A Method for Face Fusion Based on Variational Auto-Encoder

Face fusion refers to fuse two different facial images into a new face image that retains the facial features of the original image. Our network model is combine Variational Auto-Encoder(V Ae)and Generative Adversarial Networks(GAN), which achieved the end-to-end fusion task. Not only does it guarantee the quality of the fusion image(the image generated by GAN is sharp and photorealistic), but also doesn't lose specific details of the face(This is guaranteed by VAE). In the end, the experiment achieved a promising result on the CelebA dataset.