AI can reconstruct a picture of your face into these weird and wonderful 3D images
Tech's talk of the town is all about the iPhone X facial recognition feature that can scans your face and create the already infamous "animojis."
SEE ALSO: The Sony Xperia XZ1's camera can 3D scan faces into augmented reality avatars
But what if you don't have $999 to invest in the next-generation smartphone but still want to play with the 3D faces?
AI experts from the University of Nottingham and Kingston University now let you make a plausible re-creation of your face using just a selfie or a picture of yourself.
Try it yourself here, in the online demo of their paper entitled "Large Pose 3D Face Reconstruction from a Single Image via Direct Volumetric CNN Regression."
Here's what happened when I fed it with one pic of my face:
Image: gianluca mezzofiore/mashable
Here's my colleague Rachel. Not as accurate, but not bad either:
Image: gianluca mezzofiore/mashable
Aaron S. Jackson, Adrian Bulat, Vasileios Argyriou, and Georgios Tzimiropoulos, who worked on the research, admit that 3D face reconstruction is a problem of "extraordinary difficulty."
As the current systems work, multiple facial images from various angles are fed to the system which then needs to address several challenges, such as "establishing dense correspondences across large facial poses, expressions, and non-uniform illumination," researchers said.
To overcome these challenges, Jackson and others trained a neural network — called Convolutional Neural Network (CNN) — on a big fat dataset of 2D pictures and 3D facial models.
The result is that now the CNN can reconstruct the whole 3D facial geometry from a single, unseen 2D image. Not only that, it can also guess the non-visible parts of the face.
Pretty extraordinary, isn't it?
Image: http://www.cs.nott.ac.uk/~psxasj/3dme/
WATCH: Get the multipass powerup in this ‘Fifth Element’ 16-bit short
