Google can use machine learning to identify objects in videos

All the fun of image recognition applied to massive media databases.

Paul Hanna / Reuters

You've been able to use a reverse Google search to hunt down similar photos on the internet for years thanks to image recognition and processing tools, but only for static pictures. Today, at Google Cloud Next '17, the search giant unveiled a new API that uses machine learning to search within videos for nouns and verbs occurring therein. Finally, you'd be able to hunt down that one movie about the guy who does the thing.

The Cloud Video Intelligence API is now in a private beta, meaning companies will get their hands on it long before you can use it to hunt down esoterica on YouTube. At this point, it is an enterprise solution, a deep-learning tool built on frameworks like TensorFlow that companies can use to parse through their stored videos and extract metadata. If you wanted to hunt through your vast media collection to hunt down "tiger," for example, you'd get a result like this:

The API searches for "entities" like nouns found in videos and indicates when that object appears. It can even detect when scenes change. Companies have must store their media on Google Cloud Storage to run the annotating software, but getting onboard with their cloud suite would be a decent idea, given that Apple, Evernote and Spotify started using the search giant's Cloud platform this year. Don't expect the intricate search functions to find their way into YouTube just yet, but as with most of Google's tinkering with parsing data, it will probably find its way into your life soon.