Google patents object recognition software for YouTube identification
Google has just successfully patented a piece of object recognition software that could change the way in which we publish videos.
The technology works in a similar way to face recognition, except that it distinguishes between objects - specifically, those that aren’t of the human kind. The software would work automatically, labelling objects seen in a video uploaded to YouTube by identifying characteristics in each object.
Google will match these to objects stored within its database of “feature vectors”, using movement, colour, shape and texture of said objects to help determine what they are.
YouTube users will be able to help the matches be more accurate by naming and tagging their clips with keywords, which in turn will narrow the search parameters Google needs to implement.
Google has already been working on face recognition and song identification software for YouTube clips, with object recognition now looking like the latest string to its bow.