 |
|
Still interested? Cool; chexit:
-
Stable but Responsive Cloth
The paper is impressive by itself, but you really need to see
the videos to realize just how mindbogglingly impressive
skilled motion capture can be. It's like watching plastic mannequines
claw their way up the uncanny valley through sheer force of will
(and tight pants). Cool.
-
Image Based Flow Visualization
Jarke J. van Wijk put together some absolutely dynamite code for
quickly expressing...flow. Flow, as in deforming change in
moving particles -- air, M&M's, whatever. Both
Source Code and an
actual Win32 implementation have been made available. Get it, play with it,
try not to be hypnotized by it.
-
Interactive Multiresolution Hair Modeling and Editing
The closer the virtual actors get to reality, the closer they get
to needing to be primped realistically. I have this great image of a
famous stylist being asked to design the hairstyle of some virtual
character, and insisting -- because heh, why not -- that the only way
this can be done is if he can haptically interact with the character's
locks. The next day, Mr. Kim gets a call...
-
Real-Time 3D Model Acquisition
Yeah!
Hell yeah! I actually applauded as I watched
this video. These guys combined a cheap video projector with
an even cheaper camcorder to create a system for automagically discovering
the geometry of an arbitrary object. Basically, they project parallel
lines against an object, and compare the deviations they receive from
the lines they sent. The deviation becomes the curvature, and poof.
Cheap, straightforward, elegant.
- Geometry Images:
Textures and Vertexes, Vertexes and Textures, two different worlds,
never the two shall meet...right? Wrong. Hoppe et al. figured out
how to cleanly represent a large set of vertexes as the combination
of two sets of textures. The combination is a psychedelic, decidedly
non-random swirl of textures, shapes, and shadows. And, of course,
now that it's a texture, it can be compressed and decompressed using
existing (lossy!) image compression systems like DXTC. What can I say,
I'm a sucker for synasthesia :-) (Side note: Hoppe's behind one
of SIGGRAPH 2001's cooler demos, "Real Time Hatching". Non-photorealistic
rendering is going to be very, very interesting to see over the next
few years -- What Dreams May Come just scratched the surface; with the
hardware required to pull these sorts of stunts dropping in price dramatically,
low-budget, quick-release TV shows could theoretically turn into the new
showcases for cutting edge animations.)
-
DyRT: Dynamic Response Textures for Real Time Deformation Simulation with Graphics Hardware
When I was growing up, I used to play this game on my Color Computer 3 called
The Microscopic Surgeon. Great fun; I was probably the only 9 year old who
regularly dispensed Tetracycline. The idea of the game was that you pilotted
a craft, Innerspace style, through some patients body, dispensing drugs and laser
blasts as required. Certainly some creative license was taken :-), but the drugs,
anatomy, and defibrillation was all quite real. (You had to stop the heart to
swim through it.) I've always wanted a modern remake of that game, though I'm
sure there'd be all these protests about it make kids think taking drugs was cool
or some garbage. Anyway,
seeing some bodily organ eerily and accurately manipulated in realtime
gives me hope.
-
A Lighting Reproduction Approach to Live-Action Compositing
Debevec. So, it turns out that of the many things the human brain
is good at, calculating the precise amount of light that a given
object should emit given a specific amount of global and local
illumination is about at the top of the list. This is one of a
few very powerful reasons why computer generated objects often look
worse than a simple miniature -- we're not programmed to easily detect
minutely detailed objects zoomed to the size of our visual field,
but we can pretty quickly see if their surfaces reflect light wrong.
Anyway, Paul Debevec -- creator of Fiat Lux, one of the more abstract
but interesting films to presage unrecognizable CGI -- is going to some
pretty extensive lengths to accurately light human subjects who will
later be composited into virtual environments. He's...well, check
out
the video. That's one way to do it :-)
And this is just a small sampling of what was out there, and I'm forgetting all sorts
of good stuff. In particular, I need to track down the absolutely hilarious video
of two mice -- utterly synthesized from 40 minutes of video footage -- just barely missing
a couple phone books being dropped on them. It's totally fake but still quite amusing.
Hope you enjoyed what is otherwise a tour of ancient history for the CGI folk out there :-)
Quick note: I want this. I want this now! :-)
|
Access Archives |
|
| Mission
|
|
DoxPara Research exists as a repository for information security analysis, UI theory, and the miscellaneous writings of its founder, Dan Kaminsky. |
| Authorship
|
 |
|