A New Approach to Searching Photos

Posted by: Rob Hof on August 25, 2005

Munjal Shah, onetime cofounder of the auction services firm Andale, finally let slip on his new blog what he’s been working on since leaving last year. As he writes: “I am co-founding a company because I found I had 31,246 photos all named DSC0009.jpg.” In other words, his startup, tentatively named Ojos (Spanish for “eyes”), is creating a new way to search and organize photos.

Now, that might not sound at first like the most original idea. Among many others, Flickr, now part of Yahoo!, and Picasa, now part of Google, are tackling various aspects of organizing the crush of images spawned by digital cameras and sharing them with others. But while Shah remains coy about all his ambitions, he revealed the key technologies behind Ojos: face and text recognition.

Ojos plans to offer free unlimited hosting of photos, with software to upload photos to the service. An early demo I saw appeared to do a surprisingly good job of matching different photos of the same people. Though Shah admits there are many limitations, his bunch of Stanford facial recognition researchers seem likely to make further improvements.

The text recognition looks pretty interesting, too. Check out this photo of a Treo, annotated with all the words the Ojos service recognizes. Combine that with the ability to group photos taken in close time proximity to each other, add in a little social networking, and you can imagine that a large proportion of your photo collection might finally become useful.

The other key: You can assign tags, or keywords, to one photo and the service will automatically append that tag to other photos of the same people. That's potentially much less hassle than even Flickr, which already made tagging a lot easier.

Shah's cofounders include Burak Gokturk, a Stanford Ph.D. who holds 15 patents in face recognition, and entrepreneur Azhar Khan. Investors include Peter Rip of Leapfrog Ventures and John Malloy of BlueRun Ventures, the former Nokia venture unit.

No launch date yet. In fact, Ojos is an example of the new stealth mode for startups. Which is to say, not very stealthy. The new wrinkle here is that there isn't even a beta site up yet. Shah hopes that by soon letting people send in photos that Ojos will process and send back, and also inviting some alpha testers to try out the service and provide feedback, he'll tap into the open-source style. "Getting people involved with our development process will only make our products better," Shah says. With the potent competition out there, they'll need to be.

TrackBack URL for this entry: http://blogs.businessweek.com/mt/mt-tb.cgi/

Reader Comments

Munjal Shah

August 26, 2005 12:34 AM

Rob,

Thanks. You've always been fair in your report. I think Flickr's tag based system is just super, but I wanted all of my photos on there, I wanted them all tagged, and I didn't want to spend hundreds of hours doing it. So being the lazy engineers that we are, we thought maybe we can at least auto-tag some of the faces and names. Folks can fix mistakes we make but it will still be less than than tagging in the first place and in the end you will have a tagged library of photos. I hope lots of folks come to give us feedback as we want your help to make this a great product.

Thanks - Munjal

Jonathan

August 26, 2005 01:50 AM

Question for Munjal: Does your autotag program have that good of pattern recognition that it can differentiate faces?

Munjal Shah

August 31, 2005 11:51 PM

Jonathan,

Yes it can differentiate faces but not perfectly and it does better if you give it more examples to "learn" from.

Andy

September 1, 2005 05:30 PM

How would one go about being an Alpha tester for this product? It sounds very interesting. I love flickr too, but who wants to tag 20gb worth of photos...? No one. I do not see flickr adding value with tagging at all anyway. They add value by creating a medium to share photos rather than email. Tagging is just a perk. If Ojos can really automate this process, they will be huge. Look out flickr (yahoo) ...

Next question... who's going to buy Ojos?

Vikram Aggarwal

October 31, 2005 06:17 PM

Check out myheritage.com, they also tag photos based on face recognition, but they are scalable, which means they'll detect the faces you teach them in photos submitted by other users and not just by you. They take this service in the direction of genealogy by allowing users to find photos where they appear, or their family, or their ancestors, contributed by other users. Since face recognition works by creating parametric maps of faces, using a variety of different algorithms and approaches, it is surprisingly good in finding relatives, i.e. the people whose faces resemble your face the most are, as shown mathematically, your parents, siblings and other close relatives. This makes face recognition exceedingly good at discovering people you're related to, that you may or may not have known about. Check out myheritage.com if you're interested in face recognition applications for consumer photos, or in the future of genealogy research.

Post a comment

 

About

BusinessWeek writers Peter Burrows, Cliff Edwards, Olga Kharif, Aaron Ricadela, Douglas MacMillan, and Spencer Ante dig behind the headlines to analyze what’s really happening throughout the world of technology. One of the first mainstream media tech blogs, Tech Beat covers everything from tech bellwethers like Apple, Google, and Intel and emerging new leaders such as Facebook to new technologies, trends, and controversies.

Categories

 

BW Mall - Sponsored Links

Buy a link now!