We all know that the tech giant Google already has technology that can identify objects or certain types of details in images. But, now the tech giant Google’s AI (Artificial Intelligence) can identify what it is seeing in videos.
Google’s AI Can Now Identify What It’s Seeing In Videos
Google already has technology that can identify objects or certain types of detail in images. This is useful for helping Google Photos do automatic photo sorting, for example. However, it did not stop there: the company announced a machine learning system that can identify objects in videos.
Technology is an API called Cloud Video Intelligence. For now, it is only available as a beta to a closed developer group. This is a clear sign that the system needs to be cut before being offered openly. Even so, the current stage is entirely convincing.
This is the week that the tech giant Google hosts the Cloud Next conference. Fei-Fei Li, the artificial intelligence leader of the Google Cloud platform, took the opportunity to show the technology in action.
Whoever was on stage saw the API identify a dachshund dog and then understand that the video is a TV commercial. The next demonstration was just as impressive: in response to a search for “beach,” technology has been able to locate videos with scenes of that kind of place.
If you search for “beach” on YouTube, you’ll certainly find videos related to the subject. But the service primarily uses keywords, tags, and descriptions given by users to sort content. Cloud Video Intelligence can do an equivalent job, but alone, without relying on human help.
Google just showed off a new Video Intelligence API that lets you search for objects within video clips, here's an bit of the demo pic.twitter.com/wOPwsmnAfe
— Karissa Bell (@karissabe) March 8, 2017
According to Li, Cloud Video Intelligence was developed based on frameworks such as TensorFlow and then submitted to a large-scale database: YouTube, of course. The algorithm learns to distinguish objects, entities (animals, people, flowers, etc.) and actions (swimming, flying, navigating, among others). The technology is also able to contextualize: a search for Tiger also shows related content such as wildlife, zoo, and cats.
At Cloud Next, Google went to great lengths to make it clear that it’s a serious player in cloud-based applications and artificial intelligence. In this sense, the Cloud Video Intelligence API appears as part of the various features of the Google Cloud platform.
However, I will not be surprised if Google harnesses this technology on YouTube, but at least in the initial phase, the API will only be available to organizations using the Google Cloud Storage service (including companies like Evernote and Spotify).
So, what do you think about this technology? Share your views and thoughts in the comment section below.