基于多尺度Scale-Unet的单样本图像翻译

Single-sample Image Translation Based on Multi-scale Scale-Unet

随着生成对抗网络(GAN)的发展,基于单样本的无监督图像到图像翻译(UI2I)取得了重大进展。然而,以前方法无法捕获图像中的复杂纹理并保留原始内容信息。为解决这个问题,提出了一种基于尺度可变U-Net结构(Scale—Unet)的新型单样本图像翻译结构SUGAN。所提出的SUGAN使用Scale—Unet作为生成器,利用多尺度结构和渐进方法不断改进网络结构,以从粗到细地学习图像特征。同时,提出了尺度像素损失scale-pixel来更好地约束保留原始内容信息,防止信息丢失。实验表明,与SinGAN、TuiGAN、TSIT、StyTR2等公共数据集Summer■Winter、Horse■Zebra上的方法相比,该方法生成图像的SIFID值平均降低了30%。所提方法可更好地保留图像内容信息,同时生成详细逼真的高质量图像。

Single-sample unsupervised image-to-image translation(UI2I)has made significant progress with the development of generative adversarial networks(GANs).However,previous methods cannot capture complex textures in images and preserve original content information.We propose a novel one-shot image translation structure SUGAN based on a scale-variable U-Net structure(Scale—Unet).The proposed SUGAN uses Scale—Unet as a generator to continuously improve the network structure using multi-scale structures and progressive methods to learn image features from coarse to fine.Meanwhile,we propose the scale-pixel loss to better constrain the preservation of original content information and prevent information loss.Experiments show that compared with SinGAN,TuiGAN,TSIT,StyTR2 and another methods on public datasets Summer■Winter,Horse■Zebra,the SIFID value of the generated image is reduced by 30%.The proposed method can better preserve the content information of the image while generating detailed and realistic high-quality images.