Opinion: Riya recognizes objects in photos—faces, text, whatever—and auto-tags those photos automatically.When it comes down to it, one of the biggest problems with developing a big site really isn't the technology, the information architecture or even the design. Nope, in terms of time-sucking, nail-biting, blood vessel-popping problems, content is probably the biggest.
Getting it is the first big problem, and as far as I know there's not a technology in the world that will yank content out of reticent or disorganized clients. But once you've (finally!) gotten it, the other huge problem is organizing it all so that you can shovel it into the site.
Organizing the text you've received is probably the least of your problems. Text is indexable. Text is searchable. Text is easily dealt with because no matter how big a mess you've got on your hands in terms of folders, mislabeled files and odd documents, chances are any one of the new generation of search tools such as Google Desktop or Apple's Searchlight will be able to peer into those files and tell you what content is inside. More advanced software (such as Agent Software's Precision Recall) will help you find, organize and categorize even the most complex documents.
But photo assets … well, that's a different story. Photos are a heck of a lot more complicated to search with a computer because images are far more ambiguous. Most of us have hard drives overflowing with cryptically labeled image files—a fact that's bad enough if they're your image files but even worse if it's a giant CD or DVD archive from a client who hasn't taken the time to label the photo assets they're shoveling your way.
Asset management has always been a big problem, and as more assets get created the problems are only going to get bigger. (see Microsoft's MyLifeBits Project for a glimpse into just how bad it could get in the future.) And while there will of course be plenty of text and documents to wrangle, multimedia assets such as video and images will continue to stretch the capabilities (and patience) of developers who have to deal with them.
Up until now, digital asset management systems such as Canto's Cumulus have been standard tools for those who have to manage lots of digital media. But while tools like Cumulus are good at storing and organizing stuff, they're still reliant on metadata being entered by those storing the assets. Sites like Flickr achieved metadata breakthrough by allowing tagging to become a social activity, but images on Flickr are dependent on a human being actually telling the system who's in the picture.
Enter Riya.
In a nutshell, Riya's a system that's smart enough to recognize stuff in pictures after being trained—faces, text, objects—and then auto-tag all the photos in a folder containing all that stuff. Automatically. Just upload your pictures, set your privacy settings (how and if you want photos to be shared), "teach" Riya by clicking on thumbnails of the faces it finds in your photos, and sit back as it "looks" at each photo and tags the faces it finds.
But it doesn't have to be faces. If you select a recognizable object in a photo—say a picture of Big Ben in London—it'll find all the occurrences of that object in your other photos and tag them accordingly. A search function will allow you to do compound searches, theoretically permitting you to search for "Mom" and "Boat" and finding all the pictures that you took of your Mom on her boat (or whatever recreational conveyance your Mom cruises around on).
Riya's positioning itself as a consumer service—think Flickr that can automatically recognize and tag photos—but from a developer's standpoint it'd be pretty cool to use, too. If you get a disc from your client with a bunch of product shots, you could conceivably have Riya tag them all for you once you've taught it what the products looked like. Pretty cool, huh?
By the time you read this, Riya should be launched in a limited closed alpha release, but according to Michael Arrington's Techcrunch blog, Riya's the real deal and not some vaporware dream. There's no word yet when it'll be released, but it definitely sounds worth waiting for.
Now if someone would just develop an app that'd allow me to send electrical shocks to clients who haven't delivered the content they promised …
