pytorch 实现图片风格迁移

这几天把图片迁移的代码运行出来，感觉很开心！😘
之前在github上找了很多关于图片风格迁移的代码，但都没有运行出来，有可能是我的电脑不支持GPU加速。后来买了本书《python深度学习基于pytorch》，书上有相关代码的介绍。市面上关于pytorch深度学习的书籍相对较少，这本是我在豆瓣上看到利用pytorch进行深度学习评分较高的一本，兼顾了CPU和GPU。先上图让大家看看效果：
原代码链接


上面一张图的图片内容损失有点大，几乎看不出原来图形的轮廓；下面一张还行，就是天空那部分有点问题，可能多运行几次效果会好点。运行代码，差不多要半个小时，这是第一次我感觉到CPU的局限。下面直接上代码：
代码：

from __future__ import print_function

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim

from PIL import Image
import matplotlib.pyplot as plt

import torchvision.transforms as transforms
import torchvision.models as models

import copy
# 有GPU就利用GPU没有就利用CPU
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# 指定输出图像大小
imsize = 512 if torch.cuda.is_available() else 450  # 标准：450 560 / 200 300
imsize_w = 600

# 对图像进行预处理
loader = transforms.Compose([
    transforms.Resize((imsize, imsize_w)),
    transforms.ToTensor()
])

def image_loader(image_name):
    image = Image.open(image_name)
    # 增加一个维度，其值为1
    # 这是为了满足神经网络对输入图像的形状要求
    image = loader(image).unsqueeze(0)
    return image.to(device, torch.float)

style_img = image_loader('pic_1.jpg')
content_img = image_loader('pic_2.jpg')

print('style size', style_img.size())
print('content size', content_img.size())
assert style_img.size() == content_img.size()

输出图片尺寸：
(PS:上面pic_1和pic_2根据自己喜好选择相应图片，pic_1为风格图片，pic_2为要转换的图片)

style size torch.Size([1, 3, 450, 600])
content size torch.Size([1, 3, 450, 600])

代码：

unloader = transforms.ToPILImage()

plt.ion()

def imshow(tensor, title=None):
    image = tensor.cpu().clone()  # 为避免因image修改影响tensor的值
    image = image.squeeze(0)  # 去掉批量这个维度
    image = unloader(image)
    plt.imshow(image)
    if title is not None:
        plt.title(title)
    plt.pause(0.001)  # 暂停一下让图片能更新
    
plt.figure()
imshow(style_img, title='Style Image')

plt.figure()
imshow(content_img, title='Content Image')

输出图片：


代码：

class ContentLoss(nn.Module):
    
    def __init__(self, target):
        super(ContentLoss, self).__init__()
        # 必须要用detach来分离出target,这时候target不再是一个Variable
        # 这是为了动态计算梯度，否则forward会出错，不能向前传播
        self.target = target.detach()
        
    def forward(self, input):
        self.loss = F.mse_loss(input, self.target)
        return input

def gram_matrix(input):
    a, b, c, d = input.size()
    features = input.view(a * b, c * d)
    G = torch.mm(features, features.t()) # 计算内积
    
    # 对拉格姆矩阵标准化
    # 通过对其处理以特征图像总数
    return G.div(a * b * c * d)

class StyleLoss(nn.Module):
    def __init__(self, target_feature):
        super(StyleLoss, self).__init__()
        self.target = gram_matrix(target_feature).detach()
        
    def forward(self, input):
        G = gram_matrix(input)
        self.loss = F.mse_loss(G, self.target)
        return input

cnn = models.vgg19(pretrained=True).features.to(device).eval()
cnn

输出：

Sequential(
  (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (1): ReLU(inplace=True)
  (2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (3): ReLU(inplace=True)
  (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  (5): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (6): ReLU(inplace=True)
  (7): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (8): ReLU(inplace=True)
  (9): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  (10): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (11): ReLU(inplace=True)
  (12): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (13): ReLU(inplace=True)
  (14): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (15): ReLU(inplace=True)
  (16): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (17): ReLU(inplace=True)
  (18): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  (19): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (20): ReLU(inplace=True)
  (21): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (22): ReLU(inplace=True)
  (23): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (24): ReLU(inplace=True)
  (25): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (26): ReLU(inplace=True)
  (27): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  (28): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (29): ReLU(inplace=True)
  (30): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (31): ReLU(inplace=True)
  (32): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (33): ReLU(inplace=True)
  (34): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (35): ReLU(inplace=True)
  (36): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
)

代码：

cnn_normalization_mean = torch.tensor([0.485, 0.456, 0.406]).to(device)
cnn_normalization_std = torch.tensor([0.229, 0.224, 0.225]).to(device)

class Normalization(nn.Module):
    def __init__(self, mean, std):
        super(Normalization, self).__init__()
        self.mean = mean.clone().detach().view(-1, 1, 1)
        self.std = std.clone().detach().view(-1, 1, 1)
        
    def forward(self, img):
        return (img - self.mean) / self.std

# 为计算内容损失和风格损失，指定使用的内卷层
content_layers_default = ['conv_4']
style_layers_default = ['conv_1', 'conv_2', 'conv_3', 'conv_4', 'conv_5']

def get_style_model_and_losses(cnn, normalization_mean, normalization_std,
                              style_img, content_img,
                              content_layers=content_layers_default,
                              style_layers=style_layers_default):
    cnn = copy.deepcopy(cnn)
    
    # 标准化模型
    normalization = Normalization(normalization_mean, normalization_std).to(device)
    
    # 初始化损失值
    content_losses = []
    style_losses = []
    
    # 使用sequential方法构建模型
    model = nn.Sequential(normalization)
    
    i = 0 # 每次迭代增加1
    for layer in cnn.children():
        if isinstance(layer, nn.Conv2d):
            i += 1
            name = 'conv_{}'.format(i)
        elif isinstance(layer, nn.ReLU):
            name = 'relu_{}'.format(i)
            layer = nn.ReLU(inplace=False)
        elif isinstance(layer, nn.MaxPool2d):
            name = 'pool_{}'.format(i)
        elif isinstance(layer, nn.BachNorm2d):
            name = 'bn_{}'.format(i)
        else:
            raise RuntimeError('Unrecognized layer:{}'.format(layer.__class__.__name__))
        
        model.add_module(name, layer)
          
        if name in content_layers:
            # 累加内容损失
            target = model(content_img).detach()
            content_loss = ContentLoss(target)
            model.add_module("content_loss_{}".format(i), content_loss)
            content_losses.append(content_loss)
            
        if name in style_layers:
            # 累加风格损失
            target_feature = model(style_img).detach()
            style_loss = StyleLoss(target_feature)
            model.add_module("style_loss{}".format(i), style_loss)
            style_losses.append(style_loss)
            
    # 我们需要对在内容损失和风格损失之后的层进行修建
    for i in range(len(model) - 1, -1, -1):
        if isinstance(model[i], ContentLoss) or isinstance(model[i], StyleLoss):
            break
    model = model[:(i + 1)]
    
    return model, style_losses, content_losses

# 复制content_img
input_img = content_img.clone()
plt.figure()
# 我这里去掉了刻度和坐标轴
plt.axis('off')
plt.xticks([])
plt.yticks([])

imshow(input_img, title='Input Image')

输出：

代码：

def get_input_optimizer(input_img):
    optimizer = optim.LBFGS([input_img.requires_grad_()])
    return optimizer

def run_style_transfer(cnn, normalization_mean, normalization_std,
                      content_img, style_img, input_img, num_steps=600,
                      style_weight=1000000, content_weight=1):
    """返回风格"""
    print('Building the style transfer model..')
    model, style_losses, content_losses = get_style_model_and_losses(cnn,
        normalization_mean, normalization_std, style_img, content_img)
    optimizer = get_input_optimizer(input_img)
    
    print('Optimizing..')
    run = [0]
    while run[0] <= num_steps:
        
        def closure():
            input_img.data.clamp_(0, 1)
            
            optimizer.zero_grad()
            model(input_img)
            style_score = 0
            content_score = 0
            
            for sl in style_losses:
                style_score += sl.loss
            for cl in content_losses:
                content_score += cl.loss
            
            style_score *= style_weight
            content_score *= content_weight
            
            loss = style_score + content_score
            loss.backward()
            
            run[0] += 1
            if run[0] % 50 == 0:
                print("run {}:".format(run))
                print('Style Loss: {:4f} Content Loss: {:4f}'.format(
                style_score.item(), content_score.item()))
                print()
            return style_score + content_score
        optimizer.step(closure)
        
    input_img.data.clamp_(0, 1)
    
    return input_img

output = run_style_transfer(cnn, cnn_normalization_mean, cnn_normalization_std,
                           content_img, style_img, input_img)
plt.figure()
plt.xticks([])
plt.yticks([])
plt.axis('off')
imshow(output, title='Output Image')

plt.ioff()
plt.show()

输出结果：

Building the style transfer model..
Optimizing..
run [50]:
Style Loss: 71.486160 Content Loss: 13.809214

run [100]:
Style Loss: 40.491692 Content Loss: 14.122237

run [150]:
Style Loss: 25.580254 Content Loss: 13.938315

run [200]:
Style Loss: 13.806768 Content Loss: 14.290324

run [250]:
Style Loss: 6.078248 Content Loss: 14.245186

run [300]:
Style Loss: 3.660152 Content Loss: 13.365716

run [350]:
Style Loss: 3.001604 Content Loss: 12.663176

run [400]:
Style Loss: 2.334168 Content Loss: 12.076923

run [450]:
Style Loss: 2.023839 Content Loss: 11.745106

run [500]:
Style Loss: 1.812191 Content Loss: 11.482430

run [550]:
Style Loss: 1.663573 Content Loss: 11.269305

run [600]:
Style Loss: 1.579811 Content Loss: 11.157408

完结撒花！
本人小白一枚，写博客一方面是加深自己的印象，另一方面可以和大家交流学习，若有写的不对的地方，还请大家批评指正，以免误导他人学习！😁