Riya, the Silicon Valley photo recognition start-up, is expanding its ambition: It wants to become a full-fledged “visual” search engine.

The move was signaled today by chief executive Munjal Shah, who told us in a phone call that he has radically transformed the company in recent days to go after this new goal. It will take about three months to crawl the entire Web, collect images and then process them to identify their various qualities so that they are searchable, he said.

The change is not a result of any sense of initial failure, he said. Riya’s original plan was to recognize individual people in photos: After seeing several pictures of your girlfriend, for example, it can begin to recognize her in other photos, and thus auto-tag them — helping you organize. Riya can also recognize other similar patterns, such as writing in signs.

Munjal Shah

Within seven weeks of launch, Riya’s users had uploaded seven million photos — that’s compared to the two years it took for people to upload 10 million photos to Flickr, Shah said. Riya isn’t perfect, he said, comparing Riya with the smarts of a two-year old. But it worked “good enough,” he said.

Soon, though, he and his team noticed something unexpected: For every person who used Riya to search for their own photograph, twenty others were using it to search images on the public web. “They wanted a new public search engine that was smarter than the Google image search engine,” he said. “Our usage numbers kept going up,” he said.

He realized he should respond to this feedback and shift Riya’s direction, even if it wasn’t in the company’s original plans.

So how will it work? Say you find a rug on eBay with a two-inch border with a favorite cross-stitch pattern. Riya will analyze the rug image, process its pixel structure for a histogram of color and shapes, and so on, and then match it against the database of images it has collected from the entire Web. Riya will show you the other places it has found the image, for example whee Amazon.com is selling it for ten dollars cheaper. Riya gets a referral fee when you buy the Amazon.com rug. Same thing can be done for dating. Search for an image of a hot blonde, and Riya will pull up similar looking blondes at Match.com. You’ll be able to photos for your search query, so you can look for just about anything.

Building its database isn’t trivial — images which will be collected and processed over the next three months, Munjal said. “It takes a lot of horse power. We’re building a massive data center.”

Venture firm Bay Partners pumped $15 million into Riya, a round announced in January.

He will go after various markets, such as products (the rug example), dating (the blonde example), travel, and landscape. He believes there may be a $100- to $200 million advertising and referral market in any one of these verticals.

The company employs 50 people, 44 of them engineers, and 14 of those Stanford PhDs. Every time we talk with Munjal, he brings up these engineering stats. It’s probably because he doesn’t want to end up like the dozens of online video sharing sites that YouTube is taking the cleaners. Deep technology is important, he said. “I think Riya’s got a better shot than most at really creating a billion dollar market,” he said.

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.