Segmentation loss-based generative adversarial network method

Abstract

The invention discloses a segmentation loss-based generative adversarial network method. The method comprises the following steps of 1, performing parameter initialization: setting a batch size m to be 100 and a hyper-parameter k to be 1, performing the parameter initialization by using an Xavier method, determining a maximum iterative frequency and a loss switching iterative frequency parameter T, and setting an iterative frequency epoch to be 0; 2, training discriminator parameters: setting i to be 1, wherein i is a cyclic variable; and 3, training generator parameters: in epoch=epoch+1, judging whether the epoch is greater than the maximum iterative frequency or not, if the epoch is smaller than the maximum iterative frequency, repeating the steps 2 and 3, and if the epoch is greater than the maximum iterative frequency, ending the training process. According to the method, a generator can adopt loss functions in different forms in different training stages, so that the deficiency of a GAN theory under a single loss form is made up for to a certain extent, and the network training is more stable; and by introducing feature-level loss between a real sample and a generated sample,features extracted by a discriminator are robuster.

Description

technical field [0001] The invention belongs to the technical field of deep learning neural network, in particular to a generation confrontation network method based on segmentation loss. Background technique [0002] Generative Adversarial Network (GAN for short) is an unsupervised deep learning framework proposed by Goodfellow in 2014. Drawing on the idea of ​​"game theory", two players are constructed: generator (generator) and discriminator (discriminator) , the former generates images through uniform noise or Gaussian random noise with input parameters of (0, 1), and the latter discriminates the input images to determine whether the input is an image from a data set or an image generated by a generator. The discriminator feeds back the result of the judgment to the generator, so that it can be optimized toward the distribution of real data. [0003] In recent years, generative adversarial networks have been widely used in image generation and semi-supervised learning. ...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the problems existing in the above-mentioned prior art, the present invention provides a method of generative adversarial network based on segmentation loss, which can avoid the phenomenon of training instability and mode collapse of conventional generative adversarial network under a single form of loss, thereby solving the problem of discrimination The problem of poor features extracted by the generator; this method can realize that the generator uses different forms of loss functions in different training periods, and makes the network training more stable by making the generator introduce feature-level losses between real samples and generated samples. The features extracted by the machine are more robust

Images