|How graphics people summarize their work
||[Aug. 5th, 2005|12:11 pm]
I'm home from SIGGRAPH. I expect I'll post about some of the highlights in another day or so, but in the meantime, here's something I've been sitting on all week for lack of time to edit it into shape...|
The evening of the first day of the conference is devoted to something called Fast Forward. All of the paper authors line up along one side of the room. They are given 50 seconds apiece to present synopses of their papers, plus 10 seconds for the next author to get up to the podium. After the 50 seconds, a bell rings to tell the speaker to stop Right Then. Well, usually a bell rings. A few speakers were informed that their time was up when a cow mooed enthusiastically at them.
In the course of about two hours you get an overview of a hundred papers or so. (This year it was 98.) Here are a few of the things that were said that night...
(on a paper about extracting good chromakey mattes on hard-to-focus stuff like curly hair) Step 1: I called together a few of my friends. You'll notice that most of my friends either have straight hair or very little hair, which leads me to step 2: Put my friends in embarrassing wigs.
There just isn't enough quantum mechanics in SIGGRAPH nowadays, so I decided to render these video sequences onto a synthetic retina... one... single... photon... at a time.
Come and hear the details at 10:30. Come to think of it, we're the last paper. 11:15 will probably do. (SIGGRAPH is pretty good about keeping schedules, but papers sessions in general are notorious for slipping later and later as the day goes on.)
Or what if you have lots of high-speed video sitting around of darts being shot into pyramids of marshmallows? (This paper was about the dynamics of non-rigid objects)
So then we added some continuum mechanics and tricky math and ended up with the world's first entirely mesh-free bubblegum. (Meshless deformation and fracture)
But of course, in the end, the real reason for this work is that we want to play around with bunnies. (Bunnies with Stanford pedigrees, no less.)
Okay, you can have many numbers on slides, and usually they don't mean anything...
(slide of woman on beach in bikini) But they're important when we want to render our, our, uh, the environment around us...
Plastic-looking hamburgers don't whet your appetite? Do you really want fries that look like painted wood to go with it?
Title of one slide on motion search: Three Stick Guys in a Rotary Sushi Bar
For example, we can translate the style of a normal walk into that of a sneaky ninja.
The system can learn to identify graduate student motion as natural, even when they try to dance.
In this work we present gamma-ton tracing, a quick and dirty way to ruin the world. (about simulations of weathering and tarnishing)
The result is very strange moving drawings. Also you can play with your teddy bear. Thank you very much.
Okay, so we have this nice camel head and we want to make it look a lot more interesting.
The evening also featured a song, two magicians, an excellent rap about wavelet noise, one particularly brave student who went up on stage without his shirt, a children's rhyme, and a hell of a lot of good research. Welcome to my world.