Image generation system and method based on MIMO-GAN

Abstract

The invention discloses an image generation method and system based on MIMO-GAN. The system comprises a mode decomposer used for receiving a real image and decomposing the real image into a pluralityof sub-modes of the real image; a mode synthesizer which is used for receiving a plurality of sub-modes output by the MINO-GAN and fusing the sub-modes to generate a complete image; a multi-input multi-output adversarial generation network module and an MIMO-GAN module. According to the method, the real image is decomposed into a plurality of sub-modes through the mode decomposer; a generator anda plurality of discriminators in the MIMO-GAN are used to capture complete modal information of a real image in an adversarial game. By repeatedly training a plurality of generators and discriminatorsin the MIMO-GAN network, a generator set and a discriminator set which meet requirements are generated, and finally, a generated image is obtained through a mode synthesizer, so that the problem of mode collapse is avoided, the diversity of the generated image is increased, and the operation speed is increased.

Description

technical field [0001] The present invention relates to the field of image generation, in particular to an image generation system and method based on MIMO-GAN. Background technique [0002] Generative Adversarial Network (GAN for short) is a machine learning model, which is a research hotspot in the field of artificial intelligence in recent years, and is considered to be one of the most promising methods of unsupervised learning. The model was originally proposed by Goodfellow et al. in 2014, inspired by the two-person zero-sum game in game theory. The model consists of a generation network and a judgment network. Through these two networks, they constantly confront each other and adjust parameters to finally generate samples that are the same as the real data distribution. [0003] Although GAN has achieved great success in image generation, video detection, etc., it has quickly attracted the attention of academia and industry; but at the same time, the model is also kno...

AI Technical Summary

Problems solved by technology

However, the images generated by the model under the traditional GAN ​​framework of this technical solution are often concentrated in a certain sub-modality, which lacks diversity, and it is easy to cause the mode collapse of the GAN model, and the phenomenon of gradient disappearance will occur, which greatly hinders the use of GAN. Techniques for model-generated images

Images