2. Apple iPhoto ‘09

New Macs come with iPhoto ‘09 as their “free” photo editor. I wish we had the time and space to do a full write-up of this remarkable program, but for now suffice it to say that the hot ticket addition to the ‘09 version was Faces and Places. The Places aspect refers to the fact that photos taken with an iPhone embed GPS coordinates into the image’s metadata. With this, you can view a map showing where all the photos were taken—slick! If you don’t have an iPhone, you can click on the little “i” icon in the bottom corner of each image and use a Google Maps search to find your location. This obviously has nothing to do with face tagging, but perhaps soon you’ll be able to say, “Show me any photos I have of someone named Tom taken in London.”

The Faces function of iPhoto is a curious beast. Importing the 300-file album took 44 seconds, which sounds misleadingly fast. Keep in mind that iPhoto isn’t doing extensive facial analysis at this point. iPhoto does importing and preliminary image analysis at the same time, but the program is only looking for faces in general. It’s not attempting to find and group face matches yet. In fact, iPhoto puts most of the computation burden back on the user. I’ll explain.

After trying out all of these face tagging tools, I quickly came to see that brute force computation may not be the secret to face tagging success. If CPU technology and visual analysis software have evolved to the point where we can now achieve, say, 80% success on facial recognition instead of 30% five years ago, that’s a massive leap forward. But in a way, you’re still stuck in the speech-to-text trap, where a 95% accuracy rate still isn’t good enough to compensate for the time it takes to edit those 5% errors. It’s the editing process—your time doing hands-on, manual intervention—that kills you if there’s anything more than a handful of errors or false positives.

Apple seems to have a better grip on this fact than its competitors. The iPhoto Faces experience is based on user feedback (yes, that’s a match or no, that’s not a match) doled out in bits and pieces from the beginning, not a big pile of analysis that then leads to prolonged editing. Not until I really understood how iPhoto Faces operates did I realize that computer analysis time runs a far second behind user workflow time. If we can agree that the desired outcome with face tagging is to have tags applied to the faces you want in the photos you want, then it’s not the initial analysis that matters most—it’s the total time it takes to complete the tagging. This is why the user interface is so critical. The UI is the tool you use to do the real heavy lifting of tagging. Better tools yield faster results.