I work in image retrieval, the basic idea being finding images similar to a given query image. This can be posed in three different forms. First, find me images which are only near duplicates of the query image. Second, find me images that are visually similar (color, texture, etc) to the query image. And third, find me images that have the same concepts (birds, tree, ground etc) as those of the query image. All of them are interesting problems in themselves, with the degree of difficulty increasing from one -> three. Although there is an evident application to this, that being image retrieval itself, but I just saw this application that makes an interesting use of the first two solutions – Image Comparer. It allows one to refine similar images or pictures he has in his collection, saving us the trouble of finding similar images.
I installed and tried this software, but not unexpectedly was not impressed with the output. For sure it does not have the capability of finding similar images at semantic level, it does not even work at the visual level. Maybe its good for near duplicate images, which I am not really interested in. Nevertheless, I will give them kudos for the idea.