Facial expression generation method based on generative adversarial network

Abstract

The invention discloses a face expression generation method based on a generative adversarial network. The method comprises the following steps: constructing a deep learning network model, wherein the deep learning network model comprises a recurrent neural network, a generator, an image discriminator, a first video discriminator and a second video discriminator; the recurrent neural network generates a time-dependent motion vector for an input image, the generator takes the motion vector and the input image as input and outputs a corresponding video frame, the image discriminator is used for judging the authenticity of each video frame, the first video discriminator judges the authenticity of the video and classifies the video, and the second video discriminator controls the authenticity and smoothness of the generated video change; using sample images containing different expression categories as input to train the deep learning network model; and generating a face video in real time by using the trained generator. Facial features are reserved while expressions are generated, the generated video keeps continuity and authenticity, and generalization ability is achieved for different faces.

Description

technical field [0001] The present invention relates to the technical field of computer vision, and more specifically, to a method for generating human facial expressions based on a generative confrontation network. Background technique [0002] In terms of face generation, 3DMM (face 3D deformation statistical model) generates faces by changing parameters such as shape, texture, posture, and illumination. DRAW (Deep Recursive Writer) uses Recurrent Neural Network (RNN) to realize image generation, and Pixel CNN uses Convolutional Neural Network (CNN) instead of RNN to realize pixel-by-pixel image generation. [0003] After the emergence of generative confrontation network (GAN), it has been widely used in image generation, and more and more GAN-based models have been applied to facial expression conversion. For example, ExprGAN (Expression Editing Based on Controllable Intensity) combines conditional generative adversarial networks and adversarial auto-transcoders to achie...

AI Technical Summary

Problems solved by technology

[0006] After analysis, in the existing schemes for generating facial expression videos based on deep learning, videos are usually generated based on noise, but due to reasons such as a small facial expression database, the generated faces are relatively single, and faces cannot be specified; Video models are less effective at facial expressions

Images