Facial Expression Transfer Based on Conditional Generative Adversarial Networks

With the development of computer vision and image transfer, facial expression transfer has been more widespread applications. But there are still some problems, such as lack realistic expression, poor retention identity features low synthesis efficiency. In order to solve problems paper proposes a model based on conditional generative adversarial network, which can generate highly face with source target features, when gave image. The consists two parts: feature point fusion module module. Among them, uses an auto-encoder encode key face, so information corresponding points image; image, then generates through modified U-net network. is finally validated publicly available datasets, RaFD CK+, experimental results show that generated than pix2pix model, only needs be trained once complete between any expression.