Facial expression synthesis method based on geometric comparative generative adversarial network

Abstract

The invention provides a facial expression synthesis method based on a geometric comparative generative adversarial network, mainly comprising representation, composition, structuring and setting of the geometric comparative generative adversarial network. The process includes giving input faces and target expressions specified by a group of facial marks; identity maintaining faces can be generated by guidance of a target expression formula; the geometric comparative generative adversarial network is composed of a facial geometric embedded network, an image generator network and an image discriminator network; the facial geometric embedded network is used for mapping the facial marks to semantic manifold in a latent space through comparative learning; the geometric comparative generative adversarial network performs training in multi-task manner, wherein comparative learning, adversarial learning and reconstructive learning are used; a joint target is finally formed. The geometric comparative generative adversarial network is provided herein; the manifold is sensitive to geometric changes in global and local faces; the method is also suitable for other facial expression characteristics; the generated expressions have high fidelity and resolution.

Description

technical field [0001] The invention relates to the field of facial expression synthesis, in particular to a facial expression synthesis method based on geometric contrast generation confrontation network. Background technique [0002] Human face is the most important external feature of human beings. Most of human expression and communication are carried out through facial expressions. Therefore, human expressions play an extremely important role in human-to-human communication. With the in-depth research in related fields such as affective computing, the analysis and synthesis of human facial expressions has become a research direction that has attracted much attention in the fields of computer vision and pattern recognition. , content-based image retrieval and medical treatment have broad application prospects. For example, in the production of movies, advertisements, and games, producers use expression animation technology to create vivid expressions for characters, add...

AI Technical Summary

Problems solved by technology

However, facial expression synthesis remains a persistent challenge at the semantic level due to the highly nonlinear nature of facial expression variations and individual differences.

Images