Tag, you're it! | April 11, 2006
As you’ll no doubt know, tags are essentially user supplied metadata. Tags are useful as a means of data recall, so you can tag your images as SXSWi and then a search on your images using that tag will bring back all the pics you took during the conference. However I find tags much more useful for data discovery, so by searching for the tag SXSWi on everybody’s photos, you’ll be able too see all the pictures taken at the event.
This is great in theory, but in practice tagging can very quickly become a chore. When you have one or two photos you want to tag, it only takes a minute or two to tag them. However when you take a lot of pictures like me, you start to get into scalability issues. If it takes you a minute to tag a photo and you’ve taken sixty photos, you end up spending an hour just adding meta data to your images. Not what you really want to be doing with your spare time.
I want to add descriptive data to my pictures so that other people can find them, but I often find that it’s just not practical, or enjoyable. One of the benefits of technology is its ability to reduce menial and repetitive tasks, but sometimes it has the oppressive effect, enslaving you instead. However I’ve always felt that a technological solution wouldn’t be far away.
When I was younger I remember seeing a science program where a University researcher created a program that could recognize famous landmarks. It used a database of images which it compared with the target picture in order to find a match. At the time it could only pick out a few major landmarks, and only if the picture was taken at exactly the right angle. However I remember thinking that it wouldn’t be long before the database got large enough to cover most major landmarks, and processors got fast enough that landmark or location recognition became common place.
Another thing I’ve been waiting for is geographically aware photography. With GPS receivers getting smaller, they can now be fitted inside a digital camera. Take a picture somewhere and the image can automatically be tagged with the longitude and latitude. Now if you were to combine this with a locations database you would know exactly where each of your photos was taken. Hooked this up with landmark recognition software and you could automatically identify and tag a large number of your pictures.
This may seem like science fiction, but Bath University have already done something pretty similar. You can take a picture of a location using a GSM phone, and because the phone is location aware it scans a database of local pictures and tries to find a match. I can’t wait for this kind of functionality to enter the consumer market.
So that covers location photography, but what about pictures of people? If you look at my SXSW photos, you’ll notice that most of them are pictures of friends and colleagues. It would be great to have these pictures automatic tagged with a location, but it would be even better if they could be tagged with the persons name as well. Well as it turns out, facial recognition software is actually fairly advanced and a recently launched service may have the solution to my tagging problem.
Riya is a photo service similar to flickr that allows you to upload, store and share you images. However it has one fairly major trick up its sleeve in the form of facial recognition software. When you upload your images you can tag them, and overtime the software in the uploader learns what your contacts look like and begins to tag them for you. I’ve not tried the service out yet as predictably it’s in beta, but if it does what it promises, one part of my tagging dilemma may be over sooner than I thought.
[Update:] Just found out that if you’re a Yahoo! US user you can sign up to ZoneTags and automatically geotag any pictures you take on your mobile phone. Cool.
Posted at April 11, 2006 9:24 PM