Conference-goers gathered in one of the Los Angeles Convention Center's many large lecture halls on the eve of the first full day of SIGGRAPH 2004 to hear all 83 of this year's
SIGGRAPH papers succinctly summarized. The much-anticipated "Fast Forward" session gives each paper author exactly 50 seconds to explain what their paper is about and to entice
conference attendees to hear their full, thirty-minute presentation later in the conference.
As papers chair Joe Marks said,
the session is "about 4,000 seconds of pure chaos." Marks introduced the session by reminding the audience of the high-quality of the papers program. Of the 478 submissions received
by the committee, only 17% were accepted. SIGGRAPH is widely considered to be the premiere forum for presenting academic advances in computer graphics, and as such, is highly competitive. Each paper is reviewed by multiple members of
the academic community in a double-blind process where the authors do not know who is reviewing their work and the reviewers do not know who wrote the paper they are reviewing. A large number of universities and industrial labs are represented
among the paper authors, with a few institutions being particularly well-represented, such as Microsoft Research, Carnegie Mellon University,
Stanford University, MIT, the University of Tokyo, Mitsubishi Electric Research Labs, and the University of Washington.
Several research themes were common in the descriptions of this in this year's crop of papers.
Several papers were concerned with digital photography -- some were concerned with interactively combining several images (image taking a photo where your eyes were shut
and in one simple gesture having the eyes replaced with your eyes from a different photo), while others algorithmically merged
flash and no-flash photos of the same scene in order to harness the advantages of both photos. Simulation of natural phenomena is also
a popular topic, with an entire paper session being devoted to improved techniques for rendering "smoke, water, and goop." Animation of realistic human characters also continues to
be an area of research interest, while general-purpose programming on graphics processors (GPUs) is really taking off this year. Additionally, SIGGRAPH is focusing on the interactive aspects of
graphics by highlighting a few papers from other, related conferences -- UIST (User Interface Software and Technology)
and VRST (Virtual Reality Software and Technology).
Although most of the paper authors filled their allotted 50 seconds with succinct summaries of their
research contribution, several went above and beyond, finding creative ways to use their seconds in the spotlight. The authors of "Interactive Digital Photomontage" enticed
potential attendees with strangely morphed images of themselves, while the author of "Volumetric Illustration: Designing 3D Models with Internal Textures"
jokingly showed a photo of a hollow cake he received as a gift upon the acceptance of his SIGGRAPH paper, and then showed how
his volumetric rendering technique could create an image of a much more appetizing dessert!
A "Mission Impossible" theme intrigued the audience to come learn more about tensor textures in the talk on "Multilinear Image-Based Rendering,"
while the author of "Textureshop: Texture Synthesis as a Photograph Editing Tool" promised that by watching his lecture,
conference-goers would learn how to alter a digital photo to "make their cheap, white T-shirt appear more hip." Researchers from
Mitsubishi Electric Research Labs donned Village People costumes and sang about "RFIG" (Radio Frequency ID and Geometry) to the tune of "YMCA" in order
to promote their work, "RFIG Lamps: Interacting with a Self-Describing World via Photosensing Wireless Tags and Projectors."
The authors of "Target-Driven Smoke Animation" showed that computer scientists can be cultured by summarizing their
research in haiku: "turbulent smoke. so difficult to control. attracted by targets." The authors of "A Method for
Modeling Viscoelastic Fluids" showed a slightly more Dr. Suess-esque penchant for poetry by reciting a rhyme about spit and snot.
Still more aspiring poets, the authors of "Triple Product Wavelet Integrals for All-Frequency Relighting" rhymed "Triple
product integrals confusing you? Come to our talk and with wavelets you can do."
Vocal manipulations also caught the attention of the crowd. The author of "Spacetime Faces: High Resolution Capture for Modeling and Animation" surprised
everyone with his excellent "Crocodile Hunter" impersonation, while another presenter got laughs by talking quicker than many thought humanly possible to
describe enormous detail about his work within the given time, then closing with a facetious promise that, in the paper session, we would
read his paper verbatim.
Papers will be presented during the conference from Monday through Thursday. Each presenter will be given a half-hour slot to
present their ideas in more detail, followed by questions from the audience. The papers were distributed to conference attendees in a printed proceedings, but are
also available online through the ACM Digital Library.