Machine learning technology is enabling Google to search for objects in video footage, the company as revealed.
(CCM) — Google now has the power to search through videos and recognize objects using machine learning, the company revealed at its Cloud Next conference in San Francisco. It plans to make the technology freely available using a new Video Intelligence API, enabling software developers to use the technology to build applications which can scan through videos to spot specific items such as cars or airplanes, or even specific animals such as elephants or lions. Until now it has not been possible for search engines to identify objects in videos without manual tagging. "We are beginning to shine light on the dark matter of the digital universe," said Fei-Fei Li, a Google artificial intelligence expert.
During a demonstration of the new technology Google played a short commercial, and the API was able to spot the dachshund in the footage, and understand that the video was a commercial. In a second demonstration the API was tasked with searching for "beach", and succeeded in finding videos which contained beach scenes in the same way that Google Photos can search a photograph collection for images of beaches. Google Photos was launched in May 2015, but it may be some months before the first services that utilize the Video Intelligence API start appearing.
Image: © Evan Lorne - Shutterstock.com