Apparently the marketing plan for Riya is bloggers and online "word-of-mouth". This seems to be paying off very well: Many are blogging about or discussing this "amazing" product, and how it's going to revolutionize the photo tagging world. Flickr (Yahoo!) and Google are going to be knocking down their door trying to get a piece of that action!

Remarkable, though, how incredibly few people have actually used the product, and how few will actually vouch for its capabilities. If you want to sign-up currently, and try the product out yourself, it's an "invite only" affair (though strangely you have to send them your email address - That's not invite only. That's a lottery system). Despite the so-called press (see the Wired article above) heaping on strangely uncritical praise, no credible reviewer has had a round with the software. Odd, wouldn't you say? Wouldn't it make sense to get a respected reviewer to vouch for its capabilities before firing up the press wagons? Someone credible who would put it through its paces and either credit or discredit it, putting their actual career on the line if they misrepresented it.

Facial and scene recognition is easy in theory - it's something we've all imagined up, inventing our own naive ways to do it - but in reality it is extremely difficult. Yet these guys not only managed to leap the gigantic hurdle of facial recognition (including discerning among incredibly similar people - close relatives, and supposedly even twins), but they added in fantastic, unparalleled text detection as well (in one case purportedly reading a tiny car logo sloped about 70 degrees away from the camera, among other fairly impressive feats).

When it comes to revolutionary technologies like face/scene recognition, it is critically important to withhold judgement until it actually proves itself in the real world (and no - I'm not being hypocritical. I'm not saying it doesn't work - I honestly don't know - but I'm just say that without proof otherwise claims of revolution seem a little premature). Facial recognition in particular is a field filled with hucksters and fraudsters, grossly overselling the capabilities of their system with dummied up sample cases and ridiculously ideal scenarios (or even worse - "mechanical turks" have been known to occur). I consider facial recognition much like the compression market - how many times have we heard about revolutionary new compression technologies, sold through jimmied up demos and "observers" on the dole, that in the end turned out to be nothing but a fraud (or a completely impractical edge scenario that is of no value in the real world).

I have no idea if this particular product is legitimate or not, but the lack of credible analysis thus far makes the growing chorus of revolution a bit difficult to stomach.