Video: New tool animates Web photos, might raise privacy concerns

Researchers at the University of Washington have found a way to take hundreds or thousands of digital photos and create an animation of a person's face or body in mere seconds. The application can make a person appear to age quickly, or change their expression from say, happy to sad. Basically making a movie from static images of a face.

"I have 10,000 photos of my 5-year-old son, taken over every possible expression," said co-researcher Steve Seitz, a University of Washington professor of computer science, in a press release. "I would like to visualize how he changes over time, be able to see all the expressions he makes, be able to see him in 3-D or animate him from the photos."

This project is producing a very cool new tool. But it might also tap into serious privacy concerns. One can imagine, for example, being able to animate images from Facebook since photos are auto-tagged with people's names. Just yesterday a German regulator asked Facebook to disable this new photo-tagging software, over concerns that its facial recognition feature is an unauthorized collection of data on individuals, and may violate a user's privacy. Suggested automatic tagging, as it's called, scans through a user's photos and finds similar features to those portraits already uploaded and tagged. According to the New York Times, Facebook has already built an archive of more than 75 billion photos, and 450 million people have been tagged worldwide.

Consider another potentially contentious scenario: A future where a digital you is persuading the real you; where a communicating and animated clone could become a new advertising medium. Studies out of Stanford's Virtual Human Interaction Lab have proven a digital you can profoundly influence our decisions and behavior. The work is premised on the fact that we tend to be influenced by those who are similar to us, and nothing is more similar than the actual us. It's not hard to imagine the leap to advertising where suddenly you aren't just seeing any old stranger drinking a Coke but that ads tailored to you have "You" drinking that Coke.

The face animation project is funded by Google, Microsoft, Adobe and the National Science Foundation. It follows from another breakthrough four years ago where University of Washington researchers found a way to automatically stitch together photos of buildings taken by hundreds of tourists to recreate a 360 degree view, in near 3D, of say something like the Sistine Chapel. That work became Microsoft's striking Photosynth tool, which launched to standing applause at a TED conference back in 2007. (See video here.)

But the researchers say faces provide a greater challenge, because they do not remain as static as the Sistine Chapel. Over years, moments, they move, have expressions.  The researchers note that photo apps have greatly improved pointing out that Picasa and iPhoto have face-recognition tools. And as noted earlier, Facebook also has auto-tagging of people. It remains to be seen, however, how long they might be able to offer that advantage.

This new software can take Web photos or photos from our own collections that are tagged with the same name. It determines the face, the major features, nose, mouth, eyes, and then aligns them with similar features from other photos, thus recognizing the person. The tool then uses a simple cross-dissolve or quick fade in-out that provides a smooth transition and gives the appearance of motion. Or in the case of showing aging, the tool will chronologically order portraits and the result is a movie. Watch the video below as a Google employee's daughter ages 20 years in less than a minute!

"There's been a lot of interest in the computer vision community in modeling faces, but almost all of the projects focus on specially acquired photos, taken under carefully controlled conditions," Seitz said in a press release. "This is one of the first papers to focus on unstructured photo collections, taken under different conditions, of the type that you would find in iPhoto or Facebook." Again this raises potential red flags over privacy and access to such photos, especially when this is a step beyond "specially acquired photos" and "carefully controlled conditions."

Seitz and colleagues will present related research this fall at the International Conference on Computer Vision. This goes even further using personal photos to build a 3-D model of a face. Imagine very  realistic avatars, and what that could do to the video game industry or advertising industry (as noted earlier with the work of Stanford's Virtual Human Interaction Lab.) Or if this is not carefully controlled, imagine the potential for stolen identities.

Lead author Ira Kemelmacher-Shlizerman will present the research next week in Vancouver, B.C., at the meeting of SIGGRAPH, the Special Interest Group on Graphics and Interactive Techniques. Should be a fascinating discussion for all who attend.

This post was originally published on Smartplanet.com