SeHAGAN: GAN을 이용한 순차적 인간 행동 생성

초록

The Generative Adversarial Networks (GANs) have shown rapid development in different contentcreation tasks. Among them, the video generation gets its own attention due to the development of various human-centric applications like avatar animation. In this paper, we proposed a method to generate sequential human actions using a two-stage GANs pipeline. First, we produce pose skeleton with our Poses Generator, and then we textured them with a Frame Generator. Results showed that the proposed method SeHAGAN generates a plausible and high-quality video of human movements.