*Run it backwards, and it ought to INSERT objects into video.
*A big help with the ever-present AR registration problem, too.
via @mbanzi
(...)
"Google, using the software, wants to scan and analyze the content within videos, YouTube videos we assume, and look for real world objects, identify them, and make a catalogue out of those objects. The patent describes Google’s technology of scanning a video, picking out landmarks, objects and context; with subsequent tagging categorization.
"The search engine giant is of the opinion that conventional systems that may be carrying out such tasks and identifying objects more or less rely on direct human input that is neither scalable nor cost efficient.
"'… such human input is expensive, time-consuming, and cannot scale up to handle very large data sets comprising hundreds of thousands of objects and millions of images', notes Google in its patent...."
