Video Search Becomes A Reality!! With Google’s Machine Learning API That Recognizes Objects in Videos

Google is one company that can be seen to push forward in the world of artificial intelligence continually. So, the fact that announced a new machine learning API that is capable of recognizing objects in videos and making them instantly searchable is no real surprise is it?


This new Video Intelligence API allows developers to build applications that have the ability to extract objects from a video. Up until now, this has only been done through still images rather than videos, but moving forward this new API will allow so much more. Developers will be able to construct creative applications that let the user search for information relating to what it is they’re watching. This brings a whole new meaning to the words “watching a program” when you can find out more in-depth information about any aspect of the program you want.

“In tests, Google Cloud Machine Learning enabled us to improve the accuracy and speed at which we analyze the images captured by our satellites. It solved a problem that has existed for decades.” – Mathias Ortner, Data Analysis and Image Processing Lead

 

This first image illustrates how difficult it is to guess the difference between snow and clouds.


With Cloud Machine Learning, automatic detection is able to identify that clouds are on the top right side (highlighted in red).

As well as being able to extract metadata from the video, the API also lets the user tag scene changes. These videos are then stored in Google’s cloud storage platform. There are demos available online if you want a little sneak peek or if you’re a developer you can even sign up for the private beta version. While video may have previously been a pain for researchers, in terms of machine learning capabilities, Google’s new service makes extracting data from them as easy as it was to do this with still images. Google is also pushing forward with its Cloud Machine Learning Engine that uses its TensorFlow framework to allow the creating of custom learning models.



More News To Read

Comments

comments