In the labs of George Washington University, students are laboring in labs covered in black-and-white dotted paper, puzzling out how to make a machine that understands images like the human brain.It’s a question that Google is trying to tackle in typical Silicon Valley fashion with Project Tango, a smartphone that will use its sensors to map the environment around its users in real time. But this research isn't restricted to the secret labs of West Coast tech giants. A piece of it is happening right here in Washington.
The vision for Tango is to make a phone that can not only recognize places — so you can, for example, navigate within a store to find a product on a shelf — but also potentially tap into large-scale 3-D maps to help the phone better understand the scene.
Think of it as more detailed version of Google Maps. In a mall, a Tango phone could direct you to the nearest restroom by using a meaningful 3-D map of the mall.
Tango phones can also be used to enable augmented reality gaming experiences — allowing users to play 3-D games fully embedded in the real world. In essence, the phone becomes a visual portal to another world.
At George Washington, the scientists have been helping Google deal with the basics: getting the phone to understand its own sensors and the images it sees to understand how what it's sensing matches up with what it expects.
That's what the dotted pieces of paper are for: It provides a graph for the smartphone to better process its own location and set points of reference for the sensors inside the device.
"That lets us figure out how we've moved in the world: how the images look and how things change when we move around," said Jack Morrison, a doctoral student at GW.
"We know where this is in the world," he said, gesturing toward the pattern on the wall. "If you know what you're looking at, you know where the camera is."
The potential for a smartphone-size device that could process that kind of information is revolutionary, said GW Professor Gabe Sibley.
"This is robust enough for use in the real world by regular folks," Sibley said. "That is a breakthrough."
The work intersects with other GW research on new open-source camera calibration software, as well as wider work the lab is doing on teaching machines to recognize particular objects regardless of where they are in a room. That means that if your phone knows, for example, what your keys look like, you could have it scan the room and find where you've left them.
Implications
It may sound like science fiction, but it may not be that far off.
"I don't see any fundamental reason this couldn't be in a consumer phone," said Kyle Wiens, chief executive of the repair Web site iFixit, noting that other devices that do cruder versions of the same thing, such as the Microsoft Kinect, have done well on the wider market.
Kinect can read both depth and movement in a room. That was a breakthrough on its own, Wiens said.
But Tango phones have even more potential because they are smaller and can better understand its location, he said.
"The Kinect has a single view at a room," he said. "But with the phone, you could walk behind the couch and add it to the model. The gyro would tell the phone that it had moved and by how much."
Looking beyond Project Tango, mobile devices will be able compile information from millions of measurements and assemble giant 3-D maps of cities, countries and even the whole planet. Connecting those experiences will pull together human knowledge in a way that allows people to search and re-experience memories in unimagined ways — an idea, Sibley notes, that was really first proposed in 1945 by noted engineer Vannevar Bush.
"That is the big, big picture of what this is about," Sibley said. "To me, that's more exciting."
What's exciting about Tango, Sibley said, is that it brings grand ideas that computer scientists and robotics engineers have been chipping away at for years into something that's accessible, useful in everyday life and portable enough to fit in the palm of your hand.
"This could be the time where computer vision breaks through and becomes a widely used application," Sibley said.
Part of a wider push at GW
The lab’s ambitious effort is part of a major push by the university to become one of the nation's top-tier research facilities for science and engineering.
It's a bit of a shift for GW, which is known best for its policy programs. Since 2008, the school has hired 44 new faculty members in fields ranging from robotics, like Sibley, to biomedical engineering, defense science and fluid dynamics, all in the pursuit of elevating GW's profile in the STEM world.
It's also broken ground on a nearly $300 million building that will house a number of the school's science and engineering departments in an open environment. (Yes, even the floor plan promotes cross-disciplinary work.) The building, which takes up nearly an entire city block at the intersection of 22nd and H streets NW, will have 10 floors of teaching space. It's expected to be move-in ready sometime in 2015, said David Dolling, dean of the School of Engineering and Applied Science.
The hope, Dolling said, is to turn the engineering program into a crown jewel of the department for GW. The university would like to foster a community of innovation that spurs its graduates into the start-up community, he said.
The start-up community in Washington, Virginia and Maryland could easily lend itself to a high-powered network of entrepreneurs, Dolling said, though he acknowledged that it will take time to build a network that rivals either Stanford or the dedicated alums out of GW's policy programs.
But he's still optimistic that it can be done, thanks in part to the new faculty and facilities. "We're drawing world-class people and students," he said. "Why can't the next Google come out of GW?"
By Hayley Tsukayama (Washingtonpost)
No comments:
Post a Comment