tb's vids aren't showing up for me, so I may be being totally redundant.
Any time I've had to do projection or video mapping type stuff, I've had to go about it very "manually". There aren't any quick fixes for getting the correct rotations, etc. that I'm aware of; I've just had to make a replicated version of whatever I'm projecting onto, in QC (yeah, that sounds like major fun, and really easy, doesn't it ? :) There are caveats to that if you have other hardware to work with, where maybe it could be a tiny bit easier, but not much.
edit-Ok, I see tb's vid's. It will be interesting to see what Matt's referring to, b/c we have pretty different interpretations :) BTW, that stuff is some of the coolest EVER toneburst. I'm totally blown away by how cool those look.
gtoledo3 wrote: edit-Ok, I see tb's vid's. It will be interesting to see what Matt's referring to, b/c we have pretty different interpretations :) BTW, that stuff is some of the coolest EVER toneburst. I'm totally blown away by how cool those look.
Thanks, man
I've had a thing for these kind of abstract-looking volume rendered diagrams, ever since leafing through books on cosmology as a kid.
I think there's some cross-purposes talking going on here though. I suspect mattgolsen meant something else entirely. Matt?
I've loved volumetric stuff since seeing this back in 98: http://www.pouet.net/prod.php?which=255 (youtube link on the right for the modern age, realtime version recommended if you happen to have a tricked-out DOS machine lying around... ;)
There's lots of ways to render volumetric stuff, none of them are really suited to QC unfortunately. A few I've come across:
Layered 2d images. Each layer is a section of the volume data. You need tons of fillrate for anything complex...
Voxels. Each point in the volume data gets rendered separately (usually with a rectangle or a cube). 100 x 100 x 100 resolution data = 1,000,000 points, or 1 million iterations with QCs famously slow iterator ;)
3d textures. Not supported in QC, unless that's changed with openCL/4.0.
Projecting an image onto a physical wiremap of thread or some other kind of medium. TB's stuff is similar in concept I think, but onto an actual physical surface.
dont know how the wire folks did it but the way i did it was with a glsl shader.
i first measured the projector (how big is the image at a certain distance). then i wrote a shader thet would calculate a 3d position for each pixel from its x,y value and a depth (distance of the object from the projector). the x,y is already given but the depth you have to provide yourself. in my case, the projector was aligned so the image fell flat on the paper stripes and since the stripes are straight (like the wire) i just gave it the depth at y=0 as a 1d image. i made the image manually with a composition i made to help me with that.
using the measurement of the projector i calculated the "eye" position and then made a vector from there going through each pixel, normalized it and multiplied with the depth. the result is the 3d position of that pixel:). well, since i only gave it the z at y=0 i also had to make it calculate the z at the other y values. but thats not hard though if the stripes are perpendicular to the projector .
What I'm thinking of doing, and I'm not even sure if it's possible is having 8 projectors point into a rectangular space, and having a projection of the same 3d model but from different viewpoints, giving an illusion of a 3 dimensional object. Although I imagine it wouldn't be anywhere near seamless.
Really this is all complete conjecture, I'm just wondering if it's even possible.
mattgolsen wrote: Very cool stuff, thanks for the information.
What I'm thinking of doing, and I'm not even sure if it's possible is having 8 projectors point into a rectangular space, and having a projection of the same 3d model but from different viewpoints, giving an illusion of a 3 dimensional object.
Is this a correct logical assumption (mathematically)? The projectors must project light onto a surface that reflects the light to the observers eye in order for the image to be seen by a viewer. Two options for this surface:
Either: (a) surface is a 3D shape already (like say a cone) and the projected image(s) are providing nothing more than a texture which may create the illusion of a 3D shape that appears different to the shape being projected onto (cone appears like a cylinder tilted so as to appear to be receding in deep perspective to a point at infinite z coord)or
(b) surface is flat (like the floor which is visible from all points in the room) and the image projected will be no more 3D than any other movie. Multiple projects will just layer images onto the texture and some may be brighter from some vantage points than others (angle of incidence to angle of reflection differences) but estentially like billboards multiple Billboards with Add rendering intent.
Unless you have invented some kind of holographic movie projector a la certain robot and a princess, I can't see how it's possible. May have missed the aim of the exercise though ;)
Edit:
Okay just watched the link, wires in space to create 3D pixel grid. Clever. Can't see how having 8 projectors will help, 2 at most. He is getting image at about 90deg to incidence (ie perpendicular). Wonder if hitting the wires with a colour laser would be more effective. Certainly more accurate and no focusing issues. Not that I have a colour laser lying around…
Yeah, that kinda stuff is a different case. Note it uses a single projection source and (essentially) infinite reflection plains. Have only seen these on www would be cool to see one for real.
Dunno if this is the kind of thing you had in mind. I've done some other experiments in ray-casting for rendering isosurfaces.
a|x
tb's vids aren't showing up for me, so I may be being totally redundant.
Any time I've had to do projection or video mapping type stuff, I've had to go about it very "manually". There aren't any quick fixes for getting the correct rotations, etc. that I'm aware of; I've just had to make a replicated version of whatever I'm projecting onto, in QC (yeah, that sounds like major fun, and really easy, doesn't it ? :) There are caveats to that if you have other hardware to work with, where maybe it could be a tiny bit easier, but not much.
edit-Ok, I see tb's vid's. It will be interesting to see what Matt's referring to, b/c we have pretty different interpretations :) BTW, that stuff is some of the coolest EVER toneburst. I'm totally blown away by how cool those look.
Thanks, man
I've had a thing for these kind of abstract-looking volume rendered diagrams, ever since leafing through books on cosmology as a kid.
I think there's some cross-purposes talking going on here though. I suspect mattgolsen meant something else entirely. Matt?
a|x
I've loved volumetric stuff since seeing this back in 98: http://www.pouet.net/prod.php?which=255 (youtube link on the right for the modern age, realtime version recommended if you happen to have a tricked-out DOS machine lying around... ;)
There's lots of ways to render volumetric stuff, none of them are really suited to QC unfortunately. A few I've come across:
Layered 2d images. Each layer is a section of the volume data. You need tons of fillrate for anything complex...
Voxels. Each point in the volume data gets rendered separately (usually with a rectangle or a cube). 100 x 100 x 100 resolution data = 1,000,000 points, or 1 million iterations with QCs famously slow iterator ;)
3d textures. Not supported in QC, unless that's changed with openCL/4.0.
Sorry I have a tendency to not type everything I have bouncing around in my head :D
I was referring to something like this:
http://www.vimeo.com/1495617
Projecting an image onto a physical wiremap of thread or some other kind of medium. TB's stuff is similar in concept I think, but onto an actual physical surface.
Oh.. I seeeee.....
Not at all what I had in mind, in that case. Sorry for the false lead.
a|x
this is very similar to what we did in k4 once http://kineme.net/Applications/AudioVisualizations/vjinstalationK4klublj....
dont know how the wire folks did it but the way i did it was with a glsl shader.
i first measured the projector (how big is the image at a certain distance). then i wrote a shader thet would calculate a 3d position for each pixel from its x,y value and a depth (distance of the object from the projector). the x,y is already given but the depth you have to provide yourself. in my case, the projector was aligned so the image fell flat on the paper stripes and since the stripes are straight (like the wire) i just gave it the depth at y=0 as a 1d image. i made the image manually with a composition i made to help me with that.
using the measurement of the projector i calculated the "eye" position and then made a vector from there going through each pixel, normalized it and multiplied with the depth. the result is the 3d position of that pixel:). well, since i only gave it the z at y=0 i also had to make it calculate the z at the other y values. but thats not hard though if the stripes are perpendicular to the projector .
Very cool stuff, thanks for the information.
What I'm thinking of doing, and I'm not even sure if it's possible is having 8 projectors point into a rectangular space, and having a projection of the same 3d model but from different viewpoints, giving an illusion of a 3 dimensional object. Although I imagine it wouldn't be anywhere near seamless.
Really this is all complete conjecture, I'm just wondering if it's even possible.
Is this a correct logical assumption (mathematically)? The projectors must project light onto a surface that reflects the light to the observers eye in order for the image to be seen by a viewer. Two options for this surface:
Either:
(a) surface is a 3D shape already (like say a cone) and the projected image(s) are providing nothing more than a texture which may create the illusion of a 3D shape that appears different to the shape being projected onto (cone appears like a cylinder tilted so as to appear to be receding in deep perspective to a point at infinite z coord)or
(b) surface is flat (like the floor which is visible from all points in the room) and the image projected will be no more 3D than any other movie. Multiple projects will just layer images onto the texture and some may be brighter from some vantage points than others (angle of incidence to angle of reflection differences) but estentially like billboards multiple Billboards with Add rendering intent.
Unless you have invented some kind of holographic movie projector a la certain robot and a princess, I can't see how it's possible. May have missed the aim of the exercise though ;)
Edit: Okay just watched the link, wires in space to create 3D pixel grid. Clever. Can't see how having 8 projectors will help, 2 at most. He is getting image at about 90deg to incidence (ie perpendicular). Wonder if hitting the wires with a colour laser would be more effective. Certainly more accurate and no focusing issues. Not that I have a colour laser lying around…
I'm thinking about giving something like this a shot:
http://gl.ict.usc.edu/Research/3DDisplay/
Yeah, that kinda stuff is a different case. Note it uses a single projection source and (essentially) infinite reflection plains. Have only seen these on www would be cool to see one for real.
"Bring up the plans of the Death Star, Ensign…"