I. Introduction
Given an image of a person, we may envision what that person might look like in a different stance. This is possible to a model of human posture transfer that was trained by observing numerous people in various contexts. The goal of human pose transfer is to create full-body photo-realistic person representations based on target positions. This topic has several applications. in film-making, data augmentation for training human pose estimation systems or person re-identification or computer-graphics manipulations. The production of pose-guided human images is a difficult task because of the following factors:(1) The distributions of clothing, body features, backgrounds, and positions in human images vary substantially (2) One person in various poses might have drastically different visual attributes; and (3) The generator network should usually infer the appearance regions of body parts that are not visible in the source image.