Be sure to read Paul Melcher’s story in his
Kaptur Magazine about where image recognition software is headed.
Keywording or tagging can take a lot of time. Most image creators would like to avoid it.
All the major Internet and software companies are trying to find automated ways of doing this that will make the millions of unkeyworded, or poorly keyworded, images findable. Millions of dollars are being spent on a variety of projects.
Flickr has recently launched a new tool to make sure that the 11 billion images it stores are fully indexed and searchable. But is this necessarily good for image creators or customers?
Many Flickr image
creators are unhappy with this new tool. They are finding that many of the tags are inappropriate. One user found that his close-up or a squirrel was tagged as “cat” and “pet.”
Some might say this isn’t that bad as long as the image also has the keyword “squirrel.” But if a customer is searching for pets and a lot of the images delivered don’t fit the criteria will the customer stick around, or go somewhere else? With Flickr this could be a big problem because there are probably well over 10 billion images that didn’t have much in the way of tags until this new tool was introduced. Now many of them may get tags that don’t quite conform to what the image is all about.
If that is the case when someone is searching for a particular subject instead of getting 1,000 returns they may get 40,000 or 50,000 returns, many of them not on target. But, the real problem is which returns will be delivered in the first few hundred because that is all any customer is going have the time or patience to review.
The Internet already has way more images tagged with any word you can think of than anyone has time to review. The question is how to bring to the top of the search returns the most appropriate images for a particular customer. Every individual looking for a specific subject, and using the same words, will most likely have a different idea of what the image would look like that will best satisfy his or her needs. More words or more images will not necessarily solve this problem
One of the great fears for image creators who have been putting a lot of effort into good tagging, and getting good sales result, is that customers will get so many inappropriate images in their searches that they will turn to other ways to find the images they need.
As auto tagging grows it could force customers to turn to organizations that create tightly edited specialist collections and provide custom research.