基于无映射数据集的生成对抗网络图像生成算法研究

Research on Image Generation Algorithm of Generative Adversarial Network Based on Unmapped Data Set

生成对抗网络(GAN)与图像处理的结合一直受到机器视觉领域研究者的推崇,GAN凭借其强大的网络结构和特征学习能力,在图像生成和转换领域尤为出色。为了研究基于生成对抗网络的图像生成算法,对CycleGAN算法进行了重点研究,并结合Pix2Pix算法进行比较分析,得出实验结果。根据人工智能领域的发展提出了可行算法的改进方向,并对GAN在图像生成领域的发展趋势提出了预测。

In recent years,the combination of Generative Adversarial Networks(GAN)and image processing had been highly respected by researchers in the field of machine vision.With its powerful network structure and feature learning capabilities,GAN was particularly excellent in the field of image generation and conversion.In order to study the image generation algorithm based on the Generative Adversarial Network,this paper focused on the CycleGAN algorithm,and combined the Pix2Pix algorithm for comparative analysis,and obtained the experimental results.According to the development of artificial intelligence field,the improvement direction of feasible algorithm was proposed,and the development trend of GAN in the field of image generation was predicted.