Google has plenty of amazing products and it’s latest developments in the field of image identification is pretty groundbreaking.
The internet is full of images, many of which come with little or no descriptions. That may not seem like a problem until you consider how they should all be sorted or searched.
Accurate image descriptions are also vitally important for making the internet a more accessible place for people who speak different languages or those who are visually impaired for example.
A team at the Google Brain Team debuted a machine learning system to automatically caption images in 2014 and it has just been getting better ever since. As of this month, the software can identify the main elements of an image with almost 94% accuracy.
That’s pretty amazing, and comes about through a combination of smart learning software and human interaction. The program is capable of taking in information from images captioned by humans and using that knowledge to increase its accuracy with every new piece of data.
Google is sharing this technology with the world, helping to make the internet a less daunting place for everybody, one picture at a time.