Monday, December 21, 2009

I was at Siggraph Asia 8)

Saturday was the last day of Siggraph Asia 2009.
This year's event was in Yokohama, which is practically in Tokyo. About 35 minutes commute from my home. Going to the airport alone for an international flight takes a lot longer than that (^^).

Unfortunately, the lack of jet-lag meant that my sleeping habit wasn't screwed up, making it difficult to wake up on time for the initial lectures which were at 9 am !!
All those conferences starting so early, it's insane. But if one considers that most people probably go for work, then it sounds more like business and less like slacking if one starts early in the morning. Also for those coming from abroad, being able to finish the day by 6 pm leaves some room for possible touristic activities.

Overall, Siggraph Asia was actually quite nice. Not as big as the American counterpart, but it had a number of interesting talks and events.
I'm terrible at reading those schedules and I usually end up missing something that I wanted to see, or getting some hints from some more attentive coworker.

There were some parties. One from NVidia, with speeches and presents (a coworker of mine got the top prize: an HP netbook !). There was also some sort of club event (see the pictures below).

I saw a few cute girls hanging around but I didn't make any "move", because well, it's kind of lame, I think. But that got me thinking.. wouldn't it be nice to meet a girl that other than being cute, etc etc, is also somewhat in my field ? I'm not sure if that would be a good thing in the long term.
I'd probably end up dreaming of the nicer looking gals that don't care about CG.

The Electronic Theater was quite entertaining. Two hours of short films, including some clips revealing the level of compositing that goes into the modern CG.
The more complex scenes are usually built on separate layers which are rendered at different times and then composed together. Off-line rendering is really something quite different from real-time graphics.

Recently, talking to a coworker from the movies department, I learned about a particularly rendering that they had to do. It was about a single frame with two characters holding hands but the scene was so heavy that they had to render the characters separately. The hands were then stitched together at a later stage via compositing 8)

One presentation close to my current interests was that of RenderAnts. Not too many revelations if one already read the paper. But hearing about some questions from the public got me thinking and I looked at the paper again and realized that basically RenderAnts converts RenderMan shaders to the shader opcodes already supported by the GPU. It doesn't build a custom VM (which would be quite a bit of GPGPU work).
Using the GPU's shader commands makes more sense and is probably more efficient, but it makes it harder to implement some additional features that just wont work in the restricted context of a GPU built around Direct3D and OpenGL.
The author himself said that RenderAnts is more like a proof of concept than a product aimed to compete with other renderers. It proves that a REYES pipeline can be implemented on a modern GPU, but it might also prove too difficult to extend to the levels necessary for actual production rendering.

Another interesting paper was "Virtual Spherical Lights for Many-Light Rendering of Glossy Scenes" (PDF). Which basically improves the famous "Instant Radiosity" from 1997. If anything because it describes a weakness of the much talked but not so implemented method from almost 13 years ago.

Of course there was a lot more stuff, but no time to see all of it and no time to write all about all that I saw and heard... which is why one wants to go: because you can't really get all of it by reading something somewhere.
This also goes to those that claim that they just need to read the papers on-line (assuming that most are available on-line) ..I probably have claimed that myself a few times in the past 8)
Surrogates are not enough.. the overall experience of going to conventions is something different. It may be a detail, a person that you meet, a book that you buy at the convention's store, a lecture that you attended by accident.. it is about being at the right place at the right time.
Having all information potentially accessible somewhere in the web isn't the same as going to a conference.


Anyhow, here are the pictures that I took:


..incidentally, Picasa Web finally got its act together and brought back up the pictures quality to a reasonable level ..after a period of sad low quality JPEG quantization settings (Google servers recompress images as needed for displaying on the web).



  1. hhmm, looks like siggraph has gotten a lot weirder since 1998.

  2. Strange flowers, inflatable hotel rooms...

    in 1998 we just had dynamic LODs!!!

  3. They year of the LODs was 1997 ;)

    The inflatable thing I think it's for some virtual reality thing. Not an hotel room like I wrote in the comment 8P

    In any case, that's from the "emerging technologies" portion which is near the exhibition area.

    Lectures, classes, presentation of technical papers, sketches and all that stuff about sitting and listening is still there and in fact is the main reason why we go. I didn't take pictures of that because it's less meaningful (pictures of slides projected on a wall 8) and in some cases there is also an explicit request not to take pics (big deal.. ..some people just do it anyway).

    I don't remember specifically about "emerging technologies" or the "art" sections in 1997.. but I looked around for info on the history of Siggraph and on a doc (available only with ACM member login !) I found the following:

    The exhibition of interactive work was first
    presented in 1991 and is now called
    “Emerging Technologies”

    ..though these things may be on and off depending on the year.

  4. So many promises, so few girls pics... :)

  5. ..ummmm yeah.. but consider that this was Siggraph 8)