You know it’s coming…you are just not sure what to do about it. A few weeks ago, Apple released the new version of their personal DAM called Iphoto. Besides being one of the worst photo organizing application, it came out bundled with a little gadget that allows it to recognize faces and tag them accordingly.
The technology is not really new. Picassa has offered the same, on line, for quite a while. The principle is rather simple. The computer recognizes where faces are located in a picture and asks you to tag it appropriately. After a few examples, the application takes over and continues automatically based on what it has learned from your input. It will still make some errors, which you can correct, but overall, it will be pretty efficient.
Myheritage.com, a family history website has offered the same technology for a while, with a twist. It tries to match your face with a celebrity and tells you who you look like.
Those are two consumer-oriented usage. What surprises me is that no photo agency search engine has even remotely tried to apply this technology to their work flow. Especially in the editorial world, where 95 % of the searches of people are made with proper names. A photographer shoots Bill Gates at an event, the image is processed automatically to add his name to the IPTC field and added to the database. What a time saver !!!. Or , you finally find out the name of that woman sitting next to George Clooney at the last Miramax party. You upload her picture with the right info and hop! your whole database is updated in seconds.
Now, lets take this a step further. You just figure out what species that bird is. With the same technology, it can update any and all images with the same bird. Even better, you take a picture and it will scan the internet to find out what species it is and automatically add it to your images. This could be done for almost anything in your images. On the search side, a user can upload a generic image of a bird for which he has forgotten the species name and the database will return all the images with the same species in it, along with its name.
Currently, photo agencies spend a fortune on key wording. Some even have in house departments that keep on growing, as they process more and more images. Others give their content to be key worded by batteries of 9 to 5 keyworders based in India or other low wage countries. No one really mentions it, but this has become one of the highest added cost of images processing since they have become digital. A whole underground world, with its tight rules and regulations , its specialized software and its priests. It has almost brought key wording, the act of adding words to an image, to a science. Or, at least they would like it to be. They have “structured” vocabulary, words you can or cannot use, “standards” and other super secret sauce that you should respect if you want to be successful. They control a broad range of scary sounding anagrams, like IPTC, XMP, Dublin core, that they spend hours discussing during closed door sessions around the world. The more complicated, the more esoteric, the more people think it is important, if not fundamental.
They usually sit next to the IT guys and exchange complicated sounding words with looks of complicity while the rest of the room looks upon them with complete blindfolded admiration.
But all this is soon going to change. Besides the new steps of auto tagging, image search is coming out of its infancy. You currently also have similar search, color search and exact image search, giving the meaning of an image its rightful place, next to its description.
Why don’t we see any of these more often ? probably because image key wording has generated its own business, and jobs, a bit like Microsoft has created help desk jobs because it crashes all the time. It has grown out of a shortcoming. It plugs in a hole between the photograph and the person looking for it. It has replaced the fame knowledgeable researcher that agencies used to have with a generic obtuse answering machine. Image Key wording has now become an evil growth on the side of the photography business, managed by librarians who would like you to believe that they are a solution. The amusing part is none actually access you database history to see what words are actually being used by your clients and how successful the results are. For them, and the rest of the key wording industry, if a client can’t find the right image, it is because they are not using the right keyword.
A bit like stock photo agency editors who decide what image should make it through your system without ever looking at what images are actually published, the key wording industry has absolutely no relations with actual users. Nor do they seemingly care. They believe the users should be keyword friendly, and not the opposite.
Image search is going through a long awaited evolution and is now beginning to offer the proper tools to match a need to an offer and those who will win will be the ones that understand how to apply this technology to better serve their clients.