动手学深度学习之样式迁移

样式迁移

• 将样式图片中的样式迁移到内容图片上，得到合成图片
  

基于CNN的样式迁移

• 输入是一张内容的图片和一张样式的图片，目标是将内容的图片应用上样式图片的样式。
• CNN对于图片的出入，每一层都是一些特征，我们训练一个图片使得图片和内容图片，放进同样一个CNN的时候，在某一层它的输出可以匹配上另一个的内容。同样另一个样式图片在放入一个CNN后，在某一层在样式上可以匹配上。
  

代码实现

# 阅读内容和样式图像
%matplotlib inline
import torch
import torchvision
from torch import nn
from d2l import torch as d2l

d2l.set_figsize()
content_img = d2l.Image.open('/Users/tiger/Desktop/study/机器学习/李沐深度学习/d2l-zh/pytorch/img/rainier.jpg')
d2l.plt.imshow(content_img);

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-NOzsF1b0-1630551548233)(output_2_0.svg)]

style_img = d2l.Image.open('/Users/tiger/Desktop/study/机器学习/李沐深度学习/d2l-zh/pytorch/img/autumn-oak.jpg')
d2l.plt.imshow(style_img);

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-1dK4vI0F-1630551548235)(output_3_0.svg)]

# 预处理和后处理
rgb_mean = torch.tensor([0.485, 0.456, 0.406])
rgb_std = torch.tensor([0.229, 0.224, 0.225])

# 将图片变为Tensor
def preprocess(img, image_shape):
    transforms = torchvision.transforms.Compose([
        torchvision.transforms.Resize(image_shape),
        torchvision.transforms.ToTensor(),
        torchvision.transforms.Normalize(mean=rgb_mean, std=rgb_std)])
    return transforms(img).unsqueeze(0)

# 将Tensor变为图片
def postprocess(img):
    img = img[0].to(rgb_std.device)
    img = torch.clamp(img.permute(1, 2, 0) * rgb_std + rgb_mean, 0, 1)
    return torchvision.transforms.ToPILImage()(img.permute(2, 0, 1))

# 抽取图像特征
pretrained_net = torchvision.models.vgg19(pretrained=True)

style_layers, content_layers = [0, 5, 10, 19, 28], [25]  # 越靠近数据对内容的还原越好，越往下对细节还原的越好

net = nn.Sequential(*[
    pretrained_net.features[i]
    for i in range(max(content_layers + style_layers) + 1)])  # 只要28层，28层之后的丢掉

# 给定X，样式层，内容层，然后来抽取特征
def extract_features(X, content_layers, style_layers):
    contents = []
    styles = []
    for i in range(len(net)):
        X = net[i](X)
        if i in style_layers:
            styles.append(X)
        if i in content_layers:
            contents.append(X)
    return contents, styles  # 返回内容和样式

# 拿到内容
def get_contents(image_shape, device):
    content_X = preprocess(content_img, image_shape).to(device)
    contents_Y, _ = extract_features(content_X, content_layers, style_layers)
    return content_X, contents_Y

# 拿到样式
def get_styles(image_shape, device):
    style_X = preprocess(style_img, image_shape).to(device)
    _, styles_Y = extract_features(style_X, content_layers, style_layers)
    return style_X, styles_Y

Downloading: "https://download.pytorch.org/models/vgg19-dcbb9e9d.pth" to /Users/tiger/.cache/torch/hub/checkpoints/vgg19-dcbb9e9d.pth
98.8%

# 定义损失函数
def content_loss(Y_hat, Y):
    return torch.square(Y_hat - Y.detach()).mean()  # 内容直接就是均方误差

def gram(X):
    num_channels, n = X.shape[1], X.numel() // X.shape[1]
    X = X.reshape((num_channels, n))
    return torch.matmul(X, X.T) / (num_channels * n)

def style_loss(Y_hat, gram_Y):
    return torch.square(gram(Y_hat) - gram_Y.detach()).mean()  # 比较gram矩阵的均方误差

def tv_loss(Y_hat):
    return 0.5 * (torch.abs(Y_hat[:, :, 1:, :] - Y_hat[:, :, :-1, :]).mean() +
                  torch.abs(Y_hat[:, :, :, 1:] - Y_hat[:, :, :, :-1]).mean())  # tv降噪的一个算法

# 风格转移的损失函数是内容损失、风格损失和总变化损失的加权和
content_weight, style_weight, tv_weight = 1, 1e3, 10

def compute_loss(X, contents_Y_hat, styles_Y_hat, contents_Y, styles_Y_gram):
    contents_l = [
        content_loss(Y_hat, Y) * content_weight
        for Y_hat, Y in zip(contents_Y_hat, contents_Y)]
    styles_l = [
        style_loss(Y_hat, Y) * style_weight
        for Y_hat, Y in zip(styles_Y_hat, styles_Y_gram)]
    tv_l = tv_loss(X) * tv_weight
    l = sum(10 * styles_l + contents_l + [tv_l])
    return contents_l, styles_l, tv_l, l  # 返回所有的损失

# 初始化合成图像
class SynthesizedImage(nn.Module):  # 告诉图片的shape，拿到weight
    def __init__(self, img_shape, **kwargs):
        super(SynthesizedImage, self).__init__(**kwargs)
        self.weight = nn.Parameter(torch.rand(*img_shape))

    def forward(self):
        return self.weight

# 调用合成图片
def get_inits(X, device, lr, styles_Y):
    gen_img = SynthesizedImage(X.shape).to(device)
    gen_img.weight.data.copy_(X.data)
    trainer = torch.optim.Adam(gen_img.parameters(), lr=lr)
    styles_Y_gram = [gram(Y) for Y in styles_Y]  # 算好
    return gen_img(), styles_Y_gram, trainer

# 训练
def train(X, contents_Y, styles_Y, device, lr, num_epochs, lr_decay_epoch):
    X, styles_Y_gram, trainer = get_inits(X, device, lr, styles_Y)
    scheduler = torch.optim.lr_scheduler.StepLR(trainer, lr_decay_epoch, 0.8)
    animator = d2l.Animator(xlabel='epoch', ylabel='loss',
                            xlim=[10, num_epochs],
                            legend=['content', 'style',
                                    'TV'], ncols=2, figsize=(7, 2.5))
    for epoch in range(num_epochs):
        trainer.zero_grad()
        contents_Y_hat, styles_Y_hat = extract_features(
            X, content_layers, style_layers)
        contents_l, styles_l, tv_l, l = compute_loss(X, contents_Y_hat,
                                                     styles_Y_hat, contents_Y,
                                                     styles_Y_gram)
        l.backward()
        trainer.step()
        scheduler.step()
        if (epoch + 1) % 10 == 0:
            animator.axes[1].imshow(postprocess(X))
            animator.add(
                epoch + 1,
                [float(sum(contents_l)),
                 float(sum(styles_l)),
                 float(tv_l)])
    return X

# 训练模型
device, image_shape = d2l.try_gpu(), (300, 450)
net = net.to(device)
content_X, contents_Y = get_contents(image_shape, device)
_, styles_Y = get_styles(image_shape, device)
output = train(content_X, contents_Y, styles_Y, device, 0.3, 500, 50)