Facebook introduced a new artificial intelligence tool on April 5, Tuesday that provides detailed descriptions of photos for the social network's blind and very visually impaired users. The goal of the new tech is to provide them with the same experience as other Facebook users as the social media giant becomes more visual. It uses object recognition technology to automatically create text about an image.
A blog post was made by Facebook's Head of Accessibility Jeffrey Wieland. The company reports that worldwide over 39 million people are blind and over 246 million people have major visual impairment.
Visually impaired Facebook users can use screen readers on iPhones and iPads to hear a photo described. This includes meticulous details such as the number of people in an image, their facial expressions, or what they are wearing.
The update is a big improvement from the previous tool. It would say who shared the photo as well as the word "photo," but would not provide more details, according to ABC News.
Facebook's object recognition tech is based on a neutral network. Such computer systems function like a human brain by becoming more intelligent as they process more data. Facebook reports that its neural network includes millions of examples.
The audio photo captions first share how many people are in a photo then state whether or not they are smiling. Next, the AI describes the objects in the image by first naming the items the algorithm is most sure it sees. The description ends with by explaining whether the picture is indoors, a selfie, or a meme.
Facebook's AI system is still a work in progress and when it is unsure about objects in a picture it will share that the image might contain the things.
The object recognition tech is now only available for iOS gadgets and in English. However, Facebook has plans to add other platforms and languages in the future.
In related news last week Microsoft unveiled its Seeing AI app that it is developing. The app requires a smartphone camera or smart glasses to identify things in a blind person's environment including people and objects, according to Tech Insider.
Here's a video on the AI tool's development: