Facebook has launched a new tool to help the blind and visually impaired help "see" images on its site.
While a number of assistive technologies are already available to help the blind read the contents of a screen, understanding the content of images is obviously a lot more difficult.
Starting today, Facebook will describe the contents of photos to the blind and visually impaired using a tool it calls "automatic alternative text".
The tool, which will work on iOS devices in English from today and will come to Android, the web and other languages soon, will dictate "a list of items a photo may contain" as the user swipes from picture to picture.
Until now, screen readers would simply read out the name of the person who had posted the picture and announced "photo". Now, users will hear the app say, for example, "Image may contain: Tree, two people, sky, outdoor, etc".
'Billions of parameters'
Facebook's object recognition technology is based on a neural network with billions of parameters and which is, Facebook says, trained with millions of examples. The more pictures that Facebook's software scans, the more intelligent it will become.
Of course, we've been seeing the benefits of Facebook's object recognition technology for a long time, recognizing our friends' faces and differentiating between people and objects.
Facebook's automatic alternative text is an impressive and important step forward for the social network. It's still early days for the technology, but
"While this technology is still nascent," said Facebook, "tapping its current capabilities to describe photos is an important step toward providing out visually impaired community the same benefits and enjoyment that everyone else gets from photos."