Patent number 8,254,699, automatic large scale video object recognition, describes software that performs “dimensionality reduction and consistency learning on visual content items.” Basically, it reduces shapes to math, compares them, and learns which are similar.
[aditude-amp id="flyingcarpet" targeting='{"env":"staging","page_type":"article","post_id":522121,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"business,media,security,","session":"D"}']Essentially, it’s artificial intelligence for images.
This is foundational technology for teaching software to recognize objects, which would help Google determine both what is in a video, and, based on that information, where the video was shot. For example, if the Washington Monument shows up, chances are that the video was shot in D.C.
AI Weekly
The must-read newsletter for AI and Big Data industry written by Khari Johnson, Kyle Wiggers, and Seth Colaner.
Included with VentureBeat Insider and VentureBeat VIP memberships.
The technology would also presumably enable far better video search, as Google would not have to rely on partial or even missing metadata such as tags or descriptions to know what a video is about — it would have its own data about the video, auto-generated from the video itself. And better video data, of course, would also enable better YouTube monetization, since Google could apply AdSense-style mechanics (people might buy things relating to what they’re searching for) to video viewing and advertising.
Google’s image database holds at least 50,000 objects, the patent states, including things as different as phones, aardvarks, swimming pools, Porsche 911s, and Bill Clinton.
Which means you should soon be able to search YouTube for videos of Bill Clinton riding an aardvark into a swimming pool. (Surely they exist!)
Until, of course, people start hacking the algorithm by incorporating desired backgrounds into their videos.
Hat tip: C|Net
VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Learn More