Carnegie Mellon is one of two universities participating in a new research and development initiative with The Walt Disney Company.
Under a five-year agreement, Disney will fund a Pittsburgh research center, including a director and seven to eight principal investigators. Additional staff will include professors, academic interns, scientific consultants, and collaborators. The other collaborative lab is being created at the Swiss Federal Institute of Technology in Zurich. Ed Catmull, president of Disney and Pixar Animation Studios, made the announcement Aug. 11 during the 2008 SIGGRAPH conference in Los Angeles.
Besides its movie and television production arms and its theme parks, Disney is the parent company of ABC and ESPN. The scope of its activities makes the partnership especially valuable, says Jessica Hodgins (CS'89), professor of computer science and robotics and director of the new Disney lab in Pittsburgh. "The number of interesting problems within that space is huge," she says. "I think this is a really special opportunity, and a really exciting time to get to work on these problems."
One of the first projects being tackled in Pittsburgh will be improving the interactive characters and robots at Disney's theme parks, Hodgins says. Researchers are looking for ways to make the interactions between guests and robots more meaningful and less dependent on puppetry and human intervention, she says.
Method to Automatically ID Photo Locations Unveiled
Researchers have unveiled the first computerized method for analyzing a single photograph and determining where in the world the image likely was taken. It's a feat made possible by searching millions of GPS-tagged images in the Flickr online photo collection.
The IM2GPS algorithm, developed by computer science graduate student James Hays and Alexei A. Efros, assistant professor of computer science and robotics, doesn't scan photos for location clues as a person might do. Rather, it analyzes the composition of the photo, noting the distribution of colors, textures, and lines. It then searches Flickr for photos that are similar in appearance.
In tests of IM2GPS, Hays and Efros could accurately locate the images within 200 kilometers for 16 percent of the photos analyzed--up to 30 times better than chance. Even if their algorithm failed to identify the specific location, it could often narrow the possibilities by identifying the locale as a beach or a desert.
They presented their research June 26 during the IEEE Computer Society Conference on Computer Vision and Pattern Recognition in Anchorage, Alaska. For more information, visit the project Web site at graphics.cs.cmu.edu/projects/im2gps.
Spam Blocking Tool Now Recapturing Lost Texts
A research team led by Luis von Ahn (CS'03, '05) has taken the CAPTCHA technology developed at Carnegie Mellon and given it a new purpose--digitizing books and articles produced before the computer age.
When old books or newspaper articles are scanned, about one in five words can't be turned into text because fading, dirt, or other damage prevents optical-character recognition software from processing them correctly. Those words are now being served up by reCAPTCHA, which requires Web visitors to decode the scrambled text before they can register for email or post comments on blogs.
Humans are better than OCR programs at ignoring background noise in the scans; by retyping the words into reCAPTCHA boxes, they turn them into machine-readable text.
In an article for the Sept. 12 issue of Science, von Ahn's team reported that more than 440 million words were deciphered during reCAPTCHA's first year of operation--the equivalent of manually transcribing more than 17,600 books. And because more Web sites keep adding reCAPTCHA every day, says von Ahn, an assistant professor of computer science, and the rate of transcription is growing. Other authors included computer science undergraduate Benjamin Maurer, graduate students Colin McMillen and David Abraham, and Manuel Blum, professor of computer science.
Perspectives Helps Secure Web Transactions
The growth of shared Wi-Fi has increased the risk of eavesdropping on Internet communications, but a low-cost system developed at Carnegie Mellon can thwart these so-called "Man-in-the-Middle" attacks.
The system, called Perspectives, was created by David Andersen, assistant professor of computer science; Adrian Perrig (CS'99, '02), associate professor of electrical and computer engineering and public policy; and Dan Wendlandt (CS'08), a Ph.D. student in computer science. It's been incorporated into a free extension for the Firefox Web browser that can be downloaded at www.cs.cmu.edu/~perspectives.
Perspectives employs a set of friendly sites, or "notaries," that help authenticate Web sites for banks, online retailers and other transactions that need to be secure. The notaries independently query a target site; if any notary receives a different digital certificate than the others, users are warned that their connections may have been compromised. Perspectives also can protect against attacks that could exploit recently discovered flaws in the domain name system (DNS) and cause Internet service providers to connect users to malicious sites instead of the legitimate sites they're seeking.
The work was supported by Carnegie Mellon's CyLab under grants from the Army Research Office and the National Science Foundation, as well as by the Department of Homeland Security.
--Compiled from staff reports