Pixazza is looking to turn any photo on the Web into a miniature online store with a new service it calls "Product in the picture." It likens the service to Google AdSense, except for images instead of text, and has landed US$5.75 million from Google and two investment funds to roll it out.
The service will tag photos showing products with links to online stores selling those products, charging the stores a fee either per impression, click or sale, and handing a slice of it to the sites displaying the photos.
Visitors to participating sites will be able to run their mouse over a photo to see tags identifying the products and linking to a site where they can buy the items.
Pixazza will initially target the fashion industry with the service, launched Wednesday, but also has its eye on home furnishings, sporting goods and consumer electronics, it said.
Publishers that sign up for the service must add a little JavaScript to their Web pages so that Pixazza gets a tip-off whenever they post a new photo. The publisher's share of the revenue will depend on the amount of traffic they generate, the images they publish and the readers they attract, Pixazza said.
With headquarters a few miles from the Googleplex in Mountain View, California, and a reliance on targeted advertising to make money, Pixazza has much in common with Google. Its name is even reminiscent of Picasa, Google's online photo album.
However, one area where Pixazza differs from Google is in its use of paid human experts to identify the items shown in pictures.
Google typically prefers to make its services as scalable as possible by relying on unpaid users to signal problems or by fully automating services.
Two cases in point: Rather than employ an army of checkers for the millions of pages it is scanning for its Book Search program, Google simply provides a link for users to flag a scanned page as unreadable, while the Google Voice telecommunications offering the company unveiled last week includes what Google said is the only fully automated voice-mail transcription service on the market.
A voicemail transcription service from U.K.-based Spinvox, on the other hand, learns to recognize indistinct or new words by passing problem phrases to human experts.
Pixazza's reliance on "crowdsourcing" - the efforts of a large group of people - and the fact that its experts don't work for free means it must optimize its service to make the best use of their time.
Thus, Pixazza queues photos so that those generating the most traffic are examined first - usually within an hour, it said.
The company sees its crowdsourcing platform as its technical advantage - but it may be the product of that platform that interests Google.
While there's not yet a reliable way for computers to identify the make and model of a pair of shoes from a photo, that may become possible in future, given a sufficiently large database of accurately tagged images.