Face image processing model training method and device, electronic equipment and storage medium

Abstract

The invention discloses a face image processing model training method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the coding of a first faceimage through an encoder of a face image processing model, and obtaining a first feature vector of the first face image; splitting the first feature vector into a plurality of sub-feature vectors, wherein the sub-feature vectors at least comprise an identity sub-feature vector and an age sub-feature vector; determining a to-be-decoded vector according to each sub-feature vector, and decoding theto-be-decoded vector by using a decoder of the face image processing model to obtain a second face image; and optimizing parameters of the face image processing model according to the regression lossvalues of the first face image and the second face image. According to the face image processing model obtained through training, the dependence on data distribution is reduced, the method is more robust to long-tail data with unbalanced ages, and a face aging image with a better effect can be generated.

Description

technical field [0001] The present application relates to the technical field of image processing, in particular to a face image processing model training method, device, electronic equipment and storage medium. Background technique [0002] The age editing of face images is mainly based on the input face image and the corresponding target age to automatically simulate the face of the age. Related algorithms have a large number of potential applications in many entertainment and medical beauty scenarios, such as predicting your future your appearance, guess what you looked like when you were a child, find missing children, compare certifications, etc. In addition, more face data sets can also be generated through this model to assist in the training of related intelligent face analysis tasks. At present, this field mainly adopts schemes such as generative confrontation network and variational self-encoding. The well-known methods include Learning face age progression propos...

AI Technical Summary

Problems solved by technology

[0003] The inventors found that the existing models trained based on the age editing algorithm of face images are not ideal, and the generated face images are not effective at certain ages

Although the face image age editing algorithm based on generative adversarial networks (GANs) has made great progress, this method cannot synthesize a small number of face age images

Images