Google is already capable of identifying objects in images thanks to machine learning. In fact, you can try that out right now. Say "OK Google," to your smartphone, ask it for context-specific photos by commanding "Show my photos of food," and it'll do exactly that. Granted, it isn't flawless, but it's still highly impressive.
Now, that capability is expanding to video. At its Cloud Next conference in San Francisco on March 9, Google announced the launch of a new machine learning API for automatically detecting and recognizing objects in videos and making them searchable.
Google's Machine Learning Learns How To Spot In-Video Objects
The new Video Intelligence API will pave the way for developers to create applications that can automatically extract information from videos. As previously mentioned, this technology was only realized using still images prior, but with the help of a new API, developers will finally have the freedom to build applications that can allow users to search information in videos. TechCrunch notes that this means users can search for specific objects, such as cars, or even animals, such as tigers.
The API was demoed onstage by playing a commercial. It was able to identify the dachshund in the video when it appeared, and it even understood that the video was a commercial. In another demo, entering a simple search term "beach" brought up videos which had scenes from beaches in them.
Before Google's new API, computers weren't really able to recognize and understand the context of a video independently, or without the help of manual tagging.
"We are beginning to shine light on the dark matter of the digital universe," said Fei-Fei Li, Google Cloud's chief scientist of artificial intelligence and machine learning. While Google's own demo of the API was indeed impressive, it's still up to developers whether they can push the technology further, but with the API democratized, it's easy to imagine that it'll go well. Developers may sign up for a private beta via the Google Cloud Platform site.
Li also noted during the keynote that the world of pixels extends past images. Google wants to make the roster of machine learning technologies it has been cooking up in-house widely available. This new Video Intelligence API is just another fine example of that.
Google Cloud Platform
The keynote was about Google's serious commitment to being a player in the emerging game of cloud-based services and platforms. As noted by The Verge, Google, while offering a competitive platform for potential partners, is still behind Amazon and Microsoft in terms of cloud services. The problem is that enterprises Google needs to pluck from the reservoir are already tied to other cloud companies.
If Google, however, continues wooing companies with stellar, promising incentives such as a machine learning API for object recognition in videos, it might just be the next platform to start turning to.
Impressed with Google Cloud Platform's Video Intelligence API? Feel free to spitball ideas about its potential implementations in the comments section below!
