Face it Steve, you need help

March 8th, 2010 by

Steve Jobs by Dunechaser @ Flickr

There are some things in life that are always going to get you in trouble. No matter what you say to try to cover them up, or how much you apologize. Blaming alcohol, stress or even Steve doesn’t work. Trust me.

It’s not a big deal – if we are to believe Hollywood, it happens all the time. The slip of the tongue can happen if you’ve been with your partner for a day or a decade. One day, when you least expect it, you’ll accidentally call her by your ex-girlfriend’s name.

Smugly, I had always considered myself better than this. I like it when people remember my name, so I do my best to remember theirs. Especially my girlfriend’s.

Like I said, I blame stress (or more precisely long hours and a lack of free time). Stress and Steve Jobs.

An Apple a Day
It all started three years ago, when I bought my first Apple. Despite the social stigma that seems to come with the territory, I’m a very satisfied customer (check these comments on Engadget after a major Apple release to see how a blog post looks when burnt to ashes).

Even so, I wouldn’t class myself in the much-maligned category of “Apple fanboy”: I don’t have the apple logo tattooed on my forehead, nor do I worship his holiness Steve Jobs. And if one of Apple’s products is rubbish, I  feel no obligation to defend Steve and his followers out of blind loyalty.

At MacWorld 2009 Apple introduced the iLife 09 suite, which includes new versions of Garage Band, iMovie, iWeb and iPhoto. The latter program convinced me that I was in need of an upgrade. I don’t consider myself an Annie Leibovitz or Robert Doisneau, but from time to time I’m lucky enough to catch some decent shots.

Given that my previous version of iPhoto was way out of date (2006!) and in light of my sprawling, out of control archive, the new features such as geotagging, Facebook/Flicker synchronization and automatic face recognition seemed really promising. With little free time on my hands, I said to myself: why not?

A Case of Mistaken Identity
Alas, the answer to that question came soon enough but also, regrettably, too late. Putting aside the annoying issues of the social network syncing function (slowness, double tagging, uploading errors, impossibility to switch off), the real disappointment was the much-promoted Face Recognition feature.

It is supposed to work like this: you feed in your photo archive and tag a few familiar faces; then an algorithm scans the rest of the pictures for all the recurrences of people tagged, and categorizes them to help you filter your collection using real people as a search criterion. Sounds easy in theory, but in practice, I assure you, it’s not.

First of all, the software only works well if the collection is tiny and the photos are stored locally. Feed it with 30 GB of pictures on an external USB hard disk (not an unusual situation given the diffusion of affordable DSL-R models these days) and your computer will be stuck in that first scan for hours. As one might expect, the agony continues every time new pictures are added, with repeated scans keeping the CPU unacceptably busy for what seems like an eternity.

That would probably be a bearable price to pay if the face-recognition performed flawlessly. But unfortunately, people with completely different facial features and even gender are easily confused by the program. As most of my friends are not cross dressing clones, this is rather disappointing. Things improve somewhat the more manual tagging a user does – but if I have to do it myself, what exactly did I pay for?

The raw truth is that algorithms are nowhere near as good as real people at recognizing human faces and emotions. Something that for me is absolutely crystal clear seems, at times, impossibly difficult for Face Recognition to figure out. It’s not that the programs are useless, it’s just that to be effective they need the help of human instinct and insight.

The solution to this problem has a name and that name is real-time crowdsourcing.

So, dear Steve, if you are reading, I implore you to think about it. For my sake. You have no idea how uncomfortable my couch is. Or how hard it is to explain to your girlfriend how you confused her with your mother.


  • Pingback: Alio News Letter….March 2009 issue: crystallography – Windows Live | Diffraction Material Geek

  • Renato

    Being a linux user I use Picasa as my photo manager. The new version includes a face recognition option, which I have been recently checking out. Like iPhoto's feature it takes forever to scan the photo collection, straining the CPU to a good extent. However I have to say I'm extremely satisfied of the result. The automatic tagging of the faces is reduced to the minimum, and most often the user is presented with a small menu with the names of those featured in the previous pictures (Facebook-style, more or less). This avoids incorrect coupling of names and faces by the computer. It is time consuming, but never as much as checking every pic for incorrect tagging, removing the tags and reassign them manually. If I have to make a remark, I would point out that the algorithm works TOO good, picking up every single face in a crowd (quite irritating when you have pics of a sport event), faces printed on a poster or the head of a statue in the background.

  • musta

    rrrrrrrrright… picasa is just great at that at least on win and linux… and is free :)


<<

>>