You know how it goes – you want to find a photo of the Mona Lisa, so you go onto Google Images, you type in Mona Lisa, and you get hundreds of results. But what if you have an image, and you want to find out what it is of? Or what if you want to find out where else on the world wide interwebs this image is used? Cue Idee’s Tineye.com, an image search engine.
We decide to take it for a spin…
A website that proclaims that it “does for images what Google does for text”, is not going out of its way to hide its ambitions, but this is the claim that meets the internet surfer that lands on Tineye.com. Simply put, Tineye.com consists of a search engine which asks you to upload a picture or provide a link to an image. You can also install a plug-in for FireFox which allows you to right click an image and choose Tineye search from the dropdown menu.
Start a search, and Tineye.com will find how many times this picture has been represented on the internet – a feat accomplished by comparing images pixel-by-pixel, and creates a so-called “digital fingerprint” for each image found, determining which pictures are similar to the one you searched for.
The process is supposed to work even through the image is cropped, resized or photoshopped. Problems arise if the image is changed too drastically, relies on an outline, or changes the colour scheme too much.
The website itself raises a few warning flags that not all is well in the state of TinEyemark – most notably its limited search index. At the time of the introductory video, there were a little less than 500 million images in the search index, meaning that the image you have uploaded is only compared to a fraction of the pictures available on the internet – by the time we got to the site, there were just over 900 million images, and the company behind Tineye.com is aiming at entering billions more.
For now, however, the limited search index meant that the first photo I entered into the search engine, did not return any results, even though it was sourced from photocritic.org.
The second search was of a map of the world, which returned 33 results. Type in “map of the world” in Google image search, and you will be blessed with more than 60 million results. A quick comparison of the Google and Tineye results shows that the colour schemes for Tineye are much closer to the original picture, whereas Google has a much broader range of definitions for ‘map’.
The final search was done with a detail of the head of Venus (picture on the right, the original used for this review is here) from the famous painting “Birth of Venus” by Botticelli. Of the 123 results returned, there were originals, book covers, reproductions, and some slightly larger details which included her shoulders and torso, yet not a single result came up with the full painting – To be expected, perhaps, but it does show that this technology works better when the original search terms are not cropped.
So.. What’s the point?
Ultimately, we were left wondering what the use of this website might be. Google is fine for finding information which you do not already have. If you already have the picture on your desktop, or you have the link to it on a website, you clearly will not need to find the same identical picture again.
What it might be useful for is for photographers to find whether their pictures are being used in contravention of copyright. (and for that to be the case, the search index needs to be significantly larger).
As a concept, Tineye is pretty nifty technology, but how often have you found yourself wondering ‘hey, that’s a nifty picture, I wonder what it is of’? For us, that’d be ‘never’, which sort of ruins the point a bit…
This article was written by guest writer Meke Kamps
Do you enjoy a smattering of random photography links? Well, squire, I welcome thee to join me on Twitter - Follow @Photocritic