Last week I wrote a story about “
Microsoft’s Research On Captioning Photos Automatically." I argued that this technology is a long way from being of much use to stock photo customers who trying to find useful photos for their projects.
However, after reading
Paul Melcher's story on Google Photos, “a new, standalone product that gives you a home for all your photos and videos” it occurs to me that helping users find photos may not be what Google, Microsoft and Yahoo are really trying to accomplish.
They could care less whether someone who needs a great image of a particular subject will, or will not, be able to find that image quickly and easily.
There is no money in that for them.
They are interested in MINING DATA from the images being posted so they can learn more about the people taking and looking at these pictures. Then they can sell that information to advertisers.
That’s where the money is.
Melcher pointed out, “And photos provide a huge amount of extremely valuable data. In fact, as content recognition gets more accurate, researchers are actively developing tools to better understand what is in the 1.8 billion images we share every day, something we heard over and over during the last
LDV Vision Summit.”
Why Should Photographers and Stock Agents Care?
We know we have a problem. We haven’t found a technological way to surface the images that will best inform and explain a news event to someone, or those that will be useful in motivating someone to buy a product or service. We’re not even close. Visual Search, the way it is being developed, will be of little help.
To be useful for our purposes Visual Search needs to be able to take the hundreds, or thousands, of images identified through keywording, or the content mining processes being developed, and take all of that a huge step further. It must be able to apply a human editor’s ability to bring the most appropriate to the top of the search returns.
If you’re looking for an image that will be useful for commercial purposes it is important to be able to make the distinction between
a woman's hair and a cat. On the other hand if you’re just trying to deliver ads to cat lovers then sending them to the person who took this picture, or anyone who views it, may be a good idea. A certain percentage of the ads delivered will be a waste, but no one will knows and some of the ads will probably get to the right people.
The Value Of Editing
Back in the 80s and 90s stock agents went through their collections and laboriously selected what they believed were the most useful images they had to offer. Then they placed them in print catalogs. It turned out that despite the fact that they had millions more images in their collections the vast majority of images licensed were those that appeared in the catalogs – even when there were hundreds of catalogs floating around and tens of thousands of catalog images to choose from.
But when the Internet made it so easy to post many more images online (no print costs for paper catalogs), and the prices customers were willing to pay dropped to the point that it was no longer economic to pay for all the editors, it became the buyers problem to dig through huge collections to find something useful.
Buyers are going to need, and seem likely to increasingly demand, edited collections. Visual search will be of little, if any, help in producing such collections. The trick will be in finding a way to offset the expense of the editing.