Social networking site Facebook is set to introduce a new system that would allow those with visual impairments to understand the contents of photographs uploaded to the site.
The company is set to release a navigation software that can decode images and change them into a format that can be read out to the user, so that they know what is in the image.
The news comes at a time when the internet is becoming dominated by images. For a long time, the internet was extremely text heavy, but the platform is becoming more photo and image-based, with estimates suggesting that 1.8 billion photos and images are uploaded each day. Photo-led networks like Instagram and Snapchat are at the forefront of this shift.
This has lead Facebook, the world’s largest social network, to develop a system to assist those who cannot see images, or struggle to see them. At present, blind and visually impaired people use navigation software known as screenreaders that can make computers usable by reading text through speech output, or translate it into braille with no way of letting the user know what is in an image.
Through the use of artificial intelligence, Facebook’s servers are now able to decode and describe a set of images uploaded to the site and then translate them into something that can be recognised and read out by the screenreaders.
The technology behind the system is currently able to recognise up to 80 objects and activities across a variety of contexts that regularly appear in photographs. This includes forms of transport such as cars, planes, trains, boats and bikes, as well as landscape and environmental-related items such as mountains, the sun, grass, the sea and trees. Furthermore, it can define specific sports and tell the user which ones are being played in the images, as it can recognise tennis, football, basketball, swimming and golf, as well as where they are being played, such as a stadium.
As you can imagine, it can recognise food items, assuming they’ve been cooked correctly! It is so far able to recognise coffee, pizza, and even sushi. Another important category that the new software can decipher is appearances. This can tell the user who is in the photo and what their expressions may be. It can distinguish between adults and babies, and can recognise glasses and beards, as well as whether the person in the image is smiling. For those who like to keep up-to-date with the language of the millennials, it can even tell you if the photo is a selfie!
The software was developed by Facebook engineer Matt King, who lost his own sight due to retinitis pigmentosa; a disorder that destroys light sensitive cells within the retina. King says that the more objects that the system scans, the more sophisticated it will become.
As the technology is still fairly new, it describes images in a fairly basic manner by saying something along the lines of “there are three people in this picture and they are all smiling.”
According to the head of Facebook’s accessibility team, Jeff Wieland, the company is investing in strategies to enable different communities to be able to access and engage with Facebook.