Image generation and identification module and method based on BlockGAN

Abstract

The invention discloses an image generation and recognition module and method based on a BlockGAN. The module comprises an image segmentation module; a mode decomposition module; a plurality of blocks, whereinthe plurality of blocks forming a BlockGAN (block GAN), each block including a generator and a plurality of discriminators connected to the generator; a first fusion module; second fusion module. The method mainly solves the problem of mode diversity in the complex image generation process, the problem of accuracy in the recognition process and the problem of operation efficiency. Each block is composed of one generator and a plurality of discriminators and is responsible for generating and identifying one type of object, and the plurality of discriminators can discriminate a plurality of modes of the same object so as to improve the accuracy of image identification. A plurality of blocks can be processed in parallel, a single object generated by the generator can be fused into acomplex image, and the discrimination result of the discriminator is fused to form accurate identification of the complex image.

Description

technical field [0001] The invention relates to the technical field of image generation and recognition, in particular to an image generation and recognition module and method based on BlockGAN. Background technique [0002] GAN (Generative Adversarial Nets, Generative Adversarial Nets) draws on the idea of ​​​​zero-sum games. Through the game between the generator and the discriminator, the performance of the generator and the discriminator is continuously improved during the confrontation process, and finally the generator can generate fake images. , and the discriminator cannot distinguish the authenticity of the generated image. The GAN method is simple and practical, and is widely used in the field of image generation. In recent years, research has found that it also performs well in the field of image recognition. [0003] The generation and recognition of simple images has achieved remarkable results, but the generation and recognition of complex images still has big...

AI Technical Summary

Problems solved by technology

[0004] In view of the above problems, the present invention provides an image generation and recognition module and method based on BlockGAN, which mainly solves the problem of pattern diversity in the process of complex image generation, the problem of accuracy in the recognition process, and the problem of operating efficiency

[0005] On the basis of GAN technology, the present invention proposes a new GAN architecture BlockGAN (block GAN), which is used for the generation and recognition of complex images, and effectively solves the problem of pattern diversity and recognition process in the process of complex image generation. The problem of accuracy and model operation efficiency, the solution is as follows:

Images