The Automatic Alt Text system of Facebook would be able to recognize 10 times more concepts than before.
The Automatic Alt Text (AAT) system from Facebook can now detect and recognize 10 times more objects and concepts than it could when it was first launched, the company reported. Descriptions will also be more detailed, and events, landmarks, types of animals, and more will be recognizable by AAT. Facebook announced updates to AAT, a technology launched by the company in 2016 that uses object recognition for visually disabled people to create explanations of images on Facebook and Instagram.
Over 1,200 concepts are recognized by the enhanced AAT, Facebook said in a blog post. In the original version released in 2016, this is more than 10 times the amount possible. AAT will also provide data on the position location and relative size of elements in an image, Facebook said. Instead of saying “May be an image of 5 people,” for example, the AI would be able to specify that in the middle of the photo there are two people and three scattered to the fringes.
Although a compact overview for all photos will be offered by the new AAT by default, users will also have the option of obtaining comprehensive descriptions of photos that might be of more interest to them. A panel that has a more detailed explanation of the contents of a picture will be provided when they choose this option.
Facebook said it consulted users who rely on screen readers and found out that when it is from friends and family, they want more information about a picture, and less information when they are not. Positional details and a comparison of the relative importance of objects may also be part of the detailed definition.
Facebook reported that the updated AAT now reflects many advancements in technology that enhance users’ photo experience. In addition, to reduce the margin of error, terms that it could not define accurately were omitted.