有百度推广的网站,免费企业网站系统源码,半年工作总结,做家装的网站神经风格迁移在之前的博客中已经用keras实现过了#xff0c;比较复杂#xff0c;keras版本。
这里用pytorch重新实现一次#xff0c;原理图如下#xff1a; from __future__ import division
from torchvision import models
from torchvision import transforms
from PIL…神经风格迁移在之前的博客中已经用keras实现过了比较复杂keras版本。
这里用pytorch重新实现一次原理图如下 from __future__ import division
from torchvision import models
from torchvision import transforms
from PIL import Image
import argparse
import torch
import torchvision
import torch.nn as nn
import numpy as npdevice torch.device(cuda if torch.cuda.is_available() else cpu) 加载图像
def load_image(image_path, transformNone, max_sizeNone, shapeNone):Load an image and convert it to a torch tensor.image Image.open(image_path)if max_size:scale max_size / max(image.size)size np.array(image.size) * scaleimage image.resize(size.astype(int), Image.ANTIALIAS)if shape:image image.resize(shape, Image.LANCZOS)if transform:image transform(image).unsqueeze(0)return image.to(device) 这里用的模型是 VGG-19所要用的是网络中的5个卷积层
class VGGNet(nn.Module):def __init__(self):Select conv1_1 ~ conv5_1 activation maps.super(VGGNet, self).__init__()self.select [0, 5, 10, 19, 28] self.vgg models.vgg19(pretrainedTrue).featuresdef forward(self, x):Extract multiple convolutional feature maps.features []for name, layer in self.vgg._modules.items():x layer(x)if name in self.select:features.append(x)return features 模型结构如下可以看到使用序列模型来写的VGG-NET所以标号即层号我们要保存的是[0, 5, 10, 19, 28] 的输出结果。
VGG((features): Sequential((0): Conv2d(3, 64, kernel_size(3, 3), stride(1, 1), padding(1, 1))(1): ReLU(inplace)(2): Conv2d(64, 64, kernel_size(3, 3), stride(1, 1), padding(1, 1))(3): ReLU(inplace)(4): MaxPool2d(kernel_size2, stride2, padding0, dilation1, ceil_modeFalse)(5): Conv2d(64, 128, kernel_size(3, 3), stride(1, 1), padding(1, 1))(6): ReLU(inplace)(7): Conv2d(128, 128, kernel_size(3, 3), stride(1, 1), padding(1, 1))(8): ReLU(inplace)(9): MaxPool2d(kernel_size2, stride2, padding0, dilation1, ceil_modeFalse)(10): Conv2d(128, 256, kernel_size(3, 3), stride(1, 1), padding(1, 1))(11): ReLU(inplace)(12): Conv2d(256, 256, kernel_size(3, 3), stride(1, 1), padding(1, 1))(13): ReLU(inplace)(14): Conv2d(256, 256, kernel_size(3, 3), stride(1, 1), padding(1, 1))(15): ReLU(inplace)(16): Conv2d(256, 256, kernel_size(3, 3), stride(1, 1), padding(1, 1))(17): ReLU(inplace)(18): MaxPool2d(kernel_size2, stride2, padding0, dilation1, ceil_modeFalse)(19): Conv2d(256, 512, kernel_size(3, 3), stride(1, 1), padding(1, 1))(20): ReLU(inplace)(21): Conv2d(512, 512, kernel_size(3, 3), stride(1, 1), padding(1, 1))(22): ReLU(inplace)(23): Conv2d(512, 512, kernel_size(3, 3), stride(1, 1), padding(1, 1))(24): ReLU(inplace)(25): Conv2d(512, 512, kernel_size(3, 3), stride(1, 1), padding(1, 1))(26): ReLU(inplace)(27): MaxPool2d(kernel_size2, stride2, padding0, dilation1, ceil_modeFalse)(28): Conv2d(512, 512, kernel_size(3, 3), stride(1, 1), padding(1, 1))(29): ReLU(inplace)(30): Conv2d(512, 512, kernel_size(3, 3), stride(1, 1), padding(1, 1))(31): ReLU(inplace)(32): Conv2d(512, 512, kernel_size(3, 3), stride(1, 1), padding(1, 1))(33): ReLU(inplace)(34): Conv2d(512, 512, kernel_size(3, 3), stride(1, 1), padding(1, 1))(35): ReLU(inplace)(36): MaxPool2d(kernel_size2, stride2, padding0, dilation1, ceil_modeFalse))(avgpool): AdaptiveAvgPool2d(output_size(7, 7))(classifier): Sequential((0): Linear(in_features25088, out_features4096, biasTrue)(1): ReLU(inplace)(2): Dropout(p0.5)(3): Linear(in_features4096, out_features4096, biasTrue)(4): ReLU(inplace)(5): Dropout(p0.5)(6): Linear(in_features4096, out_features1000, biasTrue))
) 训练
接下来对训练过程进行解释
1、加载风格图像和内容图像我们在之前的博客中使用的一幅加噪图进行训练这里是用的内容图像的拷贝。
2、我们需要优化的就是作为目标的内容图像拷贝可以看到target需要求导。
3、VGGnet参数是不需要优化的所以设置为验证状态。
4、将3幅图像输入网络得到总共15个输出每个图像有5层的输出
5、内容损失这里是遍历5个层的输出来计算损失而在keras版本中只用了第4层的输出计算损失
6、风格损失同样计算格拉姆风格矩阵将每一层的风格损失叠加得到总的风格损失计算公式同样和keras版本有所不一样
7、反向传播
def main(config):# Image preprocessing# VGGNet was trained on ImageNet where images are normalized by mean[0.485, 0.456, 0.406] and std[0.229, 0.224, 0.225].# We use the same normalization statistics here.transform transforms.Compose([transforms.ToTensor(),transforms.Normalize(mean(0.485, 0.456, 0.406), std(0.229, 0.224, 0.225))])# Load content and style images# Make the style image same size as the content imagecontent load_image(config.content, transform, max_sizeconfig.max_size)style load_image(config.style, transform, shape[content.size(2), content.size(3)])# Initialize a target image with the content imagetarget content.clone().requires_grad_(True)optimizer torch.optim.Adam([target], lrconfig.lr, betas[0.5, 0.999])vgg VGGNet().to(device).eval()for step in range(config.total_step):# Extract multiple(5) conv feature vectorstarget_features vgg(target)content_features vgg(content)style_features vgg(style)style_loss 0content_loss 0for f1, f2, f3 in zip(target_features, content_features, style_features):# Compute content loss with target and content imagescontent_loss torch.mean((f1 - f2)**2)# Reshape convolutional feature maps_, c, h, w f1.size()f1 f1.view(c, h * w)f3 f3.view(c, h * w)# Compute gram matrixf1 torch.mm(f1, f1.t())f3 torch.mm(f3, f3.t())# Compute style loss with target and style imagesstyle_loss torch.mean((f1 - f3)**2) / (c * h * w) # Compute total loss, backprop and optimizeloss content_loss config.style_weight * style_loss optimizer.zero_grad()loss.backward()optimizer.step()if (step1) % config.log_step 0:print (Step [{}/{}], Content Loss: {:.4f}, Style Loss: {:.4f} .format(step1, config.total_step, content_loss.item(), style_loss.item()))if (step1) % config.sample_step 0:# Save the generated imagedenorm transforms.Normalize((-2.12, -2.04, -1.80), (4.37, 4.46, 4.44))img target.clone().squeeze()img denorm(img).clamp_(0, 1)torchvision.utils.save_image(img, output-{}.png.format(step1)) 写在if __name____main__后面的语句只会在本脚本中才能被执行被调用时是不会被执行的。
python的命令行工具argparse很优雅的添加参数
但是由于jupyter不支持添加外部参数所以使用了外部博客的方法来支持记住更改读取图片的位置
import sys
if __name__ __main__:#解决方案来自于博客if -f in sys.argv:sys.argv.remove(-f)parser argparse.ArgumentParser()parser.add_argument(--content, typestr, defaultpng/content.png)parser.add_argument(--style, typestr, defaultpng/style.png)parser.add_argument(--max_size, typeint, default400)parser.add_argument(--total_step, typeint, default2000)parser.add_argument(--log_step, typeint, default10)parser.add_argument(--sample_step, typeint, default500)parser.add_argument(--style_weight, typefloat, default100)parser.add_argument(--lr, typefloat, default0.003)#config parser.parse_args()config parser.parse_known_args()[0] #参考博客 https://blog.csdn.net/ken_for_learning/article/details/89675904print(config)main(config)
