Facebook to use Artificial Intelligence to Help Blind People ‘See’ Social Media

More than 20k malicious apps use Facebook APIs
More than 20k malicious apps use Facebook APIs
Facebook Johannesburg
Every day, people share more than 2 billion photos across Facebook, Instagram, Messenger and WhatsApp. That’s a lot of photos!

Every day, people share more than 2 billion photos across Facebook, Instagram, Messenger and WhatsApp. That’s a lot of photos!

While visual content provides a fun and expressive way for people to communicate online, consuming and creating it poses challenges for people who are blind or severely visually impaired. As a result, some people may feel frustrated and excluded because they cannot fully participate in the conversations around photos that take place on Facebook.

Facebook’s mission is to make the world more open and connected. This means that we want everyone to have equal access to connect with others. Worldwide, more than 39 million people are blind, and over 246 million have a severe visual impairment. As Facebook becomes an increasingly visual experience, we hope our new automatic alternative text technology will help the blind community experience Facebook the same way others enjoy it.

What is automatic alternative text?
Automatic alternative text, or automatic alt text, is a new development that generates a description of a photo using advancements in object recognition technology. People using screen readers on iOS devices will hear a list of items a photo may contain as they swipe past photos on Facebook. Before today, people using screen readers would only hear the name of the person who shared the photo, followed by the term “photo” when they came upon an image in News Feed. Now we can offer a richer description of what’s in a photo thanks to automatic alt text. For instance, someone could now hear, “Image may contain three people, smiling, outdoors.”

While this technology is still nascent, tapping its current capabilities to describe photos is a huge step toward providing our visually impaired community the same benefits and enjoyment that everyone else gets from photos.

We are testing this feature first on iOS screen readers set to English, but we plan to add this functionality to other languages and platforms soon.

Try it yourself
You can try this technology for yourself. On an iOS device, enable VoiceOver by asking Siri to “turn on VoiceOver,” or by tapping on Settings > General > Accessibility > VoiceOver. Once enabled, you can open the Facebook app and swipe to scroll through your News Feed or a profile. When you swipe past a photo, you’ll hear this technology tell you some of the items the photo may contain.

AI-Powered tech
Facebook’s object recognition technology processes images uploaded to Facebook and our family of apps. The technology is based on a neural network that has billions of parameters and is trained with millions of examples. Each advancement in object recognition technology means that the Facebook Accessibility team will be able to make technology even more accessible for more people.

What objects are recognized?
We have started by identifying objects that are commonly shared on Facebook. Here are some of the items that can be identified by our system:

– Transportation words: Car, Boat, Airplane, Bicycle, Train, Road, Motorcycle, Bus
– Nature words: Outdoor, Mountain, Tree, Snow, Sky, Ocean, Water, Beach, Wave, Sun, Grass
– Sports words: Tennis, Swimming, Stadium, Basketball, Baseball, Golf
– Food words: Ice Cream, Sushi, Pizza, Dessert, Coffee
– Words that help describe someone’s appearance: Baby, Eyeglasses, Beard, Smiling, Jewelry, Shoes And, of course, Selfie!

Getting it right
We don’t want to make mistakes that create a negative experience for the very people we’re hoping to help. We are careful to only include items where we believe there is a high degree of precision and accuracy. We are committed to building and iterating on this technology so we can expand our photo descriptions.

The Backstory
The idea for an automatic photo captioning system came from interviews with people in our community who have visual impairments. They told us what it was like when their friends shared photos on Facebook, and we decided to improve this experience. The Core Data Science team initiated this project, but quickly gained support from a wide range of teams and organizations within Facebook, including the Facebook Accessibility team, the Applied Machine Learning team, the User Experience Research team, and iOS engineering.

Facebook’s Accessibility Team
The Facebook Accessibility team began five years ago to help make Facebook’s existing products usable by everyone. Now the team also works on new products specifically for people with disabilities. Automatic alt text is part of our effort to improve Facebook for people with blindness and vision loss, but the team has made progress in helping people with other disabilities including providing closed captions to help people who are deaf or hearing impaired, and ensuring Facebook can be used by people with limited mobility who use eye-tracking devices.

Facebook is also committed to making improvements that help the whole field of tech accessibility. We created the Accessibility Toolkit, a public resource covering best practices for designing and implementing an accessibility program. React, Facebook’s open source front-end framework for building web and mobile products, supports accessibility, and Facebook is a contributor to the Accessibile Rich Internet Application (ARIA) standard.

Staff Writer