by las las
After all these years. A pointcloud renderer. Like those of the early 2000s. Fantastic. I hope the guy shuts up now and stops making those videos.

Unlimited LOL.
added on the 2015-05-16 22:37:11 by iq iq
well, at least we can see that the detail is infact quiet limited.
wasn't there something along the lines of "going to atomic level"? :)
added on the 2015-05-17 01:51:35 by abductee abductee
And let's not forget, they also suport "animation". And they compare their rendering to the one in modern video games (which sucks, according to them).

So yeah, what a crapload of bullshit.
added on the 2015-05-17 02:28:55 by xTr1m xTr1m
Thread, arise!

Ok, when I first saw that tech years ago I wasn't sure what that voxel thing is, their demos showed some boring fractalish stuff which wasn't very impressive indeed, so I was rather towards the crowd which was laughing it.
But then I saw Smash particle demos which were jawdropping and revolutionary to me. And it was revolutionary indeed because many of you jumped the wagon and made some similar stuff. Then I realized that basically Euclideon is the same thing, I also played with some laser-scanned point cloud realtime renderings and am even more amazed.

The idea/technology itself is fabulous. In future, when hardware resources permit and it evolves, I see it used along with VR to wander distant places. Computer games use seems quite plausible too. It's potentially THE technology to import and render reality in the virtual one (ok, there's also polygon-based one - check out Realities on Steam). Only a complete idiot could be ignorant about the possibilities.

So why are you laughing it so hard? I hope that the only reason are the bold claims of that CEO person? Well that guy's way of marketing is sillish indeed, but then there were a lot (if not most) of technologies in human history which were laying wasted until some looney took them, commercialized and made successful. As you say, voxels/point clouds/particles are here since a long time (a horse in 3DMark2001 is the oldest decent use of them that I remember), but it was Smash in 2010 who turned them into something breathtaking, something that now is a cliche here.
And lets be honest, Euclideon did make a progress in these 5 years. Hardware resources evolve, they have money and that stubborn guy with a vision. I hope they'll make this tech more successful than raytracking, which is another wasted tech that models reality so much more natively than the fakes we are using today.

And then in a decade or two someone will read this thread and laugh at so many otherwise respectable guys...
added on the 2016-04-13 01:18:30 by rutra80 rutra80
Quote:
So why are you laughing it so hard?

Smash never claimed his demos would display unlimited detail.
I will admit years ago I was someone who didn't like the attitudes of most people at the time, calling his old videos fakes. I didn't feel they where impossible, you could see heavy object repeatition, you could imagine it would be possible with very optimized octree or other searching methods, I wasn't feeling like it was a scam in the same way I think a scame is when someone claims he compressed a full HD movie in 1024 bytes. It looked amazing but less probable, despite I didn't liked his arguments about polygon based game and his attitude/voice was annoying. So, I believed. (I also don't like the ease how people these days call everything a scam, like when there is a kickstarter and they call it a scam, even if that's how kickstarter works, people present something that they don't have maybe the skills to provide, they don't intent to just take your money. A scam is more like when you pay 600$ to get iphone and you get a wall brick in a box.)

But watching the web player, I am very dissapointed. I mean, I like the streaming and things made out of 90s looking voxels, I like it as a clunky visualisations, but that feels so laughable now when talking about "unlimited detail" and remembering his own old videos. Maybe the old rendering where a scam in a sense that they where rendering at 1FPS and just presented as 30FPS? Although there was that video on a show where he controlled it with Xbox controller. Maybe we need to download some desktop app with more detailed data rather than stream? Why doesn't he just upload some of his old demos to prove his claims? Maybe because it was a fake, really going under 1FPS with the more detailed data, so we see the light version?

There are better voxel or point cloud engines on the net, atomontage and ken silvermans attempts and some other webbased that look more detailed. After so many years,. yeah I believe the scam claims now :(
added on the 2016-04-13 12:40:05 by Optimus Optimus
What makes me laugh is there was just over 12GB of data for the industrial scene and that was barely 100M^2. If they get more detail, then that means more data, imagine the data for a map like GTAV. XD

Not only that, they assume all empty spaces aren't actually empty and fill them with voxels, that's just wasting MB!
Maybe "Unlimited Detail" is meant to describe their bullshit.
added on the 2016-04-13 13:23:42 by Gargaj Gargaj
They're as persistent as VR.
added on the 2016-04-13 13:28:48 by rp rp
Quote:
I hope they'll make this tech more successful than raytracking, which is another wasted tech that models reality so much more natively than the fakes we are using today.


Not sure what you mean by "raytracking", but path tracing (via ray tracing on polygon-based scenes) is quite typical in offline rendering (see Arnold, Hyperion, Renderman, Manuka) and might be so in realtime stuff in the future.
added on the 2016-04-13 20:31:49 by msqrt msqrt
Ouclideon fails in being dynamic. So its only useful for highres 3d photography.
Euclideon still fails in being realistic looking or efficient.

Very limited use cases and their pr makes the misttale of hyping and misleading in desperation.
added on the 2016-04-14 01:27:20 by ollj ollj
Quote:
Ouclideon fails in being dynamic.

That.
It's one thing that they haven't even mentioned animation yet (besides of 'yeah we have it and it rules but you'll have to wait'), but even the lighting is completely static.
Once their demo fully loads a view direction it looks nice enough, but as soon as the camera moves the illusion breaks because all of the dynamic lighting cues are also baked into the voxels.
added on the 2016-04-14 01:43:00 by BoyC BoyC

login