 |
|
Page 2
 |
 |
The PCG's cluster computer.
Photos: Courtesy Cornell Program of Computer Graphics
|
Pixar Animation Studio's Rob Cook was one of Greenberg's graduate students
from 1979 to 1981. Last year he and two colleagues won an Oscar "for
significant advancements to the field of motion-picture rendering
as exemplified in Pixar's RenderMan"--software used not just on
animated films like Toy Story and Monsters Inc. but also
for the blockbuster special effects in such movies as Star Trek II,
Terminator 2, Jurassic Park, Titanic, The Phantom
Menace, The Matrix, and Gladiator. Cook's contribution
to RenderMan is directly descended from the work he was doing at the PCG
and has drawn heavily from subsequent advances pioneered at Cornell. "A
lot of the pictures that were being made then looked like plastic,"
he says, "and I was trying to figure out why." In the library
Cook found a huge volume on material properties containing a small section
on reflectivity, perhaps the only data of its kind at that time. "Someone
had measured light reflecting off of objects, but only did it at one
angle so it was very limited," he says. "But at least it was something."
Knowing how copper, rubber, and a handful of other materials behave when
hit by light allowed Cook to develop algorithms that improved the realism
of computer images.
By the early 1990s Greenberg had established a light-measurement laboratory
sensitive enough to gauge light at the wavelength level, giving the PCG
the ability to generate very accurate data of the kind Cook's research was
based on. Sophisticated as it is, the lab looks remarkably homemade--full
of oddly anachronistic Brazil-like gadgets, it resembles more the
basement workshop of an eccentric inventor than a place where you'd expect
to find a team of white-jacketed technicians. Still, it's arguably
the finest facility of its kind. "If we know the wavelength of
light, the angle at which it is hitting, and the properties of the material--like
how deep light penetrates--then we know how the light is going to scatter,"
Greenberg says. "To determine that for, say, my desk, we cut out a
piece and we shine light at it from a given wavelength and measure the energy
that goes out in all directions. And we change the angle again and again,
all around from all directions, and with all the wavelengths. It's a five-hour
job even if we automate it, which we have." Obviously it would be impracticable
to do this for every one of the countless materials used in architecture.
Instead Greenberg's team developed algorithms that predict how light will
scatter when it hits a surface based on the material's smoothness, orientation,
and composition. Graduate student Sing-Choong Foo and professor Kenneth
Torrance (who originally set up the lab) then performed experiments in which
they set up identical situations--in terms of material, illumination, and
geometry--in both the lab and the simulation, and confirmed that the
model was accurate. In early versions it was far too slow to be used in
real-time applications, but by the late 1990s postdoc Eric Lafortune had
streamlined the simulation so that it can now be done very fast.
|
|
 |
|
 |
|
|
 |
|
 |
These images are stills from a computer animation called "Cornell
in Perspective," completed in 1972. Greenberg and his students
modeled much of Cornell's campus, including the then-unbuilt Johnson Art
Museum, and put together a fly-through movie. A still appeared on the
May 1974 cover of Scientific American magazine, introducing the
world to the concept of "Computer Graphics and Architecture."
Photos: Courtesy Cornell Program of Computer Graphics
|
It turns out that to make a simulation accurate, simply modeling the light
reflecting off of the first surface it hits isn't enough. You
also have to take into account the way the reflected light bounces
around. This is called global illumination, and it requires some heavy computing.
"Let's just take this room, which has six light sources and light coming
through the windows," Greenberg says. "The light hits the table,
bounces off, hits my face, gets the specular reflection from my forehead,
bounces off--everything interacts with everything else. So there's a relationship
between all of the light sources and all of the surfaces in the room."
Mathematically you could solve this problem by dividing the room into thousands
of tiny areas, measuring the light going in and out of each, and performing
hundreds of thousands of simultaneous equations between them. "Clearly
impossible, but we would get it right." So again, a simulation was
developed and checked against data in the light lab. "We built some
environments where we knew the qualities of the materials and the geometry
of the light sources," Greenberg says. "We set up a liquid-cooled
camera, ran our simulation, and checked it against what we would record
in the digital camera in each of ten different wavelength bands. We know
it's correct. I must have had ten or fifteen Ph.D. students in the
last couple of decades who have worked only on aspects of this problem."
The PCG isn't actually a department. It's more like an interdisciplinary
think tank comprised of people from various fields--computer science,
architecture, engineering, art, psychology--who are all studying these problems
collaboratively. Psychologist James Ferwerda, for example, is working on
the most recent addition to the program: research on how the brain interprets
images so that precious processing power is spent only on rendering what
the human perceptual system can actually pick up.
Each fall undergraduate architecture students are brought into the program
as part of an elective third-year design studio. Working in groups of three
or four, they complete a project using whatever tools--digital or otherwise--best
serve their purposes, and then present the completed design online. "Students
come in with no prior knowledge, and we don't require prior knowledge,"
says Moreno A. Piccolotto, who runs the studio with Greenberg. "What
we're trying to do is assist the design process with the tools and research
done here." Last semester's assignment was a modified version
of a competition brief for a huge residential tower complex on an East River
site adjacent to the UN. "We encourage them to use traditional media
and methods, and then to mesh them with these new ways to do things,"
Piccolotto says. Rather than teaching the technology--which includes AutoCAD,
a rendering package called 3D Studio VIZ, HTML, and Flash--in separate classes,
the students learn it as they need it during the design process. "We
train them to be able to learn about the software by themselves, according
to what they want to do."
A new tool the students used was a beta version of Autodesk Architectural
Studio (AAS), software based in large part on Piccolotto's master's thesis,
created to facilitate the conceptual portion of the design process. Unlike
AutoCAD's heavily technical interface, AAS allows the user to simply sketch
with a stylus onto a tablet to create three-dimensional drawings and models
that can be worked on collaboratively over the Internet. Architects at three
firms--Richard Meier & Partners; Skidmore, Owings & Merrill;
and Kohn Pedersen Fox Associates--used AAS in conjunction with videoconferencing
technology from their offices in New York to critique the students
in Ithaca. "We could sketch on the Web--and that's how we communicated
with the firms in the city as well as with each other," architecture
student Dana Getman says. "If I needed a quick crit from Moreno, I
could sketch it and e-mail him, and he could sketch on top of it and e-mail
me back." Piccolotto was impressed at how well the software worked,
especially for senior designers who aren't necessarily known for their computer
skills. "Richard Meier sat down in New York City, picked up the stylus,
and got down to business right away." The technology became transparent.
"It was so simple, it was unbelievable," Meier says. "I was
able to sit and sketch some suggestions in a similar way that you might
if you were face-to-face in a studio."
"When I got started I went to people and said it would really be great
to consider trying to make the displays better--add coloring," Greenberg
says. "Nobody was interested. As late as 1976 they said, 'Color will
never be needed.' I went to the best: IBM, HP, Silicon Graphics, Apple."
A few years later, when the PCG began work on global illumination, Greenberg
was told it would never have any practical value. "Everybody said it
would never be used. Now it's standard presentation in every architecture
office." So believe him when he says that at some point, in the
not too distant future, you'll be working with software that will allow
you to sketch into the computer, easily turn the sketches into scalable
3-D models, and render them in real time. Whether you're designing a discotheque
or a garden deck, a skateboard or a space shuttle, reshaping breasts or
reconstructing a shattered face, you'll know exactly how it's going to look.
Eventually the distinction between model and rendering will melt away, and
design will happen in a perfect physically accurate digital environment
that looks and acts like the real world. "That's the Holy Grail,"
Greenberg says. "The Holy Grail."
|
|
|
BACK TO TOP
