Unlimited Detail

i watched that video muted so i don't know what he was talking about, just dropping in to say point clouds are cool.

i'm currently exploring lo-fi 3D scanning + depth of field + stop motion + anaglyphs and having a blast:

(9.5 MB animated gif)

left: blue - right: red

BOOBS!

  On 3/15/2010 at 4:03 PM, phling said:

i watched that video muted so i don't know what he was talking about, just dropping in to say point clouds are cool.

 

i'm currently exploring lo-fi 3D scanning + depth of field + stop motion + anaglyphs and having a blast:

 

 

(9.5 MB animated gif)

 

left: blue - right: red

Oh god , that is fucking weird

  On 3/11/2010 at 10:00 PM, Joyrex said:

There's still texture resolution as well.

well it would eliminate the need for bump mapping, and bitmaps could just be mapped to change the colour of each atom thingy? i think it could work if you could just convert high-polygon models to this, to save on rendering time, but maybe that process in itself would take ages anyway?

can't see it working for gaming, but maybe it could put super detailed CGI creation into the home, without the need for supercomputers and 2 months rendering?

rule 34 applies.

  On 3/15/2010 at 4:03 PM, phling said:

i watched that video muted so i don't know what he was talking about, just dropping in to say point clouds are cool.

 

i'm currently exploring lo-fi 3D scanning + depth of field + stop motion + anaglyphs and having a blast:

 

 

(9.5 MB animated gif)

 

left: blue - right: red

cool! what are you using for the 3D scanning? LASER BEAMS?

nope i don't have lasers, but that's the standard way of doing 3d scans… Radiohead's House Of Cards video is done with some fancy laser setup for example. i'm using a standard dlp projector, swiping a white line over the subject, and filming with one camera. then calculating the depth information from video stills. going to do a write up soon…

it's really lo-fi, there's also another process called "structured light scanning" using something called Grey Codes, which would be more accurate and also faster, but the way i do it the distortions are nicer to look at and it also has a stronger connection to film… it's almost like an anaolgue process done digitally lol.

to get back to the topic (though phling's vid is super cool)...

can't they just make the fucking polygons smaller? seems like an easy solution to me.

  On 3/15/2010 at 7:02 PM, Franklin said:

to get back to the topic (though phling's vid is super cool)...

 

can't they just make the fucking polygons smaller? seems like an easy solution to me.

uh, i think that's exactly how polygon graphics work. by making the polygons smaller, they increase the polygon count and make more rounder objects. or am i missing something.

well polygons after all are made out of triangles, which are simply planes between 3 points in space… it's a question of whether showing the points themselves or the space between them, but in the end the overall concepts are not too far apart from each other. i'd reckon a good thing would be to use a tight integration of the two, polygonal textures for straight surfaces and voxels for more fuzzy stuff… and then, how someone pointed out already, smart LODs...

well that solves that problem.

this is the way i see it. it doesnt matter about the shape of the things they use to compose images... only that they fit together easily and can be made small. beyond certain sizes we can't detect the fucking difference. I dunno what that dude in the video is so concerned about. they could be little tiny elephant shapes, who cares.

in fact I move that we demand tiny elephant shapes.

There's a big difference in how a voxel image gets rendered and how a triangle gets rendered. Polygons get rendered by rasterisation which is extremely fast. It maps the geometry to pixels and shaders take care of the look of the surface. Voxel scenes are basically just raytraced.

This is a really nice introduction on how to do Voxels in realtime:

http://www.flipcode.com/archives/Realtime_Voxel_Landscape_Engines-Part_1_Introduction.shtml

You can also just look for Siggraph papers which are often extremely in depth. There's loads of info on Voxel research that has gone way beyond the stuff this "Unlimited Detail" thing shows.

Just check the work of this nvidia guy:

http://www.tml.tkk.fi/~samuli/

He made a very cool open-source demo showing what you can do with voxels by putting them into a tree data structure.

http://code.google.com/p/efficient-sparse-voxel-octrees/

Anyway that animation thing is extremely cool phling. You have some space I can see more of your stuff?

DUDE. PHLING'S GIF RULS. I WANT MORE

Anyway, I read about the rest of this the other day and it seems kinda wack. You guys have seen Jon Olick (id Software)'s SIGGRAPH demo of a ray-casting SVO, right? If not, check it out, it's on YouTube. He zooms into like, the character's nostril.

But basically yeah, the info is totally out there if you want to read about it. I'm a layperson but SVO's seem dope and mad interesting. However the devil is in the implementation (as far as I can tell), which is why my money is on John Carmack and not this wackjob.

^^ Ego interesting… and thanks guys!

i'm don't know much about 3d engines, rendering etc… just slowly stumbling into 3d territory with my current project.

the point cloud thingie i'm working on is unfinished, but i'll post it once it's done… it actually has more impact in realtime with user interaction than as a video…

here's some older stuff: http://vimeo.com/user1603427/videos and http://www.youtube.com/user/plangoplang

the 3d point cloud scanning is a progression of the video 'nacktscanner' on the vimeo page.

in the mean time here's a little proof of concept sketch: http://www.openprocessing.org/visuals/?visualID=8253

it proves that point clouds are cool.

the graphics on the video looked like a 12 year old pc game to me

WE NEED THIS TECHNOLOGY IF WE EVER WANT TO ACHIEVE FOGLETS