Data Innovators Body Labs

Published on February 13th, 2014 | by Travis Korte

0

5 Q’s for Eric Rachlin, Co-Founder of Body Labs

The Center for Data Innovation spoke with Eric Rachlin, co-founder of 3D body modeling company Body Labs. Rachlin discussed the future of body modeling, retail applications, and why elite athletes’ bodies are an area of particular interest.

Travis Korte: Can you briefly introduce Body Labs, what you make, and who uses it?

Eric Rachlin: Body Labs has the world’s most advanced statistical model of human pose and shape. This model was developed through almost a decade of research, first at Brown University and later at the Max Planck Institute for Intelligent Systems. We make it fast and easy to create remarkably accurate 3D models of people. Our models can be generated from 3D scans, measurements, demographic information, or even consumer depth cameras such as the Xbox Kinect.

By having a model that “understands” what people look like, Body Labs is able to turn raw data into a “smart” 3D model. Data is just numbers, but when numbers become a model, they become much more powerful. When we create a body model of someone, that model can be reposed, animated, measured or even reshaped. When we have multiple models of different people, we can precisely compare their overall shape, or the shapes of specific body parts. More generally, our models provide a general-purpose interface between individuals and shape-specific products and services.

TK: Describe some of the challenges associated with moving from a static, lifeless body scan to a “smart” 3D model.

ER: It took us many years of research to be able to automatically go from static 3D data to a 3D model that understands how people move and how one person’s shape compares to another. To accomplish this, we trained our models from thousands of high quality 3D scans. This was difficult because raw scan data is often noisy, and more fundamentally, scans don’t provide an explicit correspondence between one body and the next.

Even once you’ve managed to collect thousands of body scans that represent the full range of human shapes and poses, you cannot turn that data into a useful statistical model without establishing an anatomically accurate point-to-point correspondence between the scans. Bringing 3D body scans into registration with a common template mesh is at the core of what Body Labs does. Our algorithms have been developed over many years. They’re fast, accurate, and fully automatic.

TK: Tell me about some of your favorite unexpected or interesting use cases of the body model technology.

ER: I don’t know if it’s unexpected, but our ability to simply document and compare bodies is quite interesting when applied to particularly “rare” or “elite” body types. Professional athletes, for example, exhibit truly unique body types that are the result of years of training. Being able to see how athletes compare to the rest of us is always fascinating. Part of the reason we started Body Labs was to get our technology out of academia and into the hands of real people. If we can get to the point where we’ve literally documented millions of body shapes from all over the world, we’ll know we’ve succeeded in doing something fundamentally important.

TK: You mention that you are working on software that uses a Kinect to produce 3D body scans. If such scans were easily and cheaply available in the future, how might it change the average person’s life?

ER: First off, I wouldn’t say “if.” Inexpensive, high-quality consumer-grade 3D depth sensors are coming, there’s no question. A few years from now, 3D scanners will be available to consumers not just in the form of the Kinect, but also built into laptops, TVs and smart phones. That said, the body scans produced by these devices aren’t so powerful on their own, it’s only when we turn them into a 3D body model that things really get exciting. By turning a scan into an anatomically consistent body model, we provide a platform on which companies can build shape-aware goods and services.

As an example, imagine shopping for clothes online when a store knows exactly what you’re shaped like. They could not only provide highly accurate shape-specific clothing recommendations, but also offer a much wider range of sizes to ensure that the clothing you purchase has been tailored to your specific body shape. Outside of retail and product design, imagine being able to track your body shape over time, and knowing precisely how a particular diet or exercise regimen is going to affect your physique. In all of these settings, shape-aware data analysis is greatly facilitated by cheaply available 3D scanners, but it also requires an anatomically aware 3D model that can be automatically fit to the raw 3D scans.

TK: It would seem that with such an advanced computer vision foundation, your software might be applied to 3D models of other things besides human bodies. Have you used the technology for other purposes, or do you have any plans for spinoff initiatives?

ER: You’re right that the foundations of our technology are more widely applicable, but we’ve chosen to focus on human bodies because they are really, really important. Our physical form impacts almost every aspect of our lives, and yet up until now, computers – which also impact nearly every aspect of our lives – haven’t been able to really understand what people are shaped like. In terms of extending our technology, I’m certain my former colleagues at the Max Planck Institute for Intelligent Systems will continue to apply the foundations of body models to new domains; but right now, there are still a myriad of human-specific applications left to pursue.

Tags: , , , ,


About the Author

Travis Korte is a research analyst at the Center for Data Innovation specializing in data science applications and open data. He has a background in journalism, computer science and statistics. Prior to joining the Center for Data Innovation, he launched the Science vertical of The Huffington Post and served as its Associate Editor, covering a wide range of science and technology topics. He has worked on data science projects with HuffPost and other organizations. Before this, he graduated with highest honors from the University of California, Berkeley, having studied critical theory and completed coursework in computer science and economics. His research interests are in computational social science and using data to engage with complex social systems. You can follow him on Twitter @traviskorte.



Back to Top ↑

Show Buttons
Hide Buttons