Discuss as:

Software traces faces through time

How the software works: "Exploring Photobios" from Ira Kemelmacher on Vimeo.

Leave it to the computer scientists to turn baby pictures into a slick animation that traces faces through the years.

The technique is already being put to use on Google's Picasa photo-sharing website as a feature called "Face Movie."


"I have 10,000 photos of my 5-year-old son, taken over every possible expression," Steve Seitz, a computer science and engineering professor at the University of Washington and an engineer at Google's Seattle office, said today in a news release about the research project. "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."

Seitz and his colleagues have already started down that road, thanks to the university's "Photobios" project. UW researcher Ira Kemelmacher-Shlizerman is due to present their research next week in Vancouver, B.C., at a meeting of the ACM Special Interest Group on Computer Graphics and Interactive Techniques, or SIGGRAPH.

Rapid advances in image recognition
Photobios take advantage of rapid advances in automated image recognition and tagging. In the past, such advances led to the development of Microsoft's Photosynth technology for re-creating clickable 3-D scenes from a "cloud" of images taken from many different angles. (Microsoft and NBC Universal are partners in the msnbc.com joint venture.)

Picasa's Face Movie feature takes that one step further by building in face recognition and name-tagging.

"This work provides a motivation for tagging," Seitz said. "The bigger goal is to figure out how to browse and organize your photo collection. I think this is just one step toward that bigger goal."

To build a Photobio, you'd start with a collection of photos showing the same person, whether they show your daughter or George W. Bush. You can arrange the photos chronologically, or specify the beginning and end points. The software automatically identifies the face and major features, lines up the eyes and morphs smoothly from one image to the next. Automated morphing is one of the key reasons why the results are so easy to produce and easy on the eyes.

"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. "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."

Better 3-D avatars
Kemelmacher-Shlizerman and Seitz are already working on the next step: taking a collection of photos and turning them into a movable 3-D model of a face. They'll be presenting research on that topic this fall at the International Conference on Computer Vision in Barcelona, Spain.

The researchers say such models could be used to create more realistic animated avatars for use in video conferencing or game play. More accurate face modeling also could well lead to improved face-recognition programs, for personal use (that is, sorting through the pictures on your hard drive or Facebook friend list) as well as for security applications (that is, matching up a photo taken at a security checkpoint with a database of images taken from different perspectives).

Does that sound cool, or scary? Feel free to weigh in with your thoughts on face-tracing and other image-recognition applications you'd like to see.


In addition to Seitz and Kemelmacher-Shlizerman, the authors of the SIGGRAPH presentation, "Exploring Photobios," include Eli Shechtman and Rahul Garg. The research was funded by Google, Microsoft, Adobe Systems and the National Science Foundation.

Connect with the Cosmic Log community by "liking" the log's Facebook page or following @b0yle on Twitter. You can also check out "The Case for Pluto," my book about the controversial dwarf planet and the search for new worlds.