【机器学习】GANs网络在图像和视频技术中的应用前景

# 导入必要的库
import tensorflow as tf
from tensorflow.keras.layers import Conv2D, Input
from tensorflow.keras.models import Model

# 定义生成器网络
def generator():
    # 输入层
    inputs = Input(shape=(None, None, 3))
    
    # 特征提取层
    x = Conv2D(64, 9, padding='same', activation='relu')(inputs)
    x = Conv2D(64, 3, padding='same', activation='relu')(x)
    
    # 残差块
    for _ in range(16):
        x = residual_block(x)
    
    # 上采样层
    x = Conv2D(64, 3, padding='same', activation='relu')(x)
    x = Conv2D(256, 3, padding='same')(x)
    x = tf.nn.depth_to_space(x, 2)
    
    # 输出层
    outputs = Conv2D(3, 9, padding='same', activation='tanh')(x)
    
    # 创建模型
    return Model(inputs, outputs)

# 定义残差块
def residual_block(x):
    y = Conv2D(64, 3, padding='same', activation='relu')(x)
    y = Conv2D(64, 3, padding='same')(y)
    return tf.keras.layers.add([x, y])

# 创建生成器模型
gen_model = generator()
gen_model.summary()

import torch
import torch.nn as nn
import torch.optim as optim
import torchvision.transforms as transforms
from torchvision import datasets

# 定义生成器网络
class Generator(nn.Module):
    def __init__(self):
        super(Generator, self).__init__()
        self.model = nn.Sequential(
            nn.Linear(100, 256),
            nn.LeakyReLU(0.2, inplace=True),
            nn.Linear(256, 512),
            nn.LeakyReLU(0.2, inplace=True),
            nn.Linear(512, 1024),
            nn.LeakyReLU(0.2, inplace=True),
            nn.Linear(1024, 28*28),
            nn.Tanh()
        )

    def forward(self, z):
        img = self.model(z)
        return img.view(img.size(0), 1, 28, 28)

# 定义判别器网络
class Discriminator(nn.Module):
    def __init__(self):
        super(Discriminator, self).__init__()
        self.model = nn.Sequential(
            nn.Linear(28*28, 1024),
            nn.LeakyReLU(0.2, inplace=True),
            nn.Linear(1024, 512),
            nn.LeakyReLU(0.2, inplace=True),
            nn.Linear(512, 256),
            nn.LeakyReLU(0.2, inplace=True),
            nn.Linear(256, 1),
            nn.Sigmoid()
        )

    def forward(self, img):
        flattened_img = img.view(img.size(0), -1)
        validity = self.model(flattened_img)
        return validity

# 超参数
batch_size = 64
lr = 0.0002
b1 = 0.5
b2 = 0.999
n_epochs = 200

# 初始化网络
generator = Generator()
discriminator = Discriminator()
optimizer_G = optim.Adam(generator.parameters(), lr=lr, betas=(b1, b2))
optimizer_D = optim.Adam(discriminator.parameters(), lr=lr, betas=(b1, b2))
adversarial_loss = nn.BCELoss()

# 数据加载和预处理
transform = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize((0.5,), (0.5,))
])
dataloader = torch.utils.data.DataLoader(
    datasets.MNIST('../data', train=True, download=True, transform=transform),
    batch_size=batch_size, shuffle=True)

# 训练网络
for epoch in range(n_epochs):
    for i, (imgs, _) in enumerate(dataloader):

        # 真实数据
        real_imgs = imgs.view(imgs.size(0), -1)

        # 训练判别器
        optimizer_D.zero_grad()
        z = torch.randn(batch_size, 100)
        fake_imgs = generator(z)
        real_validity = discriminator(real_imgs)
        fake_validity = discriminator(fake_imgs)
        d_loss = adversarial_loss(real_validity, torch.ones_like(real_validity)) + \
                 adversarial_loss(fake_validity, torch.zeros_like(fake_validity))
        d_loss.backward()
        optimizer_D.step()

        # 训练生成器
        optimizer_G.zero_grad()
        z = torch.randn(batch_size, 100)
        gen_imgs = generator(z)
        validity = discriminator(gen_imgs)
        g_loss = adversarial_loss(validity, torch.ones_like(validity))
        g_loss.backward()
        optimizer_G.step()

        # 打印训练信息
        if i % 100 == 0:
            print("[Epoch %d/%d] [Batch %d/%d] [D loss: %f] [G loss: %f]" % (epoch, n_epochs, i, len(dataloader),
                                                                              d_loss.item(), g_loss.item()))