Software Magic: Recognizing People in Photos

What automated face tagging in Photoshop Elements 8 can, and can't, do to help organize digital photos.

By SHARE
PSelementspeoplerec_1028tech.jpg

One of the best things about digital cameras is also one of the worst. Without the cost of film and developing, digicams enable amateurs to shoot as many photos as they want. That's a huge benefit—but also a huge burden as consumers get overwhelmed with thousands of snapshots buried on PC hard drives. There they sit, mocking us with their promise of great memories and beautiful images hidden among the anonymous, the underexposed, and the blurry.

[Slide show: The Best of What's New in Windows 7.]

A new tool has emerged to help consumers tame those virtual piles. Software now promises to automatically identify and tag photos with the names of people who appear in each shot. In theory, it's a lot easier to find a great photo of Brian, Suzie, or both—and even when they were kids, visiting Yosemite with Grandma.

Recognizing familiar faces is one of the most promising new features of modern photo software, including iPhoto for Apple users, Photoshop Elements 8 from Adobe, and the more basic but free Picasa software from Google. But does it work? To get a sense of how the technology performs, we've been running what Photoshop Elements calls "people recognition" through its paces. Here's what everyday users can expect from the latest twist in organizing personal photos:

It works. The software is toddler-age tech, meaning it has the smarts of a 2-year-old child in spotting familiar faces. Well, that's not too bad: A 2-year-old can recognize a lot of familiar faces. Once a user identifies and "tags" a face in one shot, Photoshop Elements finds the same face in other photos, although it also misses quite a few. Surprisingly, it can even find faces that have changed over time, such as a photo of a preteen boy when he was just, well, a 2-year-old. "It's really quite remarkable what face recognition can do," says Barbara Brundage, author of Photoshop Elements 8: The Missing Manual.

[Once found, precious shots can go to sites that make it easy to share photos.]

It takes time and effort. The software doesn't produce its magic without a lot of stirring. A user has to stay at the computer, correcting the program's mistakes as it tries to identify familiar faces. The more the software is trained, the better it works. For a library of 1,000 photos, it might take about an hour of working with the software to identify all the faces, says Mike Iampietro, a product manager at Adobe. But the amount of time can vary, depending on how many people appear in the photos. For snapshooters who mostly focus on their kids, the software won't have to work as long—or as hard—to identify all the faces. Also, after importing photos, give Elements some time to do its initial processing of photos, such as identifying which photos even have faces. It can take some time before it's ready to ID people.

Strategy can help. For a large catalog of 10,000 or more photos, which is not that unusual these days, users might want to study how their software works before spending hours tagging faces. Photoshop Elements, for example, offers multiple routes for identifying faces. Users can let the software run through all of the photos and help it ID all the faces. Or users can double-click on a particular face in a photo and prompt the program to try to find every photo where that person appears. For people wanting to first find photos of their immediate family, the second route would get them there more quickly.

No cure-all. Identifying faces is a big help in finding photos. But the software misses a lot of familiar mugs, especially those taken from an angle or of someone in sunglasses. Plus, faces are only one step toward lassoing photos into some kind of structure. Finding shots of the kids at Yosemite requires more work. One shortcut would be doing a search on dates of when the family was at the park, then adding a "Yosemite" tag to those frames. Photoshop Elements helps by trying to further distinguish photos, such as small groups or those with one or two faces, and it also pegs photos as quality or blurry shots, among other "smart tags." Combining tags can quickly narrow photo choices to a select and attractive few.


TAGS:
photos
technology