Pose-Guided Human Animation From a Single Image in the Wild

Input Animated person in various poses from a single image Figure 1. We present a new method to synthesize a sequence of animated human images from a single image. The synthesized images are controlled by the poses as shown in the inset image. It is capable of generating full appearance of the person at diverse poses, reflecting the input foreground and background in the presence of occlusion and 3D shape deformation, e.g., the occluded texture of the back.