Generative adversarial network training method, generative adversarial network and face image translation method and device

Abstract

The invention discloses a generative adversarial network training method, a generative adversarial network and a face image translation method and device, and relates to the technical field of computer vision, augmented reality and deep learning. The generative adversarial network comprises N processing layers, the N processing layers comprise M convolution layers, and the generative adversarial network training method comprises the steps of adding noise to a training sample to obtain first feature data; inputting the first feature data into the generative adversarial network for training; wherein in the training process, noise is added to the feature data output by each convolution layer, and then the feature data is input to the next processing layer. According to the method, various inputs in the real scene are simulated by adding noise to the training sample and the feature data output by each convolution layer, so that the generative adversarial network has higher adaptability toa large amount of rich inputs in the real scene, and the stability, robustness and fault tolerance of the generative adversarial network can be improved.

Description

technical field [0001] This application relates to the field of artificial intelligence technology, specifically computer vision, augmented reality and deep learning technology, and in particular to a method for training a generative confrontation network, a generative confrontation network, and a face image translation method and device. Background technique [0002] With the development of artificial intelligence technology, the application of generative adversarial networks is becoming more and more widely. Through generative adversarial networks, corresponding target data can be generated according to the data input by users. For example, face image translation technology based on generative adversarial networks can be based on user input Generate a matching personalized avatar from the face image. However, due to the variety of data input by users, the generative adversarial network often performs better on the training set, but it is prone to collapse in the test set e...

AI Technical Summary

Problems solved by technology

However, due to the variety of data input by users, the generative adversarial network often performs better on the training set, but it is prone to collapse in the test set environment, resulting in large flaws in the generated target data or complete failure of the generation

Images