First introduced in May, Idée announced yesterday that TinEye, the revolutionary new image recognition & visual search engine is now in open beta. If you haven’t had a chance to mess around with TinEye yet, you owe it to yourself to head on over there and check it out.
First impressions of the TinEye visual search engine
I was genuinely impressed with the speed & accuracy of the new search engine that catalogs & retrieves images based on their visual similarities. Simply upload an image, or direct the tool to a URL where an image is hosted and TinEye will scour its database of over 500 700 million images to find images that are similar.
TinEye starts off finding images that are exact matches to the source image, but can also find similar images – for instance images where the original has been color adjusted, had text added or removed, been cropped and even slightly rotated. It’s pretty funny seeing all the PhotoShopped versions of the Leonidas (300), the Mona Lisa, American Gothic and more. The TinEye FAQ page has more details on what makes it work.
I’ve done a number of experimental searches, and some of my favorites were already cataloged as TinEye “Cool Searches”. it’s pretty amazing how far the software can go in recognizing the original image, despite lots of visual changes.
Then there are the TinEye widgets. Easily embeddable, I imagine these widgetized galleries will become a common addition to posts and forum threads in the near future. Allow visitors to browse through hundreds of similar images with-out leaving your post or site.
So what would you use the TinEye visual search engine for..?
There’s a lot of interesting things that come to mind when checking out this latest toy. Far more than fodder for entertainment, this sort of image recognition can be used today to check for illegal use of copyrighted imagery.
The conspiracy geek in the back of my mind whispers that tomorrow this same technology could be used to to identify & track citizens of privacy-ignoring governments.
Any way you look at it, it’s a pretty nifty tool, and it’s in open beta now so we can all check it out and talk about it’s features and it’s possibilities.
What do you think of it? Any first impressions? Ideas? Concerns?