Join 3,497 readers in helping fund MetaFilter (Hide)


Where's my image?
August 21, 2008 8:30 AM   Subscribe

TinEye is the first image search engine on the web to use image identification technology.

Every day TinEye's spiders crawl the web for additional images. Using sophisticated pattern recognition algorithms, TinEye creates a unique and compact digital signature or 'fingerprint' for each one and adds it to the index.

When you want to find out where an image is being used on the web, you submit it to TinEye. The attributes of the image are analyzed instantly, and its fingerprint is compared to the fingerprint of every single image in the TinEye search index. The result? A detailed list of any websites using that image, worldwide.
posted by monospace (34 comments total) 9 users marked this as a favorite

 
Well, the site may be the first stop for those looking for infringers. Obviously.
posted by bz at 8:34 AM on August 21, 2008


This is for porn, right?
posted by afx237vi at 8:41 AM on August 21, 2008 [5 favorites]


Not just the first stop for copyright holders, but a stop for nobody but them. I mean, I guess I could look for versions of Mona Lisa, but it doesn't seem like there's a lot of demand for that.
posted by DU at 8:47 AM on August 21, 2008


Do you have to create an account to search?
posted by Mitheral at 8:48 AM on August 21, 2008


Yep. garbage@mailinator.com / garbage

I think it's pretty interesting, even if it's not entirely there yet. It was 2/10 on images I tried and thought it would be likely to find. These two (seems you'll have to be logged in to see the results, even). In the latter case, it finds a couple that are pretty nontrivial variations.
posted by Wolfdog at 8:52 AM on August 21, 2008


Yeah, you have to create an account to search. I'm not sure why, since it's a public beta. (This might have made sense when the site was closed off to new users.)
posted by chunking express at 8:53 AM on August 21, 2008


And yeah, when it works it's straight up voodoo magic.
posted by chunking express at 8:54 AM on August 21, 2008


The "compare" gadget is really smart about lining up the matching bits.
posted by Wolfdog at 8:57 AM on August 21, 2008


I'd be interested in knowing how they did the nontrivial variations. I accidentally chose an altered version of a famous image and it still found the originals, including cropped ones.
posted by DU at 8:57 AM on August 21, 2008


Since when is posting corporate propaganda oke on MF?

I tried TinEye many times, because image search is often very important for my business. The search engine hardly found anything. "Not in index" was it's usual reply.
posted by homodigitalis at 8:59 AM on August 21, 2008


Thanks Wolfdog.

Not too surprisingly it finds 666 matches to this kitten with gun image. I actually thought it'd be more but I imagine there is a lot of leeching going on. More impressively it finds 8 matches for the original of this 'shop job.

I like that you can link to the searches. They should do a image search crawler like webcrawler had.

homodigitalis writes "Since when is posting corporate propaganda oke on MF?"

Since it's cool.
posted by Mitheral at 9:02 AM on August 21, 2008


Not just the first stop for copyright holders, but a stop for nobody but them.

It can help copyright infringers too. Let's say you wanted to find a larger version of an album cover, you could use this to find a larger version.
posted by burnmp3s at 9:07 AM on August 21, 2008 [2 favorites]


with any luck, this will, in time, do away with the "help me identify this [usually shitty] painting" threads on AskMe.
posted by wreckingball at 9:19 AM on August 21, 2008


is a single press-release text cum link enough for an fpp?

the idea is good but it doesn't really work for me. it didn't find four images I know are used on other sites.
posted by krautland at 9:21 AM on August 21, 2008


Excellent idea, burnmp3s. I just tried to find a printably larger version of this excellent Solidarity poster. No luck, but I note that it finds other images that include the text that's partially obscured in the original. I'm impressed.
posted by echo target at 9:22 AM on August 21, 2008


Wow, I really think a lot of things are excellent today. Have I mentioned excelling?
posted by echo target at 9:23 AM on August 21, 2008


Also good for finding an original of an image that has either been photoshopped or branded with a web site's logo.

Disappointedly this (NSFW) didn't find any of the original.
posted by Mitheral at 9:36 AM on August 21, 2008


I find the name to be a strange choice. I mean, if a "tin ear" is a deficiency, why would a "Tin Eye" be any different?
posted by horsemuth at 9:45 AM on August 21, 2008 [1 favorite]


TinEye is put out by Idee which is a tiny little software company in Toronto run by some cool people.

Their actual business is scanning print publications looking for copyright infringement - their technology for analyzing magazine pages is amazing. They can do matches against very low-res images and they they have one demo where they separate a magazine cover into over two dozen source images used to create a composite. TinEye is cool but honestly, it's just the tip of the iceberg.
posted by GuyZero at 9:45 AM on August 21, 2008


I think they ripped off the name "Tineye" from the Mistborn series written by Brandon Sanderson.

Wiki - Mistborn:Tineye
posted by Nerro at 9:53 AM on August 21, 2008


The hottie has a Mac. What's not to like?
posted by KokuRyu at 10:09 AM on August 21, 2008


You could also use it to try and verify that the image you're about to use really is public domain. Or find a larger version of a free image.
posted by Harald74 at 10:26 AM on August 21, 2008


What amazes me is that it turns out that in addition to the 2000+ myspace pages that use my picture of some punk kid at fremont fair, it turns out there are some people who actually went to the trouble of re-hosting my image, instead of just hot-linking to it like everyone else. I could actually file my first DMCA take-down request. 'cause that'd be so punk =p
posted by nomisxid at 10:38 AM on August 21, 2008


I was thinking about using it for larger album covers, but noticed it couldn't find bigger copies for some fairly well known albums and gave up pretty quickly.

Then I decided to try a parody album cover someone submitted to one of my websites to see if it was getting copied by site scrapers.

The results

Even with some altercations to the shapes on the original cover it could still pick out enough of the original to spot the originals. Interesting algorithm at the very least.
posted by inthe80s at 12:26 PM on August 21, 2008


Metafilter Logo (have to be logged in to TinEye).
posted by fings at 1:21 PM on August 21, 2008


nomisxid writes "it turns out there are some people who actually went to the trouble of re-hosting my image, instead of just hot-linking to it like everyone else"

Well it's the polite thing to do.
posted by Mitheral at 2:06 PM on August 21, 2008


oh good, I can finally check to see what dirty pictures there might be of me out there.

!
posted by diablo37 at 2:09 PM on August 21, 2008


Politeness is the new punk.
posted by nomisxid at 2:20 PM on August 21, 2008


Mitheral: Your comment doesn't make clear whether you know that that image is an outtake from this. [hq][photog]

And inputting it into Tineye, I must say, my! Yes, that is a fetching image.
posted by dhartung at 4:05 PM on August 21, 2008


Ya that's what I was hoping tineye would find. Considering some of the sample searches I figured it had a chance. You can see on pg 39 of your search that it does manage to find some pictures of posters taken at oblique angles which is pretty amazing. Page 38 has a link where the image has had the girls magic wanded onto a different background.
posted by Mitheral at 7:37 PM on August 21, 2008


Which will happen first, getting bought by Google, or getting a dwarf fetishist in trouble for accidently downloading child porn?
posted by BrotherCaine at 9:33 PM on August 21, 2008


114 results.
posted by squalor at 10:48 PM on August 21, 2008


This is pretty awesome. I thought I'd get a lot of false positives on a piece of generic NWC art, but no ....

Also, it is fast. I thought it might churn away for a little while....
posted by Rumple at 11:44 AM on August 22, 2008


GuyZero beat me to it. I flat-out heckled Idée at a BarCamp two years ago and, within 15 minutes of further demonstrations, I was turned completely around and am now a believer. It is beyond amazing what their main software application can do. I was told that some magazines just do not bother preclearing images from large stock agencies with which they already have accounts; they use what they want, send Idée a PDF upon publication, and get a bill for what they used.

The software can take a scan of a thumbnail in a newspaper (e.g., a columnist’s headshot), with its attendant line screen and moiré, and find the original photo from which it was cropped.

Now, TinEye (indeed, a curious name) is a slightly different technology, but only slightly. As with all indices, it can’t find what it hasn’t indexed. But they have phenomenal technology and this thing will very likely get much better in no time at all.
posted by joeclark at 2:32 PM on August 22, 2008


« Older "Multitasking messes with the brain in several way...  |  "Grandmaster" Ashida Kim has a... Newer »


This thread has been archived and is closed to new comments