gm_Ergou的博客

1.生成mask模型（只保留人脸区域）# -*- coding:utf-8 -*-import osimport cv2root_path='data2/'save_path='data2/cut/'files=os.listdir(root_path)for file in files: if file.find('face') !=-1: filesplit=file...

参考链接：原文：https://blog.youkuaiyun.com/Exploer_TRY/article/details/81841826代码详解：https://blog.youkuaiyun.com/greepex/article/details/86360726公式理解：https://blog.youkuaiyun.com/Teeyohuang/article/details/82729047下面是自己...

5.generator.py（生成器）import tensorflow as tfimport opsimport utilsclass Generator: #定义实例（Generator）的私有属性---外部无法访问，实例内可以传递 def __init__(self, name, is_training, ngf=64, norm='instance', image_...

参考链接：https://blog.youkuaiyun.com/z704630835/article/details/84135277https://github.com/dongheehand/DeblurGAN-tf数据集：代码实现：data_loader.pyimport tensorflow as tfimport osclass dataloader(): def __init__(self, args): self.channel

参考链接：https://blog.youkuaiyun.com/dcrmg/article/details/82892909https://blog.youkuaiyun.com/leviopku/article/details/83818997棋盘效应：https://blog.youkuaiyun.com/qq_34622844/article/details/88936690https://github.com/KupynOrest/DeblurGAN/issues/41代码实现：train.py

参考链接：https://blog.51cto.com/gloomyfish/2348169?source=drhhttps://blog.youkuaiyun.com/hiudawn/article/details/80775740https://blog.youkuaiyun.com/DLW__/article/details/99942598https://blog.youkuaiyun.com/u013250416/article/details/78254444数据集：代码实现1：dcgan--mni

GAN：生成对抗网络（Generative Adversarial Networks [4]）主要由生成器 (generator) 和判别器 (discriminator)组成。它的原理也比较清晰，generator 负责输入随机噪声z，输出一个图片 G(z) ，而真实样本x，判别器 D 则用尽全力希望把真实样本和虚假样本区分开来。而 G 则希望产生的 G(z) 以假乱真，欺骗判别器，让其判断不出来。从而有了这么一种对抗的关系。gan的缺点：1.不适合处理离散形式的数据 例如——文本2....

参考链接：https://blog.youkuaiyun.com/weixin_41923961/article/details/80686419https://blog.youkuaiyun.com/qq_27898631/article/details/87607020https://blog.youkuaiyun.com/Sakura55/article/details/81514828https://blog.youkuaiyun.com/yunyi4367/article/details/80489962tensorflow