Eyealike, a video and image recognition company, has released an update to its ‘Eyealike Faces’ product that makes it easier for users to find their ideal match on dating sites. Eyealike Faces uses image matching technology to compare faces across databases consisting of millions of photos – the idea being that you can upload a photo of your ideal mate and get a listing of the best matches. Unfortunately, you won’t be able to try out a demo on your own – Eyealike only licenses the technology to third parties (a deal with a major dating site is in the works, so you won’t have to wait long).
The concept may be a little strange to some (the word superficial comes to mind), but physical attraction is obviously an important component for many people looking for a date. The new release of Eyealike includes “Face Levers”, which let users modify certain traits in the pictures they’ve uploaded. Looking for a woman with Angelina Jolie’s eyes but a slightly softer jaw? Just tweak the levers and bit and meet your future soulmate.
At least that’s what supposed to happen. The demo I saw wasn’t too impressive, but this was likely a result of a limited photo database of only a few thousand instead of the millions of images the technology is designed for. We’ll have to wait until the technology is incorporated into a major dating site to see if it is as effective as it promises.
Another technology company that offers similar image matching is Picitup, which didn’t work particularly well when we last reviewed it.
TechCrunch, you guys need a report bug/typo tool. I hate saying this publicly but best to get it before your RSS feed mailout goes out:
You have a typo in the headline. It’s “Perfect” not “Prefect”
Don’t mean to be the public downer, especially with first comment.
Thanks Chris. I actually caught the typo about one minute after the post went up. Too late, it seems
On another note, eyealike is a brilliant concept. I guess the next step is such an image recognition for the body.
It happens Jason, np. Talk to Michael about a report feature though because nobody likes leaving typo comments but they are needed if you don’t catch them yourself. I caught it in Twitter where your typo lies in limbo
Go to a bar or go to church be man and find a girl on your own things like this are goofy and signs of lack of confidence.
http://www.crunchnow.com
Wow this is insane, I am going to have to try it out.
brb finding angelina jolie photos..
http://easysumm...ey.blogspot.com
Yeah well I tried it to find matches for these gals, but no dice.
http://youtube....h?v=lLYD_-A_X5E
I guess the facial recognition stuff needs work.
i like the typos actually.
It is amazing to see that dating sites are coming up with a technology like that to find a match based on eyes you like. I run many dating sites and I feel this is NOT a useful feature at all. There are many scammers on dating websites which pose pictures of beautiful girls – are you going to waste your money by going after these unreal girls – just because their eyes in the picture match with your dreamgirl’s ?
A good dating site will come up with a feature which can identify scammers by comparing the pics with a universal database of scammers. That is what is really needed.
I have been working on similar problem in my Masters. Facial similarity is inherently a subjective problem. What person A finds similar may not be similar for person B.
However, this should not be a bottleneck as one can always define one’s parameters and work within them, which is what eyealike seems to be doing.
I wish them luck. As of now their results are not impressive but I think that they will be trying to improve it.
This is going to be another Me Too useless Web 2.0 application. Good concept but I have some serious doubts about its classification error rate which must be high.
I have done development in the domain of Computer Vision (and Machine Vision – very similar discipline) ie, image/video understanding & recognition for an automated industrial inspection system in the past. Most of these industrial vision systems (different types available today for different purposes) work well in their specific domains, since the images of items that the vision system is inspecting (in real time) are not that well diverse compared to an image database with millions of human faces which there is a huge diverse in features. Feature selections in face image matching is known to have a high error rate and very inaccurate.
The vision system development project that I got involved
in was for the semiconductor wafer manufacturing industry
to automatically detect & eliminate wafer/s which contain/s
defects during the lithographic printing and processing of
integrated circuits (ICs). These defects are so small which
you can’t see them (ie, you can’t visually identify these defects)
which they ranges in size from about 100 nm (nano-meters) to
around 800 micro-meters. Optical microscopy and scanning
electron microscopy (SEM) are used to capture the digital
images of these wafers on a moving lane as they progress
thru different stages in the production line. The vision system,
where SEM is its artificial eyes continually scan each wafer
as they’re passing by. If this artificial eyes vision system
detects a crack or defect in a wafer, then it is automatically
taken off/eliminated (via a robotic arm).
Now, these different types of defects are not that hugely diverse
which means that the artificial vision system tend to have a high accuracy
in its classification to detect/eliminate wafers which have cracks or defects.
So, vision system works well in industrial application in manufacturing.
I have serious doubts of how accurate computer vision in its application to
human face matching because of the massive diverse features available in these millions of different faces, which tend to affect its classification error rate. If it has a high error rate, then of course it will retrieve irrelevant faces which bear no similarities at all to the target query face.
No publicly viewable pricing means extra-ordinary costs. Otherwise I may have been interested.
Has anybody else noticed that the person most like Michael Arrington is Buster from Arrested Development (actor Tony Hale)?
You like the typos Michael?
I guess typos get attention which translates into pageviews. Worked to suck me in =)
Thanks for all the feedback.
I just wanted to clarify a few things that I saw written that may not be quite accurate.
Our pricing isn’t “extra ordinary” just because we don’t list our pricing model on our corporate website. I think most of you can appreciate why companies choose not to openly publish their rates – our rates are very reasonable to market.
To Exoctic Filipina – I totally agree – there are many scammers in the on-line dating business. Keep in mind the technology we are providing to dating sites is not just matching someones eyes but rather all of their facial and hair features – our platform is meant to compliment the existing text based search that all the dating sites are currently using. If while using our technology the user stumbles across a “fake” profile then it is the obligation of the site to filter out those erroneous images/profiles – of which perhaps we can discover for them.
At the end of the day this launch is meant to show-case – in a fun way – the technology that we are developing and an applicable business model. The faces/dating market is a small portion of our overall plan. Please stay tuned for much more regarding our image recognition technology and it’s applications in the next couple of months.
Greg Heuss
Eyealike – President
great idea, don’t think its marketability will be great though, not many people will use this, but again a great idea.
Never thought I’d see a pic of Steve Lemme on TechCrunch, but it makes me happy.