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