Google's new machine learning API recognizes stuff in videos, which may not necessarily seem like it at first but that's a big step for machine learning technology. Google broke the news about the new API, which they're calling the Cloud Video Intelligence API, at their Next Cloud conference earlier today, and they've showcased a little bit of detail on how the new API works and what it can do, while also showing off the API in action.
In the image below, which is a screencap Google shared, a video is shown playing with a tiger in it, and to the right of the footage is a list of different things that the API is determining might be in the video with percentages to measure what is most likely being displayed. Granted, what the video is showing is a tiger, and tiger is listed at 90.11% and is the third item down on the list, but the other two things above it are the words animal and wildlife, both of which are also correct determinations. Google uses Tensor Flow and other deep-learning models as the backbone for the API's build, and it's currently in a private beta that Google is allowing developers to sign up for if they're interested in testing the out the API.
Prior to Google's new API, computers recognizing and understanding what things being displayed in videos were was possible by utilizing manual tags, but the new API now actually makes it possible for items to be identified and understand what the video is without using those. For instance, if the video being played is a commercial for a particular product, the Cloud Video Intelligence API would essentially be able to determine what the product in the video is and recognize that the video was of a commercial. Moving back to the tiger, Google states that with this API, developers could also run a search for content in videos that are stored in Google Cloud Storage using a keyword, such as "tiger," and have the results come back with any relevant video footage that contains tigers. All of this is stuff that isn't really directed at people at the consumer level, and is really aimed at companies who have a need for cloud storage and cloud computing solutions, but eventually this technology would end up making its way to consumers in some form or another as companies adopt Google's Cloud platform and utilize the new tech in their services.