People pour more multimedia into the cloud every day. But if you want anyone to find it and make use of it, it needs metadata. Tags, folders, classifications, descriptions, and titles. In a word, words. Search engines algorithms are classically s(text)→text. Several image search offerings perform very well at s(text)→images, but they rely on the words around the pictures.
TinEye is different. It hosts a true s(image)→images algorithm for web image search; to my knowledge it’s the first of its kind (publicly available). Right now, the engine itself is more useful for those searching for copyright violations. TinEye searches for pictures that look more or less exactly like the input so searching with a picture from a personal album will most likely produce nothing. But that won’t be true forever.
It won’t be long before the search space becomes s(image)→text. Want to know what kind of architecture some pillars exemplify? Or who painted that picture hanging on the wall? Take a picture of a church in Chile, a tomb in Argentina, or some random structure on a wharf, upload it with your phone and wham: instant tour-guide.
The information is in the cloud; we just need to get smarter about locating it.
0 Response to “The <Image> is Out There”