基础GAN的原理还不懂的,先看:生成式对抗神经网络(GAN)原理给你讲的明明白白
# 数据归一化
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(0.5, 0.5)
])
# 加载内置数据
train_ds = torchvision.datasets.MNIST('data', # 当前目录下的data文件夹
train=True, # train数据
transform=transform,
download=True)
dataloader = torch.utils.data.DataLoader(train_ds, batch_size=64, shuffle=True)
# 定义生成器
# 输入是长度为100的噪声(正态分布随机数)
class Generator(nn.Module):
def __init__(self):
super(Generator, self).__init__()
self.gen = nn.Sequential(nn.Linear(100, 256),
nn.ReLU(),
nn.Linear(256, 512),
nn.ReLU(),
nn.Linear(512, 28 * 28),
nn.Tanh()
)
# 定义前向传播 x表示长度为100的noise输入
def forward(self, x):
img = self.gen(x)
img = img.view(-1, 28, 28)
return img
输入是长度为100的噪声,由于mnist数据集的手写数字图片是1*28*28,所以生成28*28的图片
linear 1: 100---->256 2:256----->521 3:512----->28*28
# 定义判别器
# 输入为(1,28,28)的图片 输出为二分类的概率值,使用sigmoid激活
# BCEloss 计算交叉熵损失
# 判别器中推荐使用LeakyReLU
class Discriminator(nn.Module):
def __init__(self):
super(Discriminator,self).__init__()
self.disc = nn.Sequential(nn.Linear(28*28, 512),
nn.LeakyReLU(),
nn.Linear(512, 256),
nn.LeakyReLU(),
nn.Linear(256, 1),
nn.Sigmoid()
)
def forward(self, x):
x = x.view(-1, 28*28)
x = self.disc(x)
return x
# 初始化模型
device = 'cuda' if torch.cuda.is_available() else 'cpu'
gen = Generator().to(device)
dis = Discriminator().to(device)
# 开始训练
D_loss = []
G_loss = []
# 训练循环
for epoch in range(50):
d_epoch_loss = 0
g_epoch_loss = 0
batch_count = len(dataloader)
# 对全部的数据集做一次迭代
for step, (img, _) in enumerate(dataloader):
img = img.to(device)
size = img.size(0)
random_noise = torch.randn(size, 100, device=device)
d_optim.zero_grad() # 将上述步骤的梯度归零
real_output = dis(img) # 对判别器输入真实的图片,real_output是对真实图片的预测结果
d_real_loss = loss_function(real_output,
torch.ones_like(real_output)
)
d_real_loss.backward() #求解梯度
# 得到判别器在生成图像上的损失
gen_img = gen(random_noise)
fake_output = dis(gen_img.detach()) #
d_fake_loss = loss_function(fake_output,
torch.zeros_like(fake_output))
d_fake_loss.backward()
d_loss = d_real_loss + d_fake_loss
d_optim.step() # 优化
# 得到生成器的损失
g_optim.zero_grad()
fake_output = dis(gen_img)
g_loss = loss_function(fake_output,
torch.ones_like(fake_output))
g_loss.backward()
g_optim.step()
鉴别器和生成器的损失各是什么?
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import numpy as np
import matplotlib.pyplot as plt
import torchvision
from torchvision import transforms
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(0.5, 0.5)
])
# 加载内置数据 做生成只需要图片就行,不需要标签 也不需要测试数据集
train_ds = torchvision.datasets.MNIST('data', # 当前目录下的data文件夹
train=True, # train数据
transform=transform,
download=True)
dataloader = torch.utils.data.DataLoader(train_ds, batch_size=64, shuffle=True)
# 定义生成器
# 输入是长度为100的噪声
class Generator(nn.Module):
def __init__(self):
super(Generator, self).__init__()
self.gen = nn.Sequential(nn.Linear(100, 256), # 输入长度为100
nn.ReLU(),
nn.Linear(256, 512),
nn.ReLU(),
nn.Linear(512, 28 * 28),
nn.Tanh()
)
def forward(self, x): # 定义前向传播 x表示长度为100的noise输入
img = self.gen(x)
img = img.view(-1, 28, 28)
return img
# 定义判别器
# 输入为(1,28,28)的图片 输出为二分类的概率值,使用sigmoid激活
# BCEloss 计算交叉熵损失
# 判别器中推荐使用LeakyReLU
class Discriminator(nn.Module):
def __init__(self):
super(Discriminator,self).__init__()
self.disc = nn.Sequential(nn.Linear(28*28, 512),
nn.LeakyReLU(),
nn.Linear(512, 256),
nn.LeakyReLU(),
nn.Linear(256, 1),
nn.Sigmoid()
)
def forward(self, x):
x = x.view(-1, 28*28)
x = self.disc(x)
return x
# 初始化模型
device = 'cuda' if torch.cuda.is_available() else 'cpu'
gen = Generator().to(device)
dis = Discriminator().to(device)
# 定义优化器
d_optim = torch.optim.Adam(dis.parameters(), lr=0.0001)
g_optim = torch.optim.Adam(gen.parameters(), lr=0.0001)
# 损失计算函数
loss_function = torch.nn.BCELoss()
# 绘图函数
def gen_img_plot(model, test_input):
prediction = np.squeeze(model(test_input).detach().cpu().numpy())
fig = plt.figure(figsize=(4,4))
for i in range(16):
plt.subplot(4, 4, i+1)
plt.imshow((prediction[i] + 1)/2) # 由于tanh是在-1 1 之间 要恢复道0 1 之间
plt.axis("off")
plt.show()
test_input =torch.randn(16, 100, device=device)
# 开始训练
D_loss = []
G_loss = []
# 训练循环
for epoch in range(50):
d_epoch_loss = 0
g_epoch_loss = 0
batch_count = len(dataloader.dataset)
# 对全部的数据集做一次迭代
for step, (img, _) in enumerate(dataloader):
img = img.to(device) # 上传到设备上
size = img.size(0) # 返回img的第一维的大小
random_noise = torch.randn(size, 100, device=device)
d_optim.zero_grad() # 将上述步骤的梯度归零
real_output = dis(img) # 对判别器输入真实的图片,real_output是对真实图片的预测结果
d_real_loss = loss_function(real_output,
torch.ones_like(real_output)
)
d_real_loss.backward() #求解梯度
# 得到判别器在生成图像上的损失
gen_img = gen(random_noise)
fake_output = dis(gen_img.detach()) # 判别器输入生成的图片,对生成图片的预测结果
d_fake_loss = loss_function(fake_output,
torch.zeros_like(fake_output))
d_fake_loss.backward()
d_loss = d_real_loss + d_fake_loss
d_optim.step() # 优化
# 得到生成器的损失
g_optim.zero_grad()
fake_output = dis(gen_img)
g_loss = loss_function(fake_output,
torch.ones_like(fake_output))
g_loss.backward()
g_optim.step()
with torch.no_grad():
d_epoch_loss += d_loss
g_epoch_loss += g_loss
with torch.no_grad():
d_epoch_loss /= batch_count
g_epoch_loss /= batch_count
D_loss.append(d_epoch_loss)
G_loss.append(g_epoch_loss)
print('Epoch:', epoch)
gen_img_plot(gen, test_input)