brainwagon

Well, I’m sitting at The Standard (a frankly far too chic hotel for a forty something computer geek like myself), it’s not quite 7 A.M. and I spent my first day at SIGGRAPH. I’m hear mostly for the benefit of recruiting: sitting in the booth, answering questions and showing up at our Pixar User’s Group Meeting. We are handing out 20th anniversary Renderman walking teapots: very nice, I managed to get one for Josh, but haven’t picked up one of my own: I’ll try to later and get a picture here.

I’m not really attending papers (you can get links here) but there seems to be quite a bit of buzz about Intel’s Larrabee architecture. Broadly speaking, the trend of GPUs has been to slowly work to expand both the number and capability of the different functional units: more shader units, that can execute more arbitrary code, and more texture units, which can present results which are available to more units. Larrabee leapfrogs this: we are back to having X86 cores (not my favorite architecture, but ubiquitous) which are fully general, linked together by a fast shared cache with scheduling done in software. To me, this represents the obvious end game to the evolution of GPUs. Companies like nVidia have been trying to tell us that we can use GPUs to do more general computation: Intel has delivered an architecture where that claim is much more obviously true.

Oh well, I’m gonna get some pancakes at IHOP, then walk the show floor a bit. I want to try to see what the state of the art in stereo monitors is, and maybe see who I can bump into. I’ve got booth duty again at 1:30 (more teapots handed out at 2:00) and then the User’s Group meeting at 6:30. If any readers are at SIGGRAPH, feel free to come by the booth and say hi.

This week I’m busy at work trying to complete dozens of SIGGRAPH sketch reviews, so it seems like an opportune time to present the usual link to  online copies of the SIGGRAPH 2006 Papers that have been accepted.  I haven’t had a chance to review these yet, but there is usually something good in this pile.

[tags]SIGGRAPH,Computer Graphics[/tags]

Josh over at tinyscreenfuls is digging some of the fancy “pencil sketch” effects that the Mac can do with its internal camera.  Back in 1998, I experimented with writing some filters that did much the same, with some examples that I generated shown on the right.   Macintosh?  I don’t need no steekin’ Macintosh.

And now, for your next project, render an entire feature length film.   Beneath your desk, you’ll find a pencil, a yellow pad, and a C compiler…

Last night we had a screening of Hoodwinked, the new animated feature by director Cory Edwards and co-directed by Todd Edwards. It’s a retelling of the classic story of Little Red Riding Hood, and features the vocal talents of Glenn Close as Granny, Ann Hathaway as Red, James Belushi as the Woodsman, veteran voice over master Patrick Warburton as the Wolf, David Ogden Stiers as Flippers, a suave investigating frog, and Andy Dick as Boingo the rabbit. The plotline is a fractured fairly tale: the story you know isn’t the real story…

I’ll get the negatives out of the way right off the bat, because I’m not really sure how to be nice about this. In terms of visual effects and art design, this movie is not exactly going to knock your socks off. The characters have a very wooden look to them: the characters have an extremely limited range of facial motion and the animation on the whole appears rather stiff. The net result of this is that the entire movie reminds you of some of the old stop-motion Rankin-Bass features you’d see around christmas. The lighting overall is, well, as near as I can tell, there was no lighting. It really bothered me for the first ten or fifteen minutes, especially when I realized that the highlight in Red’s eyes was actually painted on, and stuck to her eye as she looked around. Bleh. There was a couple of times when Granny was center stage and you could literally see some strange polygonal effects around her mouth. Double bleh. And you should never have a roller coaster like scene without motion blur. Yuck.

Oh, and the music? Mostly terrible, although the villain’s song (mercifully, the last in the movie) was somewhat better, and didn’t seem contrived.

Okay, it’s not the prettiest movie, what’s to like?

The vocal performances were on the whole quite good, although I couldn’t really understand what accent Jim Belushi was trying for. Stiers does an amazing job as Flippers, I never would have recognized him as the urbane frog if he hadn’t been listed in the credits. Andy Dick and Glenn close also do well, as does rapper Xzibit as Chief Grizzly.

The story is actually pretty good. Early in the movie I thought it was going to be dreadful, but I think that may have been more of a reaction to the problems I had with the visual look of the film, and that’s probably something that’s fairly unique to people like me who work in the industry. Once I sort of got around that, I began to find quite a bit to like about the story, and by the end, I thought it was actually pretty fun. If you spent $10 to see it, you might feel a bit cheated, but if you got in on a cheap matinee, I would think you might be pretty allright with that.

I stuck around at the end to watch the credits, and it’s pretty clear that they didn’t spend the kind of money that studios like Pixar and PDI spend: their credits are remarkably short. For them to release a movie like this at all and get a national distribution deal is a credit to them.

Overall, I’m going to give the movie a B-, but I’m probably being mean because I stare at computer generated images all day long. Read some user reviews on Yahoo! or whatever if you’d like to get a glimpse at a more well-rounded view. It’s rated PG: some very young children at our screening found the growling wolf pretty intense, they did not like him at all but I suspect most kids over the age of eight will be fine.

[tags]Movies,Movie Review,Hoodwinked[/tags]

Okay, I know this is an atrocity, but you might still find Recolored to be an interesting program for adding color to black and white images. You basically scribble hints into the image, and it propagates the color to nearby pixels that it determines should be the same color. At right, you can click and get my version of It’s a Wonderful Life, appropriate for the holiday season, if somewhat garish overall.

Okay, I’ve goofed around enough this morning. Off to Christmas shop.

Metamerist has linked to me before, it seems only fair that I should return the favor, especially since he came up with this terrific Computational Photography Link Roundup. I’ve linked to some of the articles listed before, since I’m generally interested in the topic, but there are some new bits here too.

Good stuff.

A group at Stanford has created an interesting new camera using a combination of conventional camera and a microlens array to form a “plenoptic” camera. This link hit our photography mailing list yesterday, and I spent some time reading it. It’s really quite clever. Basically it uses the microlens array to serve as thousands of tiny lenses. Each of these “sub-lenses” images the target scene over a narrow field of view from a different spatial location. Then, computer software can take these individual images and recombine them in flexible and clever ways: allowing you to refocus images, shift perspective, and all sorts of other cool ideas.

Very neat stuff.

Leo has shamed us all by writing up a webpage detailing all the things he found interesting at SIGGRAPH 2005. Very nice.

This is mostly just a reminder to myself to look on the Applied Geometry homepage for these course notes on Discrete Differential Geometry.

Holy crap, I spent the day at home today, and was scanning my usual blogs when I read this shocking article that Pixarian Joe Ranft had been killed in an automobile crash. While I didn’t know Joe personally, his infectious smile and humor were a regular feature of the halls at Pixar, as was his voice which would occasionally lapse into his Germanic Heimlich to good comic effect. He was only 45, and had many stories and smiles left to bring the world. The world has truly lost one of the good ones.

Cartoon Brew has some more recollections of Joe.

Holy crap, this sucks.

Addendum: My neighbor Sam sent me an email earlier today, asking me how things were at Pixar. I sent him back a routine email, since I hadn’t heard the bad news. Today was a scheduled group river rafting trip which I bowed out on, since I have had bad experiences with river rafting the last three times I’ve went. Only later did I realize what he was really asking, after I read the news on Boing Boing. Serious bummage.

Addendum²: Ronnie Del Carmen reminisces more powerfully than I ever could.

Every once in a while, it disturbs me that there are parts of the computer graphics world that I rarely delve into, and simulation near the top of the list. I did spend some time fifteen years ago trying to understand inverse kinematics and the like, mostly in the context of robot motion planning, but all that knowledge has long faded. Still every once in a while, I try to dust off my brain and read a few papers on topics I’m not comfortable with, and maybe even stare at some code.

Luckily for me (and many others) there is a lot of good work being done and published. In the world of fluid simulation, some of the most accessible work has been done by Jos Stam, who kindly made his publications, notes, and code available on his webpage. His demo code is about 100 lines of C, includes nothing all that mysterious, and can be adapted to do more complex fluid flows. Check out this short movie to see it in operation. I’m thinking of adapting it to make a spiffy screensaver, mostly as an excuse to delve into its mysteries.

Each year Pixar tosses a Renderman User’s Group meeting at SIGGRAPH. In the past few years, we’ve invited some of our users to present some of their techniques to a wide audience in a portion of the program we call Stupid RAT (Renderman Artist Tricks). Each year, there seems to be at least one presentation that really captures your attention, and this year the award goes to Hal Bertram and The Interaction Trick.

The trick was actually pretty simple: to use the raytracing capabilities of Renderman (essentially a batch process) in an interactive viewer. It’s brilliant really. Try watching the videos and reading the notes. We are all very impressed.

Next week I’ll be leaving for SIGGRAPH, the premiere computer graphics conference. Today I’m trying to work through my agenda, make lists of papers that I want to see, figure out the reception schedule and the list of parties that I’d like to attend. While doing so I noticed that the Computer Animation Festival trailer was available online. This is even cooler than normal because I know the Computer Animation Festival Chair: it’s my next door neighbor and former Pixarian Sam Lord Black. Well done Sam, and I’ll be mooching party invites off you later.

I’ll be trundling off to SIGGRAPH at the end of the month. Any attendees (who I don’t see everyday at lunch here at Pixar) want to get together and schmooze a bit? I haven’t been in a few years, so my dance card is likely to be pretty clear. Drop me an email and let me know what days you’ll be around and maybe we’ll try to get together for dinner or something.

Cool link today from Boing Boing which extracted elements from the game Halflife 2 and embedded them into real scenes using high dynamic range lighting. This kind of approach is one that I think amateurs could pursue in making their own hybrid CG films. Good stuff.

Boing Boing: HOWTO add Half Life 2 elements to your photos