Similar Image Search becomes mainstream
The ways of “cutting through the clutter” and surviving the information overload have changed: after trying to impose a structure on data by adding metadata, we quickly realized that was helpless and would not scale up if front of the explosion of the Digital Universe. Instead, we decided to (barely) keep head over water by using whatever techniques were available to search for the content we’re looking for in that sea of documents.
This has become relatively easy for documents, even scanned. By searching for keywords, OCR and the latest indexing technologies do a fairly good job at finding the information we need – provided we can make that search the right way.
But the next frontier lies in more complex, richer documents, such as pictures or videos. But there also, searching by content is finally becoming (relatively) mainstream.
I recently blogged about this research project where images can be searched for similar content, but this capability is now coming out of the labs. Gazopa, for example, is entering open beta stage.This plug-in allows you to search for a similar picture which has been indexed, or even create a drawing, and find similar images. Google Similar Images allows you to do the same, although the original image you search from needs to be indexed ahead of time. Even more interesting, Google’s latest Picasa version lets you index individual elements of a picture, such as faces, and then search for other occurences of that face throughout your collection.
Does it really work? I’ll let you test it for yourself – my personal experience has been relatively mixed. However, I have no doubt they will shortly improve significantly, causing a paradigm shift in the way we deal with information overload.
