Tumblr receives about 130 million posts per day. The Yahoo-owned brand is one of the more popular social photo sharing sites on the Web and is now beginning to parse posts made by users to identify brand affiliations. After updating its privacy policies, Tumblr made a deal with Ditto Labs of MIT fame, which specializes in analyzing photos for brand-related data.
The idea is that if you post photos that include branded products in them (think t-shirts with Nike logos, Coke bottles in the background, a Nissan parked in the driveway), the identifiers will be flagged and used by Tumblr to target advertising. For now, Tumblr says, this is an experiment and they are mostly gathering data rather than using it to serve ads or target anything. They want to know how well this will work and how much data is truly available for useful targeting. Mostly, it will be made available to advertisers (for a fee, of course) who want to test their branding efforts or gather fresh metrics on their brand’s proliferation.
This trend is a new one for photo publishers, but Pinterest purchased VisualGraph early this year with the same basic goal likely in mind. This purchase lead directly to Pinterest’s introduction of follow suggestions on the sharing site. Facebook is also getting into the game, having invested into Vicarious, an artificial intelligence firm that includes image recognition in its repertoire. Google is also reluctant to be left behind, also purchasing an AI company (DeepMind).
It’s about to get interesting in the photo sharing world.