Search
  • News
  • Learning in Machines & Brains

Research sparks virtual modelling company

by Lindsay Jolivet Feb 3 / 14

A new company that sells virtual human models to clothing and video game designers has grown out of research by a computer scientist in Germany — research he says could change online shopping and even help treat anorexia.

10-body-model-wpa-300x172
Body Labs’ software uses a body scan to build a 3D model that can be posed. Photo c/o Body Labs and bodyhub.com

Associate Michael J. Black, a member of the Neural Computation and Adaptive Perception program and a founding director at the Max Planck Institute for Intelligent Systems in Germany, developed a model for making virtual people from 3D body scans.

It began with a data set built from high-quality laser scans of about 4,000 different people, and incorporating a statistical understanding of the body’s movements.

“If a heavy person sits down, the fat of their stomach rolls over their pants in a different way than if someone with a very muscular stomach sits down,” Black says. “If you really want a computer to understand the human body, you have to understand how the human body moves and how its shape changes as it moves.”

Black and his team developed an algorithm using a template of the human body. They assigned points to various body parts, creating a map of the body, and then programmed software to align these points on bodies of many different sizes and shapes.

The result, which became the basis of the company Body Labs, is software that can take a static 3D scan of a person and turn it into a realistic animated model. Right now, Body Labs markets its virtual models to clothing designers, equipment manufacturers and video game companies that have access to the high quality scanners needed to build accurate models.

But Body Labs has built a new prototype with the long-term goal of letting consumers create their own virtual selves, which they could use for custom online clothes shopping, using mass market technology such as the XBox Kinect. Black says he hopes the work might also solve a little problem of his own.

“I’m sort of tall and skinny,” he says, adding that clothes on store shelves tend to be ill fitting on his frame. “I had this dream that I could shop on the web and find clothes that fit me.”

However, Black says his team is also collaborating with medical researchers to study how people perceive body image, including subjects with anorexia and body dysmorphic disorder.

“If I take your body scan and I can make you a little bit fatter, after you look at it for a little while, you kind of get used to yourself being fatter. So when I reset you to your normal weight, you’re sure that you’re too skinny,” Black says.

“Maybe this could be useful in helping people understand or adjust their perception of their body shape.”

Black says his work complements a range of new developments in artificial intelligence, such as Geoffrey Hinton’s work on deep learning. Whereas Hinton’s research could teach computers how to recognize people in video, Black’s could tell them what clothes to buy.