A Multi-person Expression Migration Method Based on Generative Adversarial Mechanism

Abstract

The present invention provides a multi-person facial expression transfer method based on a generation confrontation mechanism, including a Landmark converter (ψ) and a face generator referring to an image in a stationary state of a person as a corresponding landmark vector and a corresponding landmark vector of an expression image to be transferred as a pair Refer to the network structure for feature processing of the landmark vector in the static state of the reference person. The network structure for feature processing of the landmark vector for reference to the transfer expression of the person. The above-mentioned network structure and the features of the network structure are spliced ​​at the channel level and then sent to the network structure. The method of the present invention is based on The Py-Torch deep learning framework first converts the landmarks of human expressions in the geometric space, and then uses the design idea of ​​the confrontational generative model to combine the converted landmarks with the reference image to generate a real target image.

Description

technical field [0001] The invention relates to an expression transfer method, in particular to a multi-person expression transfer method based on a generated confrontation mechanism. Background technique [0002] Expression transfer technology has great application potential in many scenarios, such as film production, animation production, etc. With the development of deep learning and the increase of computer computing power, many expression transfer methods based on deep learning have achieved good results, which can realize the transfer of expressions between different people and generate realistic transfer face images. However, the existing methods generally can only be trained for a specific person. After the model is trained, only the expression transfer and face generation of the specific person can be performed. It cannot be applied to the scene of multi-person expression transfer at the same time, which greatly limits the Applicability of the method. In response ...

AI Technical Summary

Problems solved by technology

[0003] Aiming at the problem that the current method cannot use one model well to realize expression transfer between multiple people, the present invention proposes a multi-person expression transfer method based on a generative confrontation mechanism

Images