After patenting the ability to automatically detect faces in videos, Google has now bagged another patent for automatic detection of objects in videos. The new technique enables software to identify objects in videos automatically using “feature vectors.”
With Google’s automatic object recognition, humans don’t need to do anything, as the system would make use of “feature vectors” to aid in the recognition of the objects in a video or still image. Users can manually tag objects in videos to let Google know what is the thing called. Google will use this manual tagging when Google’s automatic tagging fails, and Google may collect the manual tagging data to improve the data in the underlying database. But while the automatic tagging system will work, the system will use images from a “visual content repository”.