After launching its 3D Photos feature back in 2018, Facebook is now using AI to create 3D photos from pretty much any standard 2D picture.
Facebook’s 3D Photos feature is highly immersive and engaging, but it has one drawback: It relies on the dual-lens “portrait mode” available on a few high-end devices. This made it impossible for all other device users to create or share 3D Photos, thus limiting the number of Facebook users who could use the feature.
Now, in an effort to democratize the creation and sharing of 3D Photos on its platform, Facebook is using “state-of-the-art machine learning techniques to produce 3D photos from virtually any standard 2D picture.”
The way it works is simple in theory but complex in its execution. As researchers from Facebook Artificial Intelligence explain, the “system infers the 3D structure of any image.” Hence, it presented a variety of challenges, “such as training a model that correctly infers 3D positions of an extensive variety of subject matter and optimizing the system so that it works on-device on typical mobile processors in a fraction of a second.”
To be able to overcome challenges, Facebook AI’s team “trained a convolutional neural network (CNN) on millions of pairs of public 3D images and their accompanying depth maps,” and used other techniques developed in the past as well.
The result is that Facebook can now turn any 2D picture into a 3D Photo – whether it’s a new shot taken on any device or a decades-old photo in digital format.
The new feature also helps high-end device users who want to take 3D selfies on their front-facing cameras. Any users on an iPhone 7 or higher, or a recent midrange or better Android device, can now try the new feature on Facebook.
If you want to find out more about how Facebook built the feature, you can read all the details technical details here.
[box]Read next: Facebook Brings 3D Photos To Android, Stories, And The Web[/box]