If you’re trying to remember something in a video that you want to point out to a friend, you’ll probably need really good memory to find out which video it is, and at which point in time in the video it appears. This might be frustrating especially if the object in question appears for a brief moment and the video is particularly long.
Obviously scrubbing through the video would be quite inefficient, but soon that might not be necessary because Google has been working on new technology in the form of Google Cloud Video Intelligence that uses deep-learning models that could analyze videos, tag them, and thus create a system in which objects within the video can be searched for.
Of course not every object carries the same weight, for example maybe a blade of grass might be less significant than a person wearing a funny hat, and it is these distinctions that Google’s AI will be trained to weigh and give a percentage value. This will determine how prominent it will feature in search results, which could give rise to a platform in the future where we can search through videos the same way we search for websites and images.
At the moment this tool is available to companies in a private beta mode and according to Google’s chief scientist in machine learning Fei Fei Li, “This API is for large media organizations and consumer technology companies, who want to build their media catalogs or find easy ways to manage crowd-sourced content.” We’re not sure if or when it will become more mainstream for end users, but for now it seems like it is a work in progress.