Ready to learn? Go to gerberu.com today.
Meet the Gerber Team: James O'Brien
The Oscars are this Sunday, and perhaps you’re already rehearsing your own fantasy Oscar acceptance speech and imagining who you would thank if you won an award? In this Oscars-themed “Meet the Gerber Team” blog series, we’ve been introducing you to Gerber employees who are perhaps a little bit closer to the Academy Awards than you or I.
Last week, we introduced you to voting Academy member and Gerber employee Scott Frankel, and earlier this week, we introduced you to an Academy Award winner who also happens to be a Gerber employee as well, Zoran Kacic-Alesic.
In our final Oscar themed “Meet the Gerber Team” blog entry, we’d like to introduce you to James O’Brien, another Academy Award winner in the Scientific and Engineering category, and member of Gerber’s Avametric team!
James is Chief Scientist at Avametric, a San Francisco-based company recently acquired by Gerber. Avametric team members are a team of fashion vets, engineering gurus, technology-obsessed scientists, savvy investors and visual effects experts who are shaping the future of #FashionTech. Check out our Q&A with James, below!
Q: In addition to your role as Chief Scientist for Avametric, you're also a Professor of Computer Science at UC Berkeley, and you have an extensive background in animation -- including time spent working with film studios such as Dreamworks. How did your path lead you to these exciting roles and where you are today in your career?
James: My interest in Computer Graphics and Animation started when I was an undergrad. My major was Computer Science, but I was also getting a minor in Visual Arts. When I took an introductory course in Computer Graphics, I realized that those two fields I was drawn to could merge together, complementing each other. Being able to write code that would take my ideas and build images from them was amazing to me. Later, when I was in grad school, I learned about using physics simulations to generate animation, and that has been the main focus of my work ever since.
I think that most people are familiar with the idea that a bridge engineer might use a computer simulation to see how a proposed structure would hold up during an earthquake, or that a airplane designer might use a simulation to explore how airflow over different wing shapes would generate lift. Those simulations use mathematical techniques such as finite element analysis, finite differencing, or smoothed particle hydrodynamics.I also think that most people are not too surprised to learn that those methods for using simulation in engineering tasks involve complex code and a lot of heavy math. However, what some people might find surprising is that those same techniques for computer simulation are used for generating animation in video games and special effects in films. If you want to create an animation of Godzilla smashing and knocking a building down, then the most realistic way to do it is to set up a computer simulation of that building using a finite element model with several million elements, include boundary conditions that correspond to the forces generated by a mega-giant lizard, and then run the simulation on a compute cluster for a day or two. When done right, it looks real.
For me, there is just something magical about developing computer algorithms that can generate fantastical imagery that looks real. Once I started working in that area, I was hooked! I finished my doctorate and then moved to California where I started teaching in UC Berkeley's Computer Science department. The Bay Area is also where companies like Pixar, ILM, and Dreamworks are located, so it's been very easy to stay connected to the VFX/Film industry.
Q: Can you tell us about your work in film and animation, and how it informs your work now at Avametric?
James: Most of the work that I've done in film and animation has been indirect, through working with my students on developing algorithms that have been published at conferences and in academic journals. Computer Graphics is a fast-moving field and it's pretty common to write a paper and then a few months later someone comes and tells you how they've implemented the idea in their production pipeline. I've also done consulting work for studios where I help them design algorithms for some particular problem that they have. However, the most direct impact that I've had has been through a previous startup company, Pixelux Entertainment, where we built a software system that has been used to animate destruction effects in well over a hundred films. So the example that I gave earlier of Godzilla smashing a building was an actual example where our software was used.
Avametric was started with the core idea of using the same sort of simulation tools that are used in engineering and in animation to realistically depict how clothing would look on a human body. This goal would combine the requirement of predictive accuracy from engineering applications with the requirement of creating realistic imagery from animation applications. The result is a software system that includes a core component called ARCSim, which stands for Adaptive Refinement and Coarsening Simulation. A key feature of ARCSim is that it automatically adapts to whatever it is that your simulating. So if the cloth drapes in a way that should form interesting folds and wrinkles, then the simulation with automatically add detail elements exactly where they are needed, and when they are no longer needed it will remove them.
The work at Avametric has, of course, focused on doing the best job that we can capturing real cloth behaviors with high fidelity, but ARCSim also has been used to model things like folding paper and breaking glass.
Q: You won a Scientific and Engineering Award from AMPAS (Academy of Motion Picture Arts and Sciences) What was it like to receive that recognition? And, we have to ask -- where do you keep your award?
I think that the destruction software we developed has had a pretty significant impact on the special effects industry. The Academy Award that I, along with my collaborators Eric Parker and Ben Cole, received in 2015 was a really wonderful way of being recognized for that impact. Attending the awards ceremony was very exciting, but the best part of the experience has been the recognition of other people in the industry who have made their own incredible contributions to the field.
As for where the award is kept, I'm going to have to disappoint you a bit. The Academy doesn't give an Oscar statue to people who get awards for Technical Achievement. But I did get a nice plaque that I have hanging in my office at home.
Q: As an award winner, I imagine you have more than a passing interest in film, animation, and the Oscars. Can you share any of your favorite films or Oscar picks for this year?
Well, I'm not really sure that my guesses are much better than anyone else's. But, if it were up to me then I'd probably pick "Isle of Dogs" for Best Animated Feature and "Bohemian Rhapsody" for Best Picture. But, I haven't seen "Green Book" yet and I've been told that it might change my mind. I've also been told that our destruction software was used for "Spider-Man: Into the Spider-Verse," which was nominated for Best Animated Feature but not for Best Visual Effects.
As an aside, one of the sorts of amusing things about making software that other people use for something as high-profile as a movie is that we don't always know when a film uses our software. I can sometimes tell right away when I watch a film, but unless I start stepping through a shot frame by frame it's really hard to tell. That's sort of the point of what the artists are trying to do: make the effect look absolutely real. I guess a similar situation must exist for everyone at Gerber when they see some stylish garment and there is a good chance that it was designed using Gerber software and made using Gerber hardware.
Q: What are you seeing in your current research or teaching about the future of animation or simulation in the world of e-commerce?
Well, one of the really exciting things that is changing animation and simulation is the same thing that is rapidly having a huge impact on nearly everything: a branch of Artificial Intelligence called Machine Learning. The technology that makes machine learning work has gotten good enough that many problems in programming have changed from "Design an algorithm that tells the computer what to do" to "Design an algorithm that looks at some examples and then figures out how to generalize." There is still a lot of complexity in building these systems, but a lot of problems that were previously not possible for computers are now suddenly very possible.
One of my Ph.D. students, Stephen Bailey, recently published a paper with collaborators from Dreamworks Animation where we looked at how the system they use to animate characters could be sped up with machine learning. The system they were using to animate the characters in the Kung Fu Panda and How to Train Your Dragon movies was really great in that it let the animators do fabulous work and give a lot of lifelike expressivity to the characters, but it was also very computationally expensive. The result was that even using a top of the line workstation, the animation system could be slow and difficult to work with. The system that Stephen built used a machine learning algorithm to learn a model of how the controls for each animated character determined the pose and movement of the character. Once the model was learned, it could be used instead of the original system for animating the characters. The big win is that the learned model was several hundred times faster than the original system even when running on an iPad!
In the world of e-commerce, I think the thing I'm most excited about is what Avametric and Gerber can do together to make it easier for people to buy clothing online. Virtual try on is one of those ideas that has been around forever, but no one really knew how to make it work. There were lab demos where people would show very simplistic garments that looked like they were made out of rubber on low-polygon geometry, but those demos were nowhere near the level of quality and realism where someone could use them to make a purchasing decision. At Avametric we built a virtual try-on system that can show users how real clothing items made from realistic fabrics would look on their own body. Each step in that pipeline, body scanning, cloth simulation, and image rendering, represents a lot of complex technology that has been harnessed to create a compelling and user-friendly experience. That same simulation technology is now also part of AccuMark 3D. I think there is a lot of very exciting work that Avametric and Gerber can do together that will totally revolutionize both the experiences of designing and buying clothing.
Questions about this blog? Email us! firstname.lastname@example.org