Facebook’s New AI App Lets You Convert 2D Photos Into 3D
Facebook becomes updated its “3D Photos” feature technology to allow users to create a 3D photo out of any 2D image from smartphones with one camera.
Facebook is making a tool that converts 2D images to 3D structures more widely available, rolling it bent most iPhone and Android users, the corporate announced Friday.
The App, called 3D Photos, was previously only available for high-end phones with multiple cameras or for desktop users willing to create their own depth map to guide the 3D rendering. Now, anyone with an iPhone 7 or later or a “recent midrange” Android device can use it.
Facebook is functioning to determine itself as a pacesetter in AI — the platform announced a slate of research projects spanning machine learning and robotics last year, signaling a willingness to compete with other tech giants like Apple and Google that are vying for dominance within the field.
This work will support us greatly 2D images more usually know the content. the thought of 3D scenes could also help robots navigate and interact with the physical world,” Facebook said during a statement. we’ll support the AI community to make progress in these areas. We expect that by sharing details of our 3D Photos system.
Here’s a breakdown of how the machine learning-powered app works and the way to use it.
For those unaware, Facebook had first revealed its 3D Photos feature at the 2018 F8 developer conference, which was unrolled a couple of months later. Until now, this feature trusted the dual-lens “portrait mode” capabilities, which was only available in new, high-end smartphones.
The app uses a neural network that estimates how “far” each pixel during a photo is from the camera to make a “3D photo.”
The social media giant has now improved its 3D Photos feature technology to support images taken from single, rear-facing camera phones. The feature is now available for anyone with an iPhone 7 or higher, or “a recent midrange or better Android device,” says Facebook.
The AI was trained on an outsized set of knowledge, meaning it works on content starting from paintings to photos, consistent with the corporate.
“To bring this new visual format to more people, we’ve used state-of-the-art machine learning techniques to supply 3D photos from virtually any standard 2D picture. this technique infers the 3D structure of any image, whether it’s a replacement shot just taken on an Android or iOS device with a typical single camera, or a decades-old image newly uploaded to a phone or laptop,
“This advance makes 3D photo technology easily accessible for the primary time to the various many people that use single-lens camera phones or tablets. It also allows everyone to experience decades-old family photos and other treasured images during a new way, by converting them to 3D.”
Facebook says it’s continuing to enhance its AI to urge better at depth estimation in real-time scenarios, like augmented reality.
To create this effect for single-camera photos, Facebook has deployed a machine learning-based method that infers the three-dimensional structure of a picture to make the 3D effect.
The app is out there as of Friday to people using iOS and Android, Facebook announced.
“We trained a convolutional neural network (CNN) on many pairs of public 3D images and their accompanying depth maps, and leveraged a spread of mobile-optimization techniques previously developed by Facebook AI, like FBNet and ChamNet,” the corporate added.
When posting a photograph to Facebook, users are going to be ready to click a “3D Photos” button to convert their pictures, consistent with Facebook’s announcement.
How To Use The Feature?
To use the feature, you’ll first get to update your Facebook app. Further, follow the regular process to post content on the News Feed. to try to do this, you would like to travel to “What’s on your mind”, select “3D Photos”, and pick any image of your choice from your camera roll to post it with a 3D effect.
Note, this is often just a simulation and therefore the feature won’t make images truly three-dimensional just like the ones captured by a dual-lens camera. However, it’ll still respond once you scroll, pan, or tilt your screen to ascertain it from a couple of different angles.