Tech Watch | InfoWorld Staff » Google adds facial search

June 01, 2007 | Comments: (0)

Google adds facial search

TAGS: Search

Google's popular image search has quietly added a feature that picks out facial images from the sea of graphics on the Web, allowing Web searchers to better zero on in pictures of individuals. As noted by the great folks over at Ars Technica and first reported by Google Blogoscoped, the popular Google-centric blog. The new features are unofficial and allow anyone to narrow a search to facial images or news related images by appending the terms "&imgtype=face" or "&imgtype=news" to any Google image search.

For example, a generic image search for "InfoWorld" yields a hodge podge of different Web images, heavy on the InfoWorld cover art. The same search with the facial search suffix added on turns up a rogues gallery of people who have passed through InfoWorld, or been depicted in our magazine, from founding editor Allan Lundell, to current editor in chief Steve Fox to current and former contributors like Jon Udell.

Adding the imgtype=news modifier narrows the search results to images that have been attached to news stories.

One thing that's particularly powerful about the feature is the way that it creates an ersatz network of individuals who have associations with the search term. For example, the InfoWorld search ended up cobbling together a list of InfoWorld alumni and countless other contributors through the years from all across the Web. You can see the same effect searching for Microsoft faces.

In their review, Ars Technica notes that the facial search features likely came from the company's 2006 purchase of Neven Vision and could one day be used to locate all kinds of different items on the Web -- akin to the kinds of features that startups like Riya are working on for the retail/e-commerce sector.

Cool stuff. Eerie -- but cool. One does wonder whether the new feature might make it a tad easier to ferret out images of individuals by scouring Web sites of organizations they're affiliated with. In that sense the feature could be misused, though it seems the onus falls on Web site owners to beware of what kinds of data is getting crawled on their site.

Posted by Paul Roberts on June 1, 2007 09:39 AM

RATE THIS ARTICLE: