Moving photographs from Harry Potter novels may not be far from reality, thanks to AI

In what could be exciting news for Harry Potter fans out there, those magical, moving photos in J.K.Rowling's novels may not be far from reality. Magical photos in the magical world of Hogwarts School of Witchcraft and Wizardry are perfectly normal except that characters in those photos smile, wave and some cases, disappear. Those things must be imaginary for ordinary muggles, isn’t it? Well, not anymore.

Thanks to Chung-Yi Weng and the team at the University of Washington in Seattle. They have built an AI-powered software called Photo Wake-Up. It can create 3D animations from still images. In a nutshell, articulating a 3D human pose from a 2D image is not easy for machine vision and deep-learning algorithms. Experts and computer science teams across the world have often faced difficulties in this area in the past.

Also Read | Meme lovers beware, Hackers control malware using Memes shared on social media

To deal with Human Pose Estimation, Weng and the team use SMPL, a program developed by Microsoft in association with the Max Planck Institute for Intelligent Systems in Germany. In this process, a 3D skeleton is superimposed on a 2D cutout of a human body. The sense of movement is created using the animation, which essentially solves the problem of pose estimation to some extent.

It can handle simple movements if not the complex ones due to certain limitations. Still, this process cannot create a realistic animation. That’s where Weng and the team come into the picture. Their main focus is to tweak 2D cutouts in such a way that produces realistic 3D animations. Their focus remains on the head, right arm, right leg, left arm, left leg, and torso.

Also Read | Xiaomi Mint browser contradicts own claim about the ad-free experience

Here is what Weng and the team have to say:

“We believe the method not only enables new ways for people to enjoy and interact with photos, but also suggests a pathway to reconstructing a virtual avatar from a single image while providing insight into the state of the art of human modelling from a single photo.”

Via