Audio Driven Face Reenactment Method Integrating Facial Deep-perception
Face reenactment is a conditional facial generation mission.The current face reenactment method based on audio driven is difficult to generate a complete and high quality face.As for such a problem,the audio driven face reenactment method integrating facial deep information is proposed.Meanwhile,the method adopts a light-weight model framework to reduce the model size and to improve the running speed.The experiment is compared with three kinds of the latest audio driving face reenactment methods on the AnnVI dataset.The results show that the proposed face reenactment method,integrating facial deep perception can greatly improve the quality of generated facial images driven by the audio.
NETL
NSTL
万方数据
