Video search will be a lot more refined, thanks to Google’s new machine learning API

Google has just made it easier to search within videos, thanks to a new API it has developed that mixes cloud learning with location tracking. Google unveiled the new feature it has developed during its Google Cloud Next Conference held in San Francisco recently.

What is it?

Essentially, it is a smart tool that allows for big companies which deal in media applications to search for videos using nouns, adjectives and verbs that are attributed to those videos. All of those videos will be searched through, and identified by Google’s video storage cloud. The API is called Cloud Video Intelligence and makes it now much easier to search videos that have certain objects and making them discoverable. It bridges the Machine Learning from their famous Tensor Flow technology and applies it on a wider scale, making the database smarter in identifying the objects in the videos.

The application is not open to just regular consumers, but for businesses who want to expand their video catalogue and to integrate it into their curations. It is a great way to search for crowd-sourced content as well, so they can use it to make video development software. Google’s blog indicates that the company has been working closely with media companies over the years to expand the level achievability when it came to searching for video content, especially for a format like video which is unstructured and cannot be verified by the database as much as an image or a website is.

How it works

An API, or application program interface, allows a connection to form between service and an application. This API connects to the Google Cloud Machine Learning platform to find the videos specified in the search, which are annotated by Google Cloud Storage. Once connected to the Machine Learning platform, the user can take advantage of it search for what they want, like moments within a video, frame by frame, each denoted with their own identification markers.

The API will learn the video and its contents in real-time and attribute key identification markers such as nouns and verbs to it. Similar to how one would search through a text document, for comparison. The API will also allow for companies to search for specific signals that are hidden underneath noise in a video for purposes of development and improvement of those signals.

The videos will be searchable only in the company’s directories, much like how we search for photos in our Google Photos app. The API does this by attaching labels to its videos and whenever a certain object appears in a specific scene or portion of the film, it is marked upon, so the user will know exactly how many times it has appeared in the video.

Is it available?

Yes, and no. Currently, it is beta and is accessible by only select partners of Google such as Cantemo, who has a video management tool and it will allow the program to connect to this API for easier searching of videos and elements within videos.