Riya: A Big Leap In Visual Search Engines

by : Danny Wirken

Watch out for new software that will give a new face to search engines. Rather, a program that includes faces in the search function. A new California-based company, Ojos, developed the online photo-based search service named Riya. With the use of avant-garde technology that allows face recognition within photos, the system is made to recognize individual persons as well as text in the user's digitally captured photos and labels them automatically-no more tedious searching for people in unnamed photos.

Despite its only-recent use in search engine technology, the face recognition software is not entirely new. This is a development of a company called Evolution Robotics, used in software that operates in the revolutionary Sony AIBO robot. In addition, Adobe's Photoshop Elements 4.0, a photo-modification program, has a somewhat similar function, although it does not have the ability to label individual photos with exact names.

Of course, Riya will not be able to recognize certain faces immediately after you have installed the software. You need to put in several minutes of training the program to identify a certain person whose pictures you need to find, although once it's done, you can easily carry on with the search. Not a bad service at all for tech savvy individuals, who are fast becoming shutterbugs.

Moreover, training the service can be pretty easy. While organizing your photos, just be sure to input the name and email address that corresponds to each new face you happen upon. The second time you see that same face, a drop down list of options will include the associated name, so you can simply choose from there. In due time, the service will become accustomed to identifying your family, relatives, friends, business contacts, and others that pop up often enough in your photo collection. Tags will then be made automatically, and the best way to ensure it does this is to train it often.

For the most part, Riya is a Web-based system that requires the user to have a small-sized file that helps your computer upload pictures to servers of the company. You just need to indicate on the PC module the part of your computer to look for your photo collection in, and you can continue to do work as it serves up the files in the background. However, uploading hundreds of photos have taken more than a few hours during post-launch software tests. This is a shortcoming that Ojos must work on to continue developing Riya into a top-notch application. Meanwhile, the company has recommended that heavy users schedule the uploading during their sleeping hours.

The service will then scan for people's face in your photos once you have uploaded them. At first, no tagging will be done, as it will only collect the thumbnails that correspond to every unique face found in your set of pictures. Of course, there's no guarantee that two different pictures of the same person will be recognized as one distinguished face; unless it has been trained to recognize specific faces, photos of one person in, say, five different angles may very well be identifies as five different people. Although it may be able to tell correctly that two different photos contain the same person, its capacity to do so will only increase as it is told whom the different faces belong.

The reason for using an email address to help identify a person on a photo is to provide each face a matchless classifier. This way, the amount of training you've spent time on can be instantly shared with other users of Riya. For instance, if you've trained the system to recognize the individual members of your family, then other relatives can also gain from your work in training the system. It is up to you if you would like to share your Riya training, which would then allow the other user, given your permission, to search a member of your family once, using the exact same email address you used to tag the family member, and it will at once single his or her pictures out from the other user's personal photo collection.

Something about the workings of this system might not seem appealing to some people; they may even find it frightening, to think that tons of photos with their faces in it and labeled with their email addresses may be accessible to others, maybe even strangers, at any time. It would be comforting to know that only the company can completely access this record of faces. Those who may post pictures with your face in it can allow others to view them, as well as make the system training open to them, but rest assured that the email address linked to the person is by no means revealed through the software.

In effect, another user will fail to identify you with the use of the Riya database if he or she does not already know your email address and possess photos with your face in them. Ideally, those who do know are people you know and can entrust with personal data, yet some users are still reluctant to entertain the idea of having a software company gathering email addresses without their knowledge or assent, despite its non-disclosure policy.

Initially, Riya was only able to recognize faces in photos that are deliberately posed, with the person staring right into the camera. Even a post-launch internal testing was able to reveal a number of serious bugs to Ojos' team, which they were able to fix in a matter of days, thankfully, but only future upgrades will unveil its full potential. Until then, imagine the convenience of having Riya in your computer system. Image files containing text will be easily searchable, visually, while finding people in your digital pictures collection will be a breeze. In the future with Riya, locating the files you need while you work on a website design or composite photo will be much faster and possibly hassle-free.