Face detection is undeniably clever. Born out of the slightly creepy biometrics industry, it's way of identifying the characteristic shapes of the human face in an image. The camera can then set the focus and exposure to reproduce these faces perfectly in your photos. Early face detection systems were pretty slow and unreliable, but today's cameras can pick out faces quickly and reliably, identifying them with a rectangular frame which can even follow them around if they move or if you change the camera position.
Face detection looks like it works, and it sounds like it ought to be useful, but does it really bear close scrutiny?
For a start, does it really produce better pictures of people? Yes, it ought to, of course, but has anyone actually demonstrated that it does? After all, a camera without face detection is perfectly capable of finding the correct focus point (usually the subject nearest the camera) and the correct exposure (producing a pleasing rendition of the whole scene, and not just a single face).
There ought to be a word for technologies like this, which so obviously ought to work by the very principles involved that no-one bothers to prove it.
But there's more. What happens if there's more than one face in the frame? Camera makers proudly list the number of faces their cameras can identify in a single frame, but don't address a rather obvious and central problem: which one is the camera going to pick, and what can you do about it if it's not the one you want? A few cameras allow you to change the 'chosen' face (laborious and time-consuming), while others may enable you to register 'favourite' faces and give them priority in the future, but already we're moving away from spontaneous, casual photography into something much more complex and involved.
And that's before we've even started on smile and blink detection, features which are now becoming common on digital compacts. Again, it's an idea which has such obvious merit it's seldom questioned. But look, here's how it works:
With smile detection, the camera won't take the picture until your subject is smiling.
Nice idea. We've all got pictures of people frowning, grimacing or looking the wrong way when we wanted them to be smiling at the camera instead. Normally, you'd take a few shots to be sure (it's what digital cameras are good at - instant playback). But with smile detection, the camera doesn't actually take the shot until it sees what it considers to be a smile. Now even assuming the camera can get this right, it produces another, much larger, problem. You don't quite know for sure when it's going to take the picture. And if there's one central, crucial thing about photography it's timing. You absolutely have to know the camera's going to take the picture at exactly the moment you want it to. Without that, you're not a photographer, you're a passenger. A digital camera's shutter lag is bad enough, but face detection adds unpredictability to delay.
With blink detection, if the camera detects closed eyes it takes another shot a moment later.
Again, it's a nice idea. If you actually need it. But here you don't quite know if the camera's finished taking pictures, and it's not very clear what's going to happen if it's a group shot and one person is blinking but the rest aren't.
Some makers are taking it further still, with self-timer modes which only start when a new face (yours) enters the frame. The idea is that it gives you time to get in the frame yourself and doesn't start the countdown until you do. But what's so hard about ordinary countdown timers? You know what you've got to do, you've usually got more than enough time to do it in, and you know when the camera's going to fire. Why make it complicated? (Except, perhaps, to sell more cameras.)
The point about face detection, smile detection and blink detection is that they all leave you and your subjects uncertain about what the camera is going to do, when, and why. Many of the best portraits are candid shots taken in unguarded moments, and any imperfections are part of their charm.
What face detection, smile detection and blink detection do is attempt to perfect portrait shots using a set of perfectly reasonable-sounding processes which, however well-meant, are likely to turn your people shots into deadened, uncertain and uncomfortable encounters.
Surely what we all want is a lot simpler: a camera that it does exactly what you expect it to at exactly the moment you want it to. Face detection technologies, however reasonable they may sound, are taking us in exactly the opposite direction.
