🌈🌈🌈现代卷积网络实战系列 总目录

本篇文章的代码运行界面均在Pycharm中进行
本篇文章配套的代码资源已经上传

1、MNIST数据集处理、加载、网络初始化、测试函数
2、训练函数、PyTorch构建LeNet网络
3、PyTorch从零构建AlexNet训练MNIST数据集
4、PyTorch从零构建VGGNet训练MNIST数据集
5、PyTorch从零构建GoogLeNet训练MNIST数据集
6、PyTorch从零构建ResNet训练MNIST数据集

8、VGGNet

2014年，牛津大学计算机视觉组（Visual Geometry Group）和Google DeepMind公司的研究员一起研发出了新的深度卷积神经网络：VGGNet，并取得了ILSVRC2014比赛分类项目的第二名（第一名是GoogLeNet，也是同年提出的).论文下载 Very Deep Convolutional Networks for Large-Scale Image Recognition。论文主要针对卷积神经网络的深度对大规模图像集识别精度的影响，主要贡献是使用很小的卷积核(3×3)构建各种深度的卷积神经网络结构，并对这些网络结构进行了评估，最终证明16-19层的网络深度，能够取得较好的识别精度。 这也就是常用来提取图像特征的VGG-16和VGG-19。

VGG可以看成是加深版的AlexNet，整个网络由卷积层和全连接层叠加而成，和AlexNet不同的是，VGG中使用的都是小尺寸的卷积核(3×3)。
我这里使用的是VGG-16，但是又因为这个系列全部是处理MNIST数据集的，所以我这里的VGG网络只用了3个VGG块，FC也减少了很多参数。

9、VGGNet网络架构

VGGNet(
 (vgg1): VGGBlock(
  (conv1): Conv2d(1, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (relu1): ReLU(inplace=True)
  (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (relu2): ReLU(inplace=True)
  (conv3): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (relu3): ReLU(inplace=True)
  (maxpool1): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
  )
 (vgg2): VGGBlock(
  (conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (relu1): ReLU(inplace=True)
  (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (relu2): ReLU(inplace=True)
  (conv3): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (relu3): ReLU(inplace=True)
  (maxpool1): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
 )
 (vgg3): VGGBlock(
  (conv1): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (relu1): ReLU(inplace=True)
  (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (relu2): ReLU(inplace=True)
  (conv3): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (relu3): ReLU(inplace=True)
  (maxpool1): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
 )
(classifier): Sequential(
  (0): Linear(in_features=12544, out_features=1024, bias=True)
  (1): ReLU(inplace=True)
  (2): Linear(in_features=1024, out_features=512, bias=True)
  (3): ReLU(inplace=True)
  (4): Linear(in_features=512, out_features=10, bias=True)
 )
)

VGG实际上就是很简单，主要是由VGG块组成：
前两组卷积形式一样，每组都是：conv-relu-conv-relu-pool
中间三组卷积形式一样，每组都是：conv-relu-conv-relu-conv-relu-pool
最后分类的三个全连接层：fc-relu-dropout-fc-relu-dropout-fc-softmax

10、PyTorch构建VGGBlock

class VGGBlock(nn.Module):def __init__(self, in_channel, out_channel, num_conv):super(VGGBlock, self).__init__()self.num_conv = num_convself.conv1 = nn.Conv2d(in_channel, out_channel, kernel_size=3, stride=1, padding=1)self.relu1 = nn.ReLU(inplace=True)self.conv2 = nn.Conv2d(out_channel, out_channel, kernel_size=3, stride=1, padding=1)self.relu2 = nn.ReLU(inplace=True)self.conv3 = nn.Conv2d(out_channel, out_channel, kernel_size=3, stride=1, padding=1)self.relu3 = nn.ReLU(inplace=True)self.maxpool1 = nn.MaxPool2d(3, stride=2, padding=1)def forward(self, x):x = self.relu1(self.conv1(x))x = self.relu2(self.conv2(x))if self.num_conv==3:x = self.relu3(self.conv3(x))else:x = self.maxpool1(x)return x

11、PyTorch构建VGGNet

class VGGNet(nn.Module):def __init__(self, num_classes):super(VGGNet, self).__init__()self.vgg1 = VGGBlock(1,64,2)self.vgg2 = VGGBlock(64,128,2)self.vgg3 = VGGBlock(128,256,3)self.classifier = nn.Sequential(nn.Linear(256 * 7 * 7, 1024),nn.ReLU(inplace=True),nn.Linear(1024, 512),nn.ReLU(inplace=True),nn.Linear(512, num_classes))def forward(self, x):x = self.vgg1(x)x = self.vgg2(x)x = self.vgg3(x)x = x.reshape(x.shape[0], -1)x = self.classifier(x)return x

D:\conda\envs\pytorch\python.exe A:\0_MNIST\train.py

Reading data…
train_data: (60000, 28, 28) train_label (60000,)
test_data: (10000, 28, 28) test_label (10000,)

Initialize neural network
test loss 2303.1
test accuracy 10.1%

epoch step 1
training time 8.9s
training loss 204.3
test loss 39.6
test accuracy 98.8%

epoch step 2
training time 8.3s
training loss 48.8
test loss 39.7 test
accuracy 98.8%

epoch step 3
training time 8.1s
training loss 35.9
test loss 26.4
test accuracy 99.1%

Training finished
3 epoch training time 25.4s
One epoch average training time 8.5s

进程已结束，退出代码为 0