By samoore (1251644871|%a, %b %e at %I:%M%p)
TinEye: Image Identification Technology
A useful new image search engine is TinEye by Idee (the self-described “visual search company”). TinEye, which is currently in beta, is the first search engine on the web to use image identification technology. In other words, rather than submitting a written query, you upload an image to find out where and how it appears on the web. Basically, it’s the exact opposite of your typical Google Image Search.
How It Works
According to Idee, TinEye creates a unique and compact digital structure for each image that is added to the search engine’s index. Idee appropriately refers to these digital structures as ‘fingerprints.’ Each time you upload an image to be searched for, its fingerprint is compared to every fingerprint in TinEye’s index. It’s important to remember that TinEye looks for the specific image you upload, not the content of the image. Therefore, TinEye cannot identify people or objects in an image. Likewise, it cannot find different images with the same people or things in them. It can however, find different versions of the particular image that you provide. Idee claims that TinEye has indexed 1,013,140,121 web images to date. The company states that new images are being added regularly, but an average rate is not given.
What Can TinEye Do For You?
Besides being fun to play around with, TinEye actually has some pretty useful features. Here are just some of the common uses of TinEye.
- Find websites containing more information about an image
- Attribute an image to an author or source
- Find high (or higher) resolution versions of an image
- Research the usage and placement of editorial or stock images
- Find modified or edited versions of an image
- Research corporate imagery or brand usage online
- Track the popularity or distribution of an image online
There’s actually an entire TinEye blog if you’re interested. If nothing else, there are some pretty interesting images that are worth taking a look at.
First of all, signing up for a TinEye account is ultra-simple and takes all of 30 seconds (depending on how long it takes to get that oh-so-annoying activation email). In order to test TinEye out, I picked a random image off my computer and submitted it. This particular image was of a red corvette that I had once used as my desktop background. Needless to say, TinEye found over a dozen different websites that held this exact image. So I decided to give TinEye more of a challenge. This time I uploaded an image of an Atlas rocket being launched from Cape Canaveral, which oddly enough I just downloaded from Ohio State’s library website. TinEye did it again! This time it returned eight different sites that use this image, with the OSU website being the number one result.
My next test was to search for an image of my friends that I had uploaded on Facebook several months ago. Oddly enough, TinEye did not return any results for the one image that I knew for certain was on the web. Also, I submitted one of those ‘sample images’ that were preloaded on Windows Vista. TinEye returned several pages of matching results. The only problem… not a single one of the results was a Microsoft website. As I continued to search for images (all of which I knew were downloaded off the web), I found that TinEye yielded results about 50% of the time. So an image search with TinEye can at best be considered hit-or-miss. Another major downfall of TinEye is that uploaded images have a maximum file size of 1 megabyte. This constraint rendered several of my searches useless. With that said, it’s a really cool concept that will undoubtedly improve over time. Just think, this kind of technology was reserved for the likes of the C.I.A. and F.B.I. just a few years ago. Now it’s online for the world to test out.
I can’t say that I’d actually use TinEye to research the usage and placement of editorial images, but I could easily see myself utilizing it in other ways. How often have you downloaded an image to put as your desktop background, only to find that it’s blurrier than you originally thought? Wouldn’t it be nice if you could easily find a higher resolution version of that same image? This is how I see myself using TinEye the majority of the time. It’s also extremely useful in finding out more information about an image. For example, when I searched for the Atlas rocket image, I had no idea what kind of rocket it was. Using TinEye, I learned the following in 1.780 seconds:
- it was an Atlas 120D rocket
- named Friendship 7
- which rounded the Earth three times
- on February 20th, 1962
- piloted by John Glenn
Obviously TinEye can serve as a convenient and useful research tool … when it works.