Face Swapping Method Based On Convolutional Neural Network

Face swapping is using image processing technology to replace the face of the target image with the face of the source image.This technology has important research significance and application value in social entertainment,virtual reality,film and television production,games and other aspects.Existing face swapping methods are limited by face pose consistency and cannot overcome the influence of occlusion on swapping.To solve the above problems,this paper proposes a face swapping method based on convolutional neural network.Firstly,the cascade convolutional neural network is used to detect the face and then,using the full convolutional network to segment the face to determine the replacement area.The Wallis transform is applied to adjust the skin color of the face in the source image to make it consistent with the skin color of the face in the target image.Then,the face triangulation network is constructed by using the facial key points detection technology based on regression forest and the Delaunay triangulation algorithm.Finally,Poisson fusion is applied to eliminate splicing traces between different images.The results of qualitative and quantitative experiments show that the proposed method is effective and can solve the problem of face pose consistency limitation and face occlusion.On the basis of theoretical research,this paper also constructs a minority faces database in order to designs and implements the minority face swapping system.When using the ethnic face image swapping system for face replacement,the input face can be replaced to characters of different ages,genders and nationalities in the database,and it can also be used to be replaced with any face in the group photo.The system interface is friendly,and the output result is lifelike and natural.Compared with image process software,our system doesn't need complicated man-machine interaction operation,which can save a lot of manpower and time,and has certain practical application value.