Image attribute editing method and device for generative adversarial network, equipment and medium

Abstract

The embodiment of the invention provides an image attribute editing method and device for an adversarial generative network, equipment and a medium, and belongs to the technical field of image processing.The method specifically comprises the following steps: sampling a submerged space; labeling all the initially generated images; establishing a learning model; carrying out decoupling operation on all label values; obtaining a feature vector of each attribute library; obtaining a prediction value; calculating a loss function; updating the trunk model and the head model according to the loss function and the regularization constraint; forming an attribute editing model; obtaining a target latent code; and inputting the target latent code into the face generation model to obtain a target image. Through the scheme of the invention, the existing face generation model is sampled to obtain the initial generation image, the learning model is established and independently trained at the same time after data corresponding to the initial generation image is decoupled, the attribute editing model is obtained, and the target latent code is generated through the attribute editing model to edit the specific attribute, and the editability of image attributes is improved.

Description

technical field [0001] The present disclosure relates to the technical field of image processing, and in particular to an image attribute editing method, device, device and medium against a generative network. Background technique [0002] At present, the image generation technology based on machine learning technology is becoming more and more mature. In particular, the Generative Adversarial Networks (GAN for short) technology that appeared in 2014 has been able to generate very realistic and high-resolution images after years of development. The leading technologies are BigGAN and StyleGAN, etc. This type of technology is usually an unsupervised algorithm. Gaussian noise is sampled from a multidimensional normal distribution as the input of the GAN model. The GAN model obtains a very realistic image through a series of operations. But such unsupervised algorithms have a common shortcoming: lack of control, that is, it is difficult to specify a specific individual, and it...

AI Technical Summary

Problems solved by technology

[0004] In view of this, the embodiments of the present disclosure provide an image attribute editing method, device, device and medium against generative networks, at least partially solving the problem of poor editability of image attributes in the prior art

Images