基于改进的生成对抗网络漫画风格迁移的图片生成

Image generation based on improved generation adversarial network style transfer

在VGG生成对抗网络的基础上,提出了一种改进的基于残差网络的生成对抗网络漫画风格迁移的图片生成技术,用于图片特征提取及图片生成,使具有漫画家卡通风格的特征迁移到现实图片中,和VGG生成对抗网络相比,在一定程度上缓解了梯度消失、过拟合等问题.实验表明,相较于VGG生成对抗网络,改进后的模型在图像特征提取及生成都表现了更好的性能.

In the process of cartoon production,it is very time-consuming and laborious to hand-draw these realistic scenes with cartoon style.The existing drawing software can only draw pictures with specific features,so it is of great practical significance to transfer the cartoonist s painting style to a realistic picture by using machine learning.Based on VGG generative adversarial network,an improved generative adversarial network cartoon style transfer technique is proposed,which is used for image feature extraction and image generation and makes the features of cartoonist cartoon style transferred to the real image..And compared with VGG generated against network,it alleviates the gradient disappearance and fitting to some extent.Experimental results show that the improved model has better performance in image feature extraction and generation compared with VGG.