Face detection is a common sight in still photography, but it's a rarity in video outside of certain research projects. Google may be keen to take some of the mystery out of those clips through a just-published patent application: its technique uses video frames to generate clusters of face representations that are attached to a given person. By knowing what a subject looks like from various angles, Google could then attach a name to a face whenever it shows up in a clip, even at different angles and in strange lighting conditions. The most obvious purpose would be to give YouTube viewers a Flickr-like option to tag people in videos, but it could also be used to spot people in augmented reality apps and get their details -- imagine never being at a loss for information about a new friend as long as you're wearing Project Glass. As a patent, it's not a definitive roadmap for where Google is going with any of its properties, but it could be a clue as to the search giant's thinking. Don't be surprised if YouTube can eventually prove that a Google+ friend really did streak across the stage at a concert.
Google patent filing would identify faces in videos, spot the You in YouTube originally appeared on Engadget on Tue, 03 Jul 2012 15:11:00 EST. Please see our terms for use of feeds.
Permalink | Email this | Comments