Riya 2.0 On the Way; Major Strategy Shift

Let the games begin!

I would like to see what I will look like or what that mythical gorl will look like 20 years from now. seems even more important

My wife had the idea of developing a website called WhoLooksLikeMe.com, it looks like Riya beat us to the punch

cool.I got to check

Howard, you can do that - sort of - here:
http://www.dcs.st-and.ac.uk/~m.....index.html
Just upload your picture and select “Older”.

If you’re already old, then you’re out of luck, check yourself out in the mirror, or see how you’d look if you were back in time by selecting “child” (unexpected things may happen

Note: I’m not affiliated with that site in any way.

Their technology blows my mind.
Riya makes redundant tagging obsolete - and if they can really build out these facial recognition algorithms to deal with the horribly unpredictable way we take pics, then they can find a way to deal with privacy. They may have to think outside their box but they’re capable.

You know this project out of Carnegy Mellon where that’s inadvertently tagged millions of pics by key words from making doing such addicting? All with human eyes dialated in glee? See http://www.espgame.com - amazingly addictive photo tagging – AND it is inadvertently not-creepy because the popular tags produced by the masses of people for the picture of your cousin will not have her name, rather the photos descriptions.

Excellent! I didn’t get Riya at all the first time around. Face recognition was interesting, but seemed pointless to 99% of users, and the government already has that sort of tech for themselves But this makes much more sense.

I am just afraid Riya are thinking bigger than they should.
My thoughts here:
http://blog.startup.gr/blog/_a.....35598.html

Riya seem to have some awesome technology, but I am not convinced they can implement it in any way useful enough for people to care about. How often will they be able to update their visual image search if it takes “months” to crawl the web? It seems cool, but almost a novelty. The dating idea is definitely just a novelty.

Despite their seven million photos in the first seven weeks, their traffic is still pretty weak:
http://www.alexaholic.com/riya.com+flickr.com

Also… I just tried their Mac uploader. It immediately crashes after you choose to upload photos from iPhoto. It looks like in the future there will be other ways to get photos in, but for now all those options are grayed out. Weak.

If they’ve truly nailed the face recognition problem there’s an obvious set of customers for them that has absolutely no price issue at all.

Homeland security, the CIA, the military, law enforcement and so on…

im sure the CIA have a much better (and private) face recognition software

1. Get incoming job applicant’s photo from security camera
2. Run through Riya
3. Discover photos from Spring Break, sophmore year

How many marriages will this end?

It might be similar like http://www.cortina.ch

But visual search is still a largely unsolved problem, calculated similarity is often not what people expect at a semantic level.

I am eager to see what they will offer, but I don’t assume they can jump ahead of the whole Computer Vision Research Community.

And: Scenarios like in post 12. won’t be possible for many years I guess. Scaling the exact facial search to the whole Internet is something very different than your rather small personal photo colelction.

im sure the CIA have a much better (and private) face recognition software

You’d figure the FBI would have computerized their case file system by now too.

But they haven’t.

http://www.washingtonpost.com/.....02329.html

I’m really impressed by Riya technology. Great technology, nice team (the CEO blog is a must to read). BUT, do Riya expects to build (and maintain) a billion dollar data center?

Because that’s exactly what they gonna need to browse internet pictures.

They actually use 64 dual opteron boxes with cost them $300K (http://munjal.typepad.com/recognizing_deven/2006/05/episode_1_march.html). With that hardware they can process search over more than 7 millions of pictures
(http://munjal.typepad.com/recognizing_deven/2006/06/episode_5_april.html). It gives us a (hardware) cost of 0.04$/picture.

Let’s say that we have 8 billions of webpages on Internet (Google indexed 8 billions of pages in September 27, 2005 -> http://www.nytimes.com/2005/09.....ml?ei=5090).

Let’s say again that we have an average of 10 “interesting” pictures on each of those website. It gives us 80 billions of pictures to index.
80 000 000 000 x 0.04 = 3 200 000 000$

Who can afford that kind of IT cost? And I’m not talking about the engineers army to maintain all that, the electricity bill and the bandwidth costs…

And what do we have on the other side? Some funny uses for geeks (“who looks like me?” “Am I on pictures that I don’t know?”) and some creepy uses for weird people (“where can I find a single girl who looks like my mother?” (lol)). The first category will use Riya for 5 minutes and then go back to their next AJAX project. Hopefully the second category is not so important (re-lol).

In a nutshell, IMHO I don’t see this making money. But I’m not at all a specialist, what do you guys think?

I think it’s a stupid move. There’s honestly no reason to attempt to spread a niche technology to the masses.

“Cool! I can find similar images.” ::sigh:: Now what?

With such great technology, they should be able to find attractive clients. Seriously, what a waste of VC funds.

this is impossibly complicated technology. the market will not tolerate a 50% success rate. it either has to work perfectly or users will give up trying the service. imagine if you found a “rug on ebay” then do a phto search and find a photo of a zebra on a nature site. How many times are you going to use the service? Perfect execution or nothing….that’s what they are up against. i just don’t believe that they’ve discovered the holy grail.

Yeah, searching for rugs - sure. No, (if Riya lets them), the biggest use of this will be to search for porn, as usual:
http://google.com/trends?q=porn%2C+rugs

If they get it right this service will explode with success, but getting it right won’t be easy.

I agree with Marcus and others that suggest that its a novelty with limited real-world utility. Looks like Riya came up with some cool tech first (and it certainly is neat), but is now flailing around to try and find some way to put it to use (other than getting acquired) - which, of course is the (generalizing) the wrong order to do things in.

I mean really - if the rug example is typical, how much use do they expect to get? Have they done focus groups? Built a quantified business case? Or are they really in throw-it-at-the-wall-and-hope mode?

Anyway, good luck to them.

Like many commenters, I believe that Riya has great, great technology. And it has the kewl factor on top of that.

What I don’t get with this strategy shift is what problem are they solving?

The problem that I have, the one that got me interested in Riya, and that I believe many others have, is the ability to find a specific type of picture from my personal file set in order to do something with it. As an example select the family picture that was taken in 2006 to use on a Christmas card. I want this ability on both my local and virtual drives.

There is a market for this and they could own it.

The problem being solved here is how to flip the company. Munjal thinks that by adding the search component, Google will finally bite.

Can’t wait to see this out. Google has been the only reliable service I have been using to get pics. Unfortunately google is only so intuitive. It will be great to see what the differences are with Riya!

It’s not going to happen folks, sorry to be the harbinger of bad news. You see,
computers aren’t performant to the level required for a product like this. And last but not least, the science isn’t there yet.

They better start thinking of another exit strategy real fast; maybe this is the exit strategy.

I think Lance got right idea. They need to improve their pattern recognition and try to beat google with picture search.

Oh, and their desktop apps.

They what they need to do not something silly.

*I am learning a lot from these companies and great for me since I am trying to do something in that market.

Referral fees for rug patterns? Sorry, there is no business model here. This sounds like desperation.

Wait, so they are adding object and pattern recognition? Facial recognition has a specific set of parameters, while generic object recognition is much, much more difficult.

I also can’t think of a single time in my life I’ve ever been shopping for something and wanted to find other things that looked similar to it. I suppose there are examples, like rugs, or flannel shirts, but otherwise I can’t see this being big business.

Their technology rocks, but they need to find a better business model for it.

IMHO, the facial recognition aspect of their technology was what made them interesting. This new strategy seems to be a recognition that maybe that part of the technology wasn’t quite there.

There’s another company, InQTel funded, that’s been working on the general problem of shape/color/pattern matching problem for a long time, (www.pixlogic.com) with limited success.

I’ve seen the Pixlogic stuff in action and it’s pretty impressive. However, as good as it is in doing pattern matching/recognition, its still is far from being able to anything useful for the average net surfer.

But, the elephant in the room is that nobody really cares about shape/color/pattern matching or searching. All anyone cares about is the meaning of the photo.

I want to search for happy, sad, fast, party, baseball, skyscraper photos. Basically, Flickr tags. Tell me why you need all the fancy recognition stuff for that?

Why won’t you call these guys on any of this? Rugs on eBay? Are you kidding me? Dating? Ha!

VC_flipper has it right. Built to flip.

Google has got to be working on the same type of image-searching internally (based on more than contextual content) … if Riya’s solution works, and can be bought for less than it’ll cost them to develop their own, then they’ll be bought.

Why are they so adverse to just going after the photo sharing market? Sure, Flickr and others have a big head start, but Riya has some cool tech that actually differentiates itself in a big way.

When Riya was originally announced, I saw it as having more potential as a Flickr-killer than anything out there. Then came the clarification that they weren’t even going after this market. Munjal has never clearly indicated on his blog (unless I missed it) why on earth they ever decided to not take this approach. It seems like such a perfect fit with the tech.

Besides, there’s a very good reason why Google hasn’t found a way to montetize its image search. Once they start making big money from it, then the ‘fair use’ defense goes out the window. But then, as is obvious from their current homepage, Riya seems to be all but begging for a copyright infrigment lawsuit.

It’s one thing to be hosting stolen images on your service that users upload. It’s quite another to be using them on your home page to promote your service.

Interesting how this has been announced a week after Tara leaves…and the home page changed only less than a week after she left.

I believe Munjal blogged that the reason behind the switch in strategy is that their search traffic (especially from unregistered users) is overwhelmingly higher than their browsing traffic and that search related advertising monetizes better than contextual placement ads. That sounds perfectly reasonable, but I am dubious as to how well ads on image searches will do compared to more general searches. A lot of regular search is directly related to purchasing, which is probably not true of image search. I guess that’s why they’re talking about referral fees instead of ad revenue.

I suspect this won’t be Riya’s last strategic change. It is just possible that the best application of their technology isn’t a consumer-facing web application at all. Just hope they will be able to find it before the VC cash runs out.

A more likely use of this kind of service would be by copyright owners wanting to find unauthorized use of their images on the Internet. But that’s a pretty small niche - artists and publishers willing to pay to track down and prove abuse of their copyright.

You dinks gotta stop using Alexa traffic data. It’s horrendously unrepresentative of reality, and referring to it makes you look clueless. Just letting you know.

Eric E: There are a whole bunch of companies that do that already - PicScount in particular has an interesting revenue sharing model for photographers looking to recoup lost revenue. The technology is, in my understanding significantly different than what Riya is doing though - I am not sure the two models are complimentary from a technical perspective.

why would anyone want to find “someone that looks like someone else” ? I mean really, if you’re searching for porn, do you want to find the same pair of titties? Do you really want to find all the bastards that look like you? Isn’t looking in the mirror every morning enough? And who on God’s big earth would want to find a girl that looks like the girlfriend who just dumped them?

“Hey, let’s find someone who looks like my ex…so I can start a relationship and pretend that we never broke up.”

The only uses are for the military and this fascist administration (search for all dark skin men with full beards). Nothing commercial can come from this technology.

“You dinks gotta stop using Alexa traffic data. It’s horrendously unrepresentative of reality, and referring to it makes you look clueless. Just letting you know.”

For your blog that gets 10 visitors a day, yeah it can be way off.

For well-trafficked sites, it’s pretty close to accurate.

“For your blog that gets 10 visitors a day, yeah it can be way off.”

That’s not my blog. But you should read it anyway. You might learn something useful.

“For well-trafficked sites, it’s pretty close to accurate.”

No, it’s really not. That’s what I’m telling you.

““For well-trafficked sites, it’s pretty close to accurate.”

No, it’s really not. That’s what I’m telling you. ”

Actually it is. When you get down to blogs with a few visitors a day, of course it can be way off.

When you get up to sites with tens of thousands to hundreds of thousands and even millions of users a day, it’s pretty close.

Is it perfect, of course not, and what many people don’t realize is, Alexa admits this upfront. They clearly state that any website/blog that they rank out of the 100,000 shouldn’t be considered ‘reliable’, and that the closer they rank a site to #1, the more accurate their rankings.

Which is pretty much what I said.

“That’s not my blog. But you should read it anyway. You might learn something useful.”

I did, and learned something interesting. You think Alexa sucks, yet are using their data to judge the effectiveness of Ask.com’s TV advertising?!?

http://www.mattcutts.com/blog/.....lexa-data/

Seems to be a bit of a disconnect here…

“I did, and learned something interesting. You think Alexa sucks, yet are using their data to judge the effectiveness of Ask.com’s TV advertising?!?”

@Mack, FYI, that poster is certainly not Matt Cutts.

It’s a pretty article!
Why, you say. Because I understood that information of this post may help me
(and not only me) in my web life. Thanks to author!

I saw these guys launch at http://www.demo.com Good team of people and I’m really interested to see them branch out so quickly. I would have thought they would have nailed the core first, but maybe this is more about expanding customer acquisition than a big part of the core.

Will you be able to find Riya pics when you do a google image search? If they are flagged public I think it would be ok.

niiice…now i can find out if my bitch of a girlfriend is on those dating sites again!

Here are my thoughts on the topic. I also posted them on my blog here:

http://www.mrmarkets.com/2006/....._to_d.html

Google Image Search, say hello to Riya 2.0, because you will soon have company.

Riya 2.0 is certainly an interesting development. I liken it to the development of voice recognition technology and taking some low hanging fruit. IBM, despite being a major driver of the technology research, has focused on the holy grail of 100% accurate voice recognition. The problem is there are many other places voice recognition works in much simpler situations such as using your call phone or navigating calling menus on places like Sprint customer service.

My point is that this is an important move for Riya because they could go from being pretty good at doing something to being the best at doing something. That is, perception may not be totally accurate, but Riya’s product for identifying people based on images across the internet has not worked as well as planned. Celebrities have popped up as looking like other people, and that does not exactly inspire confidence in the product.

Riya could also make waves in comparison shopping. Some online retailers, in an attempt to skirt price comparison shoppers, have renamed products or changed IDs, but still use the standard product image provided to all retailers. With Riya 2.0, which should be released in a few months, you could select an image and it will generate a list of similar images across the internet. For products, price comparisons would be generated.

In short, Riya could dominate consumer facing search.

Up until these new developments, image search has not been that great because you do not have many options for filtering. But with Riya, perhaps you could filter a search for “beach” by finding images with a similar pattern as another completely unrelated image. It seems there could be a bevy of uses only limited by the creativity of the users and the capability to do filtering.

Thanks man, i agree