Abstract

We present a method for imparting motion to a still 2D image. Our method usesdeep learning to segment a section of the image denoted as subject, then usesin-painting to complete the background, and finally adds animation to thesubject by embedding the image in a triangle mesh, while preserving the rest ofthe image.

 

Quick Read (beta)