CycleGan风格迁移

最新推荐文章于 2024-07-02 16:27:58 发布
最新推荐文章于 2024-07-02 16:27:58 发布
阅读量3.9k 收藏 14
点赞数 3
分类专栏： 深度学习
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/bblingbbling/article/details/107253930
版权

CycleGan风格迁移

简介

CycleGan全称： Cycle-Consistent Generative Adversarial Networks，循环一致性生成对抗网络，是一种Unpaired Image-to-Image Translation的深度学习网络。

效果展示

在这里插入图片描述

网络结构

生成网络 G：X → Y
生成网络 F：Y → X
判别网络 DX：判别输入图像是否为X
判别网络 DY：判别输入图像是否为Y

前向cycle-consistency损失：x → G(x) → F(G(x)) ≈ \approx ≈ x
后向cycle-consistency损失：y → F(y) → G(F(y)) ≈ \approx ≈ y
在这里插入图片描述
判别网络代码

self.d_A = self.build_discriminator()
self.d_B = self.build_discriminator()

img = Input(shape=self.img_shape)
# 64,64,64
d1 = conv2d(img, 64, normalization=False)
# 32,32,128
d2 = conv2d(d1, 128)
# 16,16,256
d3 = conv2d(d2, 256)
# 8,8,512
d4 = conv2d(d3, 512)
# 8,8,1
validity = Conv2D(1, kernel_size=3, strides=1, padding='same')(d4)
return Model(img, validity)

生成网络代码

self.g_AB = self.build_generator()
self.g_BA = self.build_generator()

img_A = Input(shape=self.img_shape)
img_B = Input(shape=self.img_shape)

fake_B = self.g_AB(img_A)
fake_A = self.g_BA(img_B)

reconstr_A = self.g_BA(fake_B)
reconstr_B = self.g_AB(fake_A)

img_A_id = self.g_BA(img_A)
img_B_id = self.g_AB(img_B)

model = get_resnet(self.img_rows,self.img_cols,self.channels)

Loss函数

1.优化目标
在这里插入图片描述
2.总损失函数
其中 λ \lambda λ 控制G和F的相对重要性。
在这里插入图片描述
3.对抗损失 LGAN
X和Y共享生成器和判别器，使生成的图片更接近于目标图片。
在这里插入图片描述
4.循环一致性损失 Lcyc
使模型重构的图像F(G(x)) 与输入的图像x十分接近。
在这里插入图片描述

训练过程

生成网络和判别网络交替训练。

生成模型训练

self.combined = Model(inputs=[img_A, img_B],
                     outputs=[valid_A, valid_B,
                               reconstr_A, reconstr_B,
                               img_A_id, img_B_id])
                                        
g_loss = self.combined.train_on_batch([imgs_A, imgs_B],
                                      [valid, valid,
                                      imgs_A, imgs_B,
                                      imgs_A, imgs_B])

判别模型训练

fake_B = self.g_AB.predict(imgs_A)
fake_A= self.g_BA.predict(imgs_B)

dA_loss_real = self.d_A.train_on_batch(imgs_A, valid)
dA_loss_fake = self.d_A.train_on_batch(fake_A, fake)
dA_loss = 0.5 * np.add(dA_loss_real, dA_loss_fake)

dB_loss_real = self.d_B.train_on_batch(imgs_B, valid)
dB_loss_fake = self.d_B.train_on_batch(fake_B, fake)
dB_loss = 0.5 * np.add(dB_loss_real, dB_loss_fake)

d_loss = 0.5 * np.add(dA_loss, dB_loss)

网络计算参数

判别网络：下采样

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_2 (InputLayer)         (None, 128, 128, 3)       0         
_________________________________________________________________
conv2d_6 (Conv2D)            (None, 64, 64, 64)        3136      
_________________________________________________________________
leaky_re_lu_5 (LeakyReLU)    (None, 64, 64, 64)        0         
_________________________________________________________________
conv2d_7 (Conv2D)            (None, 32, 32, 128)       131200    
_________________________________________________________________
leaky_re_lu_6 (LeakyReLU)    (None, 32, 32, 128)       0         
_________________________________________________________________
conv2d_8 (Conv2D)            (None, 16, 16, 256)       524544    
_________________________________________________________________
leaky_re_lu_7 (LeakyReLU)    (None, 16, 16, 256)       0         
_________________________________________________________________
conv2d_9 (Conv2D)            (None, 8, 8, 512)         2097664   
_________________________________________________________________
leaky_re_lu_8 (LeakyReLU)    (None, 8, 8, 512)         0         
_________________________________________________________________
conv2d_10 (Conv2D)           (None, 8, 8, 1)           4609      
=================================================================
Total params: 2,761,153
Trainable params: 2,761,153
Non-trainable params: 0

生成网络：下采样 ——上采样

__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_3 (InputLayer)            (None, 128, 128, 3)  0                                            
__________________________________________________________________________________________________
zero_padding2d_1 (ZeroPadding2D (None, 134, 134, 3)  0           input_3[0][0]                    
__________________________________________________________________________________________________
conv2d_11 (Conv2D)              (None, 128, 128, 64) 9472        zero_padding2d_1[0][0]           
__________________________________________________________________________________________________
activation_1 (Activation)       (None, 128, 128, 64) 0           conv2d_11[0][0]                  
__________________________________________________________________________________________________
zero_padding2d_2 (ZeroPadding2D (None, 130, 130, 64) 0           activation_1[0][0]               
__________________________________________________________________________________________________
conv2d_12 (Conv2D)              (None, 64, 64, 128)  73856       zero_padding2d_2[0][0]           
__________________________________________________________________________________________________
activation_2 (Activation)       (None, 64, 64, 128)  0           conv2d_12[0][0]                  
__________________________________________________________________________________________________
zero_padding2d_3 (ZeroPadding2D (None, 66, 66, 128)  0           activation_2[0][0]               
__________________________________________________________________________________________________
conv2d_13 (Conv2D)              (None, 32, 32, 256)  295168      zero_padding2d_3[0][0]           
__________________________________________________________________________________________________
activation_3 (Activation)       (None, 32, 32, 256)  0           conv2d_13[0][0]                  
__________________________________________________________________________________________________
zero_padding2d_4 (ZeroPadding2D (None, 34, 34, 256)  0           activation_3[0][0]               
__________________________________________________________________________________________________
res0_branch2a (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_4[0][0]           
__________________________________________________________________________________________________
activation_4 (Activation)       (None, 32, 32, 256)  0           res0_branch2a[0][0]              
__________________________________________________________________________________________________
zero_padding2d_5 (ZeroPadding2D (None, 34, 34, 256)  0           activation_4[0][0]               
__________________________________________________________________________________________________
res0_branch2c (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_5[0][0]           
__________________________________________________________________________________________________
add_1 (Add)                     (None, 32, 32, 256)  0           res0_branch2c[0][0]              
                                                                 activation_3[0][0]               
__________________________________________________________________________________________________
activation_5 (Activation)       (None, 32, 32, 256)  0           add_1[0][0]                      
__________________________________________________________________________________________________
zero_padding2d_6 (ZeroPadding2D (None, 34, 34, 256)  0           activation_5[0][0]               
__________________________________________________________________________________________________
res1_branch2a (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_6[0][0]           
__________________________________________________________________________________________________
activation_6 (Activation)       (None, 32, 32, 256)  0           res1_branch2a[0][0]              
__________________________________________________________________________________________________
zero_padding2d_7 (ZeroPadding2D (None, 34, 34, 256)  0           activation_6[0][0]               
__________________________________________________________________________________________________
res1_branch2c (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_7[0][0]           
__________________________________________________________________________________________________
add_2 (Add)                     (None, 32, 32, 256)  0           res1_branch2c[0][0]              
                                                                 activation_5[0][0]               
__________________________________________________________________________________________________
activation_7 (Activation)       (None, 32, 32, 256)  0           add_2[0][0]                      
__________________________________________________________________________________________________
zero_padding2d_8 (ZeroPadding2D (None, 34, 34, 256)  0           activation_7[0][0]               
__________________________________________________________________________________________________
res2_branch2a (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_8[0][0]           
__________________________________________________________________________________________________
activation_8 (Activation)       (None, 32, 32, 256)  0           res2_branch2a[0][0]              
__________________________________________________________________________________________________
zero_padding2d_9 (ZeroPadding2D (None, 34, 34, 256)  0           activation_8[0][0]               
__________________________________________________________________________________________________
res2_branch2c (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_9[0][0]           
__________________________________________________________________________________________________
add_3 (Add)                     (None, 32, 32, 256)  0           res2_branch2c[0][0]              
                                                                 activation_7[0][0]               
__________________________________________________________________________________________________
activation_9 (Activation)       (None, 32, 32, 256)  0           add_3[0][0]                      
__________________________________________________________________________________________________
zero_padding2d_10 (ZeroPadding2 (None, 34, 34, 256)  0           activation_9[0][0]               
__________________________________________________________________________________________________
res3_branch2a (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_10[0][0]          
__________________________________________________________________________________________________
activation_10 (Activation)      (None, 32, 32, 256)  0           res3_branch2a[0][0]              
__________________________________________________________________________________________________
zero_padding2d_11 (ZeroPadding2 (None, 34, 34, 256)  0           activation_10[0][0]              
__________________________________________________________________________________________________
res3_branch2c (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_11[0][0]          
__________________________________________________________________________________________________
add_4 (Add)                     (None, 32, 32, 256)  0           res3_branch2c[0][0]              
                                                                 activation_9[0][0]               
__________________________________________________________________________________________________
activation_11 (Activation)      (None, 32, 32, 256)  0           add_4[0][0]                      
__________________________________________________________________________________________________
zero_padding2d_12 (ZeroPadding2 (None, 34, 34, 256)  0           activation_11[0][0]              
__________________________________________________________________________________________________
res4_branch2a (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_12[0][0]          
__________________________________________________________________________________________________
activation_12 (Activation)      (None, 32, 32, 256)  0           res4_branch2a[0][0]              
__________________________________________________________________________________________________
zero_padding2d_13 (ZeroPadding2 (None, 34, 34, 256)  0           activation_12[0][0]              
__________________________________________________________________________________________________
res4_branch2c (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_13[0][0]          
__________________________________________________________________________________________________
add_5 (Add)                     (None, 32, 32, 256)  0           res4_branch2c[0][0]              
                                                                 activation_11[0][0]              
__________________________________________________________________________________________________
activation_13 (Activation)      (None, 32, 32, 256)  0           add_5[0][0]                      
__________________________________________________________________________________________________
zero_padding2d_14 (ZeroPadding2 (None, 34, 34, 256)  0           activation_13[0][0]              
__________________________________________________________________________________________________
res5_branch2a (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_14[0][0]          
__________________________________________________________________________________________________
activation_14 (Activation)      (None, 32, 32, 256)  0           res5_branch2a[0][0]              
__________________________________________________________________________________________________
zero_padding2d_15 (ZeroPadding2 (None, 34, 34, 256)  0           activation_14[0][0]              
__________________________________________________________________________________________________
res5_branch2c (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_15[0][0]          
__________________________________________________________________________________________________
add_6 (Add)                     (None, 32, 32, 256)  0           res5_branch2c[0][0]              
                                                                 activation_13[0][0]              
__________________________________________________________________________________________________
activation_15 (Activation)      (None, 32, 32, 256)  0           add_6[0][0]                      
__________________________________________________________________________________________________
zero_padding2d_16 (ZeroPadding2 (None, 34, 34, 256)  0           activation_15[0][0]              
__________________________________________________________________________________________________
res6_branch2a (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_16[0][0]          
__________________________________________________________________________________________________
activation_16 (Activation)      (None, 32, 32, 256)  0           res6_branch2a[0][0]              
__________________________________________________________________________________________________
zero_padding2d_17 (ZeroPadding2 (None, 34, 34, 256)  0           activation_16[0][0]              
__________________________________________________________________________________________________
res6_branch2c (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_17[0][0]          
__________________________________________________________________________________________________
add_7 (Add)                     (None, 32, 32, 256)  0           res6_branch2c[0][0]              
                                                                 activation_15[0][0]              
__________________________________________________________________________________________________
activation_17 (Activation)      (None, 32, 32, 256)  0           add_7[0][0]                      
__________________________________________________________________________________________________
zero_padding2d_18 (ZeroPadding2 (None, 34, 34, 256)  0           activation_17[0][0]              
__________________________________________________________________________________________________
res7_branch2a (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_18[0][0]          
__________________________________________________________________________________________________
activation_18 (Activation)      (None, 32, 32, 256)  0           res7_branch2a[0][0]              
__________________________________________________________________________________________________
zero_padding2d_19 (ZeroPadding2 (None, 34, 34, 256)  0           activation_18[0][0]              
__________________________________________________________________________________________________
res7_branch2c (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_19[0][0]          
__________________________________________________________________________________________________
add_8 (Add)                     (None, 32, 32, 256)  0           res7_branch2c[0][0]              
                                                                 activation_17[0][0]              
__________________________________________________________________________________________________
activation_19 (Activation)      (None, 32, 32, 256)  0           add_8[0][0]                      
__________________________________________________________________________________________________
zero_padding2d_20 (ZeroPadding2 (None, 34, 34, 256)  0           activation_19[0][0]              
__________________________________________________________________________________________________
res8_branch2a (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_20[0][0]          
__________________________________________________________________________________________________
activation_20 (Activation)      (None, 32, 32, 256)  0           res8_branch2a[0][0]              
__________________________________________________________________________________________________
zero_padding2d_21 (ZeroPadding2 (None, 34, 34, 256)  0           activation_20[0][0]              
__________________________________________________________________________________________________
res8_branch2c (Conv2D)          (None, 32, 32, 256)  590080      zero_padding2d_21[0][0]          
__________________________________________________________________________________________________
add_9 (Add)                     (None, 32, 32, 256)  0           res8_branch2c[0][0]              
                                                                 activation_19[0][0]              
__________________________________________________________________________________________________
activation_21 (Activation)      (None, 32, 32, 256)  0           add_9[0][0]                      
__________________________________________________________________________________________________
up_sampling2d_1 (UpSampling2D)  (None, 64, 64, 256)  0           activation_21[0][0]              
__________________________________________________________________________________________________
zero_padding2d_22 (ZeroPadding2 (None, 66, 66, 256)  0           up_sampling2d_1[0][0]            
__________________________________________________________________________________________________
conv2d_14 (Conv2D)              (None, 64, 64, 128)  295040      zero_padding2d_22[0][0]          
__________________________________________________________________________________________________
activation_22 (Activation)      (None, 64, 64, 128)  0           conv2d_14[0][0]                  
__________________________________________________________________________________________________
up_sampling2d_2 (UpSampling2D)  (None, 128, 128, 128 0           activation_22[0][0]              
__________________________________________________________________________________________________
zero_padding2d_23 (ZeroPadding2 (None, 130, 130, 128 0           up_sampling2d_2[0][0]            
__________________________________________________________________________________________________
conv2d_15 (Conv2D)              (None, 128, 128, 64) 73792       zero_padding2d_23[0][0]          
__________________________________________________________________________________________________
activation_23 (Activation)      (None, 128, 128, 64) 0           conv2d_15[0][0]                  
__________________________________________________________________________________________________
zero_padding2d_24 (ZeroPadding2 (None, 134, 134, 64) 0           activation_23[0][0]              
__________________________________________________________________________________________________
conv2d_16 (Conv2D)              (None, 128, 128, 3)  9411        zero_padding2d_24[0][0]          
__________________________________________________________________________________________________
activation_24 (Activation)      (None, 128, 128, 3)  0           conv2d_16[0][0]                  
==================================================================================================
Total params: 11,378,179
Trainable params: 11,378,179
Non-trainable params: 0
bblingbbling
关注
  • 3
    点赞
  • 踩
  • 14
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
专栏目录
[教学]cyclegan风格迁移+qt界面
分享知识
05-22 512
完整代码下载地址： cyclegan风格迁移+qt界面-深度学习文档类资源-CSDN下载本套代码的运行环境是python+pytorchCsdn上上传的代码都是完整的，大家找到对应的博客更多下载资源、学习资料请访问CSDN下载频道.https://download.csdn.net/download/qq_34904125/85436573 演示效果： 代码属于简单易上手代码，环境安装好即可直接使用。 整个代码如下： 01demo是直接训练使用的， 02demo是用于qt界面的.
基于CycleGAN的艺术风格迁移
04-22
我们使用了循环一致性生成对抗网络（ CycleConsistent Generative Adversarial Networks, CycleGAN）实现了将绘画中的艺术风格迁移到摄影照片中的效果。这种方法从图像数据集中学习整体风格，进行风格转换时只要将目标图片输入网络一次，不需要迭代的过程，因此速度较快。我们使用了一些自己制作的数据集训练了 CycleGAN 风格迁移模型，并分析了这种方法的优点和局限性。为了使风格转换后的图片保留原来的色彩分布，我们实现并分析对比了多种颜色匹配的方法。我们还利用了 Mask R-CNN 模型生成的掩膜图像来做图像运算而实现了局部风格转换和混合风格转换等效果。
风格迁移（CycleGAN）
如果想成为中心，那么就到中心去吧。
05-06 4171
文章目录概述CycleGANCycleGAN有两个Generator网络CycleGAN有两个Discriminator网络CycleGAN的网络结构应用参考 概述 Cycle-Consistent Adversarial Networks(CycleGAN)是一个新颖的方法，它能将图像从源域映射到目标域b。CycleGANs的一个很酷的功能是它不需要训练数据配对就能产生惊人的风格传递的结果。 在许多风格迁移的应用中，成对数据是训练所必需的。 配对 CycleGAN不需要配对数据输入到模型中 无需
CycleGAN代码解析(1)
X763321的博客
07-02 1002
cyclegan代码解析
cyclegan风格迁移
03-07
cyclegan图像转换压缩包，橘子苹果数据集及相关项目代码，可直接运行。
一文教会你风格迁移CycleGAN从入门到高阶再到最终成功魔改（附成功魔改代码）
记录学习的小白
07-11 1万+
🔥🔥本文的创新点部分，我将把我的整个项目过程所使用到的绝大多数trick以及创新部分进行开源。只有淋过雨的孩子才会懂得给别人打伞，所以我开源的目的很简单，希望可以帮助到有缘相遇的初学者快速了解并掌握该方向内容。有能力的同学可以进行二次改进创新。我的最终模型在定性和定量的评估中效果均有提升。后续代码将更新到GitHub上，如果对大家有帮助，希望可以得到您的免费star✨。本人万分感谢！！！ 本文的结构图均为本人绘制，如有需可以评论区留言❤ 五万字的文章创作不易，如有帮助劳烦免费点赞收藏一波，谢谢！
基于CycleGAN实现图像风格迁移
03-27
课程作业，内附源码
cycleGAN风格迁移数据集.rar
最新发布
07-17
资源1：CycleGAN_apple2orange。苹果橘子风格互换数据集。...Mindspore框架CycleGAN模型实现图像风格迁移算法实现：（https://blog.csdn.net/beauthy/article/details/140443871?spm=1001.2014.3001.5501）
风格迁移图像生成-基于Pytorch实现CycleGAN风格迁移图像生成算法-附项目源码+流程教程-优质项目实战.zip
05-28
风格迁移图像生成_基于Pytorch实现CycleGAN风格迁移图像生成算法_附项目源码+流程教程_优质项目实战
【毕设】基于CycleGAN的风格迁移【一】环境搭建及运行代码
记录学习的小白
04-24 1万+
源代码地址：CycleGAN源码 一、搭配环境 1.首先把代码包下载并解压到本地（我是直接放到桌面）。 2.打开Anaconda prompt，先切换路径到代码包解压的路径中。 3. 然后通过使用作者的环境包来搭建设定好的环境，输入： conda env create -f environment.yml 随后产生一个叫pytorch-CycleGAN-and-pix2pix的环境名称。我们可以通过输入： conda info --envs 来查看已有的虚拟环境名称。 ..
CycleGAN实现图像风格迁移的神作
jackzhang11的博客
05-11 7532
论文链接：http://openaccess.thecvf.com/content_ICCV_2017/papers/Zhu_Unpaired_Image-To-Image_Translation_ICCV_2017_paper.pdf 在CycleGAN出现之前，pix2pix网络在处理image-image translation问题上比较state-of-the-art。但是pix2pix需要利用成对（pair）的数据进行模型训练，如下图所示： 成对的数据在自然界中是非常稀有的，因此pix2pix对
基于CycleGAN的风格迁移---用debug对CycleGAN源码解读
m0_61985580的博客
05-11 586
CycleGAN源码解读
【GAN系列三】CycleGan---Unpaired Data做风格迁移
qq_44572656的博客
03-11 340
Cycle的实现原理
cycleGAN网络风格迁移，将黑夜转变成白天，低照度图像复原
weixin_44474671的博客
04-03 2053
cycleGAN网络风格迁移，将黑夜转变成白天，低照度图像复原 链接: [link](用CycleGAN网络复原低照度图像，风格转换，将黑夜颠倒成白昼，图像质量更优 - 小拍的文章 - 知乎https://zhuanlan.zhihu.com/p/362092765). ...
基于CycleGAN的山水风格画迁移
angula
09-05 7061
绘画是人类重要的一种艺术形式，其中中国的山水画源远流长，具有丰富的美学内涵，沉淀着中国人的情思。游山玩水的大陆文化意识，以山为德、水为性的内在修为意识，咫尺天涯的视错觉意识，一直成为山水画演绎的中轴主线。从山水画中，我们可以集中体味中国画的意境、格调、气韵和色调。随着互联网行业的发展，社交媒体逐渐兴起，人们经常在抖音、快手、微博等软件发布短视频或图片。发布图片前，人们一般要对视频或图片进行一些美化，最常见的就是添加一些滤镜。
【转】CycleGAN：图片风格，想换就换 | ICCV 2017论文解读
lanyuxuan100的博客
08-09 1109
转自：http://www.sohu.com/a/225520144_500659 源项目主页：https://junyanz.github.io/CycleGAN/ 本期推荐的论文笔记来自 PaperWeekly 社区用户@LUOHAO。本文提出的模型名为 CycleGAN，作者希望在不借助 paired example 情况下，来实现图片的风格转换。 关于作者：罗浩，浙江大学博士研究生，...
(五)从头开始构建风格迁移CycleGAN
寒冰屋的专栏
03-13 880
​ 在本系列文章中，我们将展示一个基于循环一致对抗网络 (CycleGAN)的移动图像到图像转换系统。我们将构建一个CycleGAN，它可以执行不成对的图像到图像的转换，并向您展示一些有趣但具有学术深度的例子。我们还将讨论如何将这种使用TensorFlow和Keras构建的训练有素的网络转换为TensorFlow Lite并用作移动设备上的应用程序。 ​
风格迁移CycleGAN开源项目代码运行步骤详细教程
qq_46231566的博客
01-20 6728
最近在学习Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks这篇论文，论文下载地址，想要复现一下文中的代码，过程中遇到了很多问题，因此记录下来。遇到其他问题欢迎在评论区留言，相互解答。
GAN | CycleGAN | 风格迁移
热门推荐
liongxiong的博客
07-01 1万+
GAN是解决风格迁移的深度学习方法，本文将做出简要介绍。一、GAN      GAN由生成网络G和对抗网络D组成，G用于接收一个噪声z，从而生成图片G(z)；D是一个判别网络，判别一张图片x是不是“真的”我们所需得到的目的图片或者是由G生成的（此时图片是“假的”），即D(x)。      生成网络G和对抗网络D在训练时候互为动态的博弈过程。      损失函数：                 x...
cyclegan风格迁移中fid数据范围
05-30
在CycleGAN风格迁移中，fid数据的范围是正实数，通常在0到无穷大之间。FID（Fréchet Inception Distance）是用来评估生成图片与真实图片之间差异的一种指标，它是通过计算生成图片特征与真实图片特征之间的距离来...
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值