What "problem" does Riya solve?

The hype and rumors around Riya are perplexing to me. In fact, count me as skeptical. Some of the uber-bloggers, Robert Scoble, Om Malik, and Michael Arrington have been praising Riya, yet saying little about what Riya actually does. These guys are very influential in the tech community and can create a lot of buzz...and they have. So, there must be something there... Riya is in a controlled beta right now, but their web site indicates that they do picture categorization, tagging, search, and face detection. I am not in the beta so I have no idea what they really do, but the stated features are certainly not new.

Image search has been around for 8 years. Nick Whyte and our team at AltaVista were the inventors of multi-media search and it is still in use today at Yahoo. Nick now works for Microsoft in the search group. All the major players have reasonably good image search. It is much more difficult and sophisticated than text search, but takes advantage of many of the same techniques. We had some crude face detection technology and some pretty sophisticated skin detection technology to screen out pornography.

Face detection has been around for more than 30 years. The Defense Department, CIA, NSA, and all the defense/intelligence guys have been buying this stuff for years. There are also lots of PhD's from MIT, CMU, Stanford and other places that have done work on face detection and image identification. There is lots of technology already in the marketplace. Tagging and categorization has been around for a long time too.

Bradley Horowitz was one of the founders of Virage, a leader in search within video, is now at Yahoo doing multi-media search. Virage had very sophisticated speech recognition and video pattern recognition technology for searching for specific clips within a long video.

The real question is not how sophisticated Riya's technology is, but what problem do they solve? There are lots of photo sites that allow you to store, categorize, tag, and search photos.  So, for anyone who actually uses Riya, I have a few questions. Actually, these are the same questions I ask any new start-up;

1. What problem does it solve?
2. How are people dealing with this problem today?
3. Who will pay money for the solution to this problem?
4. How much will they pay? What is your revenue model?
5. What do you have that the search or photo service companies don't have?
6. What unique technology do you  have that can't be easily implemented by the big guys?
7. Is this a "vitamin" (nice to have) or a "pain killer" (got to have it)?
8. Why would one of the big players be interested in acquiring the company?

Maybe there is good reason for all the hype around Riya. But, until I see the service and understand the problem it solves I remain skeptical.

In my job I see hundreds of new software / service companies every year. Most of them are all about technical innovation. I love technology and always want to know how it works, but technical innovation only matters when it is applied to a real problem that satisfies a pain point. Asking the eight questions above can help focus the discussion on the real pain point and value proposition.

Can anyone answer these questions with regard to Riya?

UPDATE:  Several of you have responded in the Comments section with descriptions of what Riya does. It appears that face detection and automatic tagging of the faces it recognizes is the main feature. Riya strikes me as a "nice to have" feature of a larger service, but not something, on its own, that you could build a profitable company around.

I know of lots of other face detection technologies on the market and completed projects in research labs that do the same thing. The task is made much simpler since it only has to recognize a very small set of faces...the dozen or so faces that are in 90% of your pictures.

I know Microsoft has similar technology sitting on the shelf in the research labs. I am pretty sure Yahoo does too. I don't see a big barrier to entry for any of the big guys to enter this business if and when they want to.

Still not much information on the business or revenue model. I don't see how ads can be effectively targetted to pictures with keywords like "Robin, Caelan, and Devan".  I have learned painful lessons about cool technologies, even those that attract millions of users,  not translating into a sustainable business. AltaVista and Napster are two examples from my personal experience.

Riya probably does have some new twists on the technology, and may eventually figure out how to attract lots of users, and monetize the service.  But at this point, the hype is way ahead of the reality.

Comments

[Vaibhav Domkundwar - Lavarithms]

Don,

You are so correct here. Being an entreprenuer myself, I find it very strange when so much hype is built around Riya which doesn't seem to solve anything new and parts of most of what they do has been done before, as you have pointed out. If they add tagging and relate recognized pictures, it is nothing more than database application features. My buddy at Cal did his Ph.D. in a similar field and he addressed a much bigger features recognition problem with mission critical application.

I was totally surprised with the rumors of acquisition because the technology didn't seem a new innovation they claim, the business model is just not clear ( I asked this question on the Riya CEO blog sometime back with no response) and if a company that is so passionately claiming to be focused on solving a big problem is ready to sell so early it makes you wonder....and question the hype yet again.

We all haven't seen Riya live as yet, so let us give them the benefit of doubt but all what you have said still stands true.

[Robert Scoble]

I'll work on that tonight. It's a photo service where you upload your pictures. It recognizes faces and puts an entry box around them. You tell it a face, then it goes and looks for other versions of that same face. It works pretty well right now (I'd guess it recognized about 80% of the times my wife was in a picture, for instance).

Why does this matter? It makes it a lot easier to identify who is in a picture for your friends.

[Patrick Dodds]

Doesn't the thought of a programme crawling the web and finding pictures of you / your wife / your children / your school friend make you smile (albeit anxiously)? A hundred ways to use such a programme come easily to mind. Being non-technical I imagine I am being hopelessly naive but the idea is just so exciting, especially as storage capacities increase and camera resolutions available to even the most amateur of photographers become enormous.

[Philipp Schumann]

You don't have to solve a problem. Or put differently, sometimes you only know afterwards what problem you solved. Of course they will be targeting consumers first, businesses second, but there is a single skill that they market (face-based image search) that will likely open up business opportunities in both areas.

Just imagine the possibilities: "find me all the pictures on the web with Matt and Max, but not Sarah". Or "the ones with [your_favorite_celebrity_here]".

Websites today can SEO or SPAM their way to the top result with text-based image search, something I presume will not be so easy for face-based image search.

All of which reminds me, Sig had a nice post recently about why you don't always need to a have a problem to solve:

http://thingamy.typepad.com/sigs_blog/2005/12/ambitions_and_l.html

[Chris Whalen]

I also spend most of my time looking at new technologies and have seen a few similar to Riya in the homeland defense arena. From what I can tell, Riya, and a few I have seen like it are different from other photo search products because of the ability to recognize unique shapes within pictures. Basically a very complicated algorythm that breaks a photo into hundreds of pieces and then looks for the pieces.

One demonstration of a seemingly similar technology I saw involved looking for a particular window, from the Ma and Pa Kettle painting, on the web and only finding that very specific window type regardless of angle, how much of the window was showing, etc. Pretty effective. Very specific. Riya seems to be a type of this shape recognition technology.

For the problem it solves, very useful for applications where you need to find a very specific item. I have seen it used in evidence databases, looking for specific facial features to ID someone wearing a mask etc. For personal photos useful if you need to find an attribute that is very specific and unique and not searchable by other means. I have seen Betas of using this as secure sign in as well having your PC recognize you when you are in front of the PCs camera and signing you in. Also sorting photos by a theme such as "all photos with John in it" or "all photos at the Pyramids".

I really like the technology, but I am not sure you can build a real business on it. It seems to me to be a great complimentary technology for another product line. Perhaps the folks at Riya realize this and this is why we hear so many acquisition rumors.

[Myron Kassaraba]

I've used the Riya Alpha quite extensively (on close to 4000 of my photos) for an article to be published in Future Image Report (www.futureimage.com).

It is true that Ojos/Riya has attracted a lot of attention. That is mostly a function of their using an "open blog" approach to what they are doing and the network amplification effect that the blogoshpere can unleash. IMHO they have not been irresponsible in their marketing. Their software/service pretty much does what they claim - is finds people in your pictures, lets you train their facial recognizer with a fun UI and then does its best to automatically recognize more of those people in your pictures. It also detects and recognizes text which is not fully functional in the Alpha but will be equally valuable. There are many startups who who would love to have Riya's waiting list of consumers dying to try their service.

Though there has been a lot of work done on face detection and recognition technologies over many years, most of it has been in research labs and universities or for commercial applications. Riya is unique in a several ways, they are doing some things differently on the technology front and they have actually put a system out for consumers to use on their snapshots. They have aimed their guns at consumer snapshots - this is some of the hardest content to automatically process because of the variety of poses, lighting, exposure, backgrounds, etc. Their goal is not to search the entire web for a picture of Uncle Joe, it is to let you develop a training set of the 50-300 people that show up in 95% of your pictures you take so that you never have to manually tag a photo in the future.

The problem Riya solves is that consumers do not tag or categorize their photos. The capabilities have been in Picasa, Photoshop Album and many, many other software products. At Kodak we had a software product called Shoebox that shipped in 1992 that had great organizational capabilities that nobody used.

A photo without metadata is nothing but a bucket of bits. Digital photos get lost and forgotten, once forgotten, it is really hard to go back and add that metadata. Since Kodak, HP, Shutterfly make money from prints, cards and books - making it easier to go back into your archives to find all the pictures you have of some close friends or loved ones so you can print them a photobook for their anniversary means real $$$ for their businesses.
Also, photos (and video) without metadata are almost useless as advertising content since you can't target the ads. Is the metadata generated by Riya good for Google (or Yahoo or Webshots), you bet. It turns buckets of bits into gold.

Riya's contribution will be to provide a service for consumers to use (or for other services to plug into as a utility) that offers good value even in its current form and because the recognition is done on proxy images that are on Riya's servers, they can easily iterate and extend the intelligence of the system as better algorithms and techniques are developed. Stand-alone business, maybe, maybe not - but once it is up and running with several hundred thousand users (and hopefully a robust infrastructure to handle the traffic & processing) it will have value.

Will Microsoft and others throw a bunch of smart people at this - absolutely. Ojos and their investors are betting they can run faster and get a MySpace-like exit vs. selling out now and getting lost in Googleville or Yahooland.

[Dan Hawkson]

Your argument is similar to this:

When 3Dfx (the company that produced the first video cards with 3D acceleration for gamers) released the Voodoo 3D accelerated video card, what was new?

Video cards had been available to general customers for tens of years. 3D acceleration was available in military applications and very high end engineering workstations.

What 3Dfx did was take 3D acceleration from military applications and make it available to everyday people.

Let's discuss Riya. Yes, face recognition technology is available. But it's not available TO ME or TO YOU, to have it sort your photos!

This IS a big deal. I would certainly buy a program which sorts my photos according to the people in them, and enable me to search by name and maybe by scenary (like "find all pictures of Sandra and me on the beach, shot before 2002").

So Riya IS a big deal.

However, I see this more like a lean and mean utility which costs $40 - $50. I don't see it as a VC-money-guzzling-web-2.0-site. It may also be very overhyped.

However, at the center of it may be a very useful real product.

[Chris Whalen]

I agree with both postings above. Having looked at several of these technologies currently being used by or presented to various government entities, it can be a very useful piece of software. I know I could use it for my photo selection. I am just not sold on it as the centerpiece of a viable business. Rather a nice add on or added service to a Photoshop, Shutterfly etc. adding value there. The play may be for Riya to stay a technology company and license to the above players. THat maight be where the business lies if it can be done as a stand alone business. Thoughts?

[Bala]

From an application standpoint, Riya's alpha release is ways from being complete. It failed some basic image recognition test cases. The good folks at Riya are working hard towards having a better release out sometime in Feb. More details of my test with sample images here: http://balak.blogspot.com/2005/12/riya-alpha-release-so-alpha-it-hurts.html

[Myron Kassaraba]

Hi Don,

I don't disagree with anyone's comments about the question of "is this a feature or is it a company"? That remains to be seen. The revenue model could be premium subscription services, it could be lookup fees from other photo hosting services it could be technology licensing.

I will add a few more points that I think are relevant. In the world of "recognition" of anything, the currency is "training sets". Those faces or objects that were identified by someone to be used to match against. You could see how Riya could get a big lead on building a database of these training sets and then allowing the sharing of training sets between users or across services.

Once I have built a training set of the people in my pictures (mine are typical family photos & business events and I have 110 unique people recognized 2171 times in 3800 photos with about 1000 faces detected and not yet recognized). Once I've invested the time to do the training once, I'm not going to want to do that again on Flickr. Riya lets me keep my original pictures on my hard drive or on an online service. They keep an index, the training sets and will give users the ability to export their tags to their host services.

Second point is that though "Robin or Devan" are not valuable as an ad trigger for Google the fact is that as a user, once you start tagging and getting the benefits of being able to search your photos, it is hard to stop. Metadata is addictive. As long as Riya can make it fun and as easy and as automatic as possible, they will be encouraging and facilitating the capture of metadata with those photos that goes beyond the names of the people in the photos to include things like location, venue, events, etc. Eventually many of those too can be automated.

When I look at some of the me-too companies attracting venture money these days, tagging automation for user generated image content (and video I'm sure in the future) seems like a pretty reasonable place to place a bet!

AltaVista and Napster were great efforts that were ahead of their time. Nobody ever said being a pioneer was easy! (I know how that feels, I was selling ads on online services in the mid-80's and selling CCD color scanners for $13K in 1988).

The organization and retrieval of personal images problem is very real and here today so we'll see come Demo in February (when Riya says they will open the doors to the public) how they do.

Happy Holidays!

Myron

[Paul Montgomery]

I worry about photo sites like Riya getting sued by celebrities defending their right to publicity.

http://tinfinger.blogspot.com/2005/12/flickr-of-tiger-woods.html

Someone tell me I'm being too paranoid.

[Myron Kassaraba]

Here's a link to a post on my Camera2web.com blog about some interesting research on Mobile Media Metadata going on at Yahoo's lab at Berkeley.

http://c2web.blogspot.com/2005/12/new-scientist-technology-camera-phone.html

[Myron Kassaraba]

Update:

Riya raises another $12 million in venture investment. Plans public launch at DEMO.

http://munjal.typepad.com/recognizing_deven/2006/01/going_to_the_mo.html