Facial expression editing system and method and terminal

Abstract

The invention provides a facial expression editing system and method and a terminal. The method comprises the steps: calculating a relative action unit vector from a source expression action unit vector to a target expression action unit vector through a facial action unit vector difference module; taking the relative action unit vector as an input condition, sending the relative action unit vector and the source face image into a synthesis network generator together, and synthesizing a picture conforming to the target face expression; predicting a source expression action unit vector and a target expression action unit vector by utilizing an expression condition discriminator; meanwhile, predicting action characteristics of the generated face picture to calculate expression loss, and optimizing the face expression edited by the synthesis network generator; and a synthetic network discriminator is utilized to perform true and false scoring on the generated face image to calculate the adversarial loss, and the quality of the generated face image synthesized by the synthetic network generator is optimized. According to the method, a coding and decoding network structure, a multi-scale feature fusion mechanism and a face action unit vector difference module are fused, and high-quality expression synthesis is realized.

Description

technical field [0001] The present invention relates to the technical field of computer vision, in particular to a human facial expression editing system and method, and a terminal. Background technique [0002] Facial expression editing is a very challenging task in the field of computer vision. Its goal is to transform the facial expression in the original image according to the input relative vector of facial expression while maintaining the head posture of the human face. , identity information and other information remain unchanged. This technology has quite a wide range of applications and can be extended to many emerging application scenarios such as photo editing, film production, and virtual character animation. Although generative adversarial networks have achieved excellent performance in domain transfer tasks such as face attribute conversion and makeup conversion in recent years, there is still a lot of room for improvement in character clothing conversion task...

AI Technical Summary

Problems solved by technology

[0004] Secondly, the existing methods are difficult to meet the requirements of high-quality facial expression editing.

The existing facial expression editing methods either use the traditional geometric deformation method, which usually requires a third determined face picture as a reference, so the ability to edit fine-grained facial expressions is very limited and cannot Meet high-quality, flexible editing requirements

Not only that, but the existing methods use the input characteristics of the global facial expression category as conditional information to train the network, which does not make full use of finer-grained and more local expression methods, so it is not enough to support fine-grained expression face generation

Images