BEACON TRANSCRIPT – Perhaps you were impressed by Facebook’s auto tagging feature, but they are preparing more photo features. Now, Facebook will be able to recognize where you took a photo, what you were wearing in the photo, and also what you were doing.
Joaquin Candela is Facebook’s director of the department of applied machine learning. He explained how Facebook could recognize all the objects in photos even when no other clue words were posted.
Facebook already use Lumos, which is an AI image recognition program that can see what is in a picture even if it doesn’t feature any description. They used this program to recognize nudity or spam. Now, they are upgrading the program so that it will be able to identify actions, not only images.
Facebook has an automatic alt text that helps those with sight problems understand what is going on in photos. Now, this alt text can recognize up to 12 actions. The same technology will allow all users to search for images taken in a particular place or containing particular objects. You may say that this was possible before, but only for photos with a description. Now, any photos may appear in the search field, even if there is no text attached to them.
The technology used by Lumos is machine learning. This means that, the more photos it gets to tag, the more accurate its recognizing capabilities become. This program learned actions after about 13,000 users tagged photos without description and it was used these labels the humans applied.
Thus, if you type something into the search tab and then click on images, the search box will display photos from your own news feed, as well as public photos that do not necessarily feature a description.
Candela announced that the next step is the labeling of videos. This development is quite revolutionary and it represents only the beginning of what can be made with machine learning technology. This technology will surely get more and more advanced and Lumos will get better at recognizing photos and even moves further to videos.
Thus, the experience of the Facebook users will greatly improve and offer them a tool that is meant to reduce their efforts when searching for information.
Image Source: Pixabay