PyTorch實現VGG

import numpy as np
import torch
from torch import nn
from torch.autograd import Variable

定義VGG的block

VGG幾乎全部使用3x3的卷積核，以及2x2的池化層，使用小的卷積核進行多層堆疊和一個大的卷積核感受野是相同的，同時VGG小的卷積核還能減少參數，具有更深的結構。

def vgg_block(num_convs, in_channels, out_channels):
    # 定義第一層，並轉化爲 List
    net = [nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=1),nn.ReLU(True)]
    
    # 通過循環定義其他層
    for i in range(num_convs - 1):
        # List每次只能添加一個元素
        # 輸入和輸出channel均爲out_channels
        net.append(nn.Conv2d(out_channels, out_channels, kernel_size=3,padding=1))
        net.append(nn.ReLU(True))
        
    # 定義池化層    
    net.append(nn.MaxPool2d(2, 2))
    # List數據前面加‘*’表示將List拆分爲獨立的參數
    return nn.Sequential(*net)

打印出模型結構

# 3個卷積層，輸入通道 64， 輸出通道 128
block_demo = vgg_block(3, 64, 128)
print(block_demo)

輸出：

Sequential(
  (0): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (1): ReLU(inplace)
  (2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (3): ReLU(inplace)
  (4): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (5): ReLU(inplace)
  (6): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
)

輸入尺寸驗證，定義輸入爲（1, 64, 300, 300）

input_demo = Variable(torch.zeros(1,64, 300, 300))
output_demo = block_demo(input_demo)
print(output_demo.shape)

輸出：

torch.Size([1, 128, 150, 150])

經過一個VGG模塊，輸入大小減半，通道數爲128

定義循環函數，對VGG模塊堆疊

def vgg_stack(num_convs, channels):
    net = []
    for n, c in zip(num_convs, channels):
        in_c = c[0]
        out_c = c[1]
        net.append(vgg_block(n, in_c, out_c))
    return nn.Sequential(*net)
        

利用循環函數，定義一個簡單的VGG結構，其中有8個卷積層

vgg_net = vgg_stack((1, 1, 2, 2, 2),((3, 64),(64, 128),(128, 256),(256, 512),(512, 512)))
print(vgg_net)

輸出：

Sequential(
  (0): Sequential(
    (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): ReLU(inplace)
    (2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (1): Sequential(
    (0): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): ReLU(inplace)
    (2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (2): Sequential(
    (0): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): ReLU(inplace)
    (2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (3): ReLU(inplace)
    (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (3): Sequential(
    (0): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): ReLU(inplace)
    (2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (3): ReLU(inplace)
    (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (4): Sequential(
    (0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): ReLU(inplace)
    (2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (3): ReLU(inplace)
    (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
)

5個最大池化層，圖片最後會縮減5倍(2^5)

添加全連接層

class vgg(nn.Module):
    def __init__(self):
        super(vgg, self).__init__()
        self.feature = vgg_net
        self.fc = nn.Sequential(
            nn.Linear(512, 100),
            nn.ReLU(True),
            nn.Linear(100, 10)
        )
    def forward(self, x):
        x = self.feature(x)
        x = self.view(x.shape[0], -1)
        x = self.fc(x)
        return x

VGG_Net = vgg()

print(VGG_Net)

輸出：

vgg(
  (feature): Sequential(
    (0): Sequential(
      (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (1): ReLU(inplace)
      (2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    )
    (1): Sequential(
      (0): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (1): ReLU(inplace)
      (2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    )
    (2): Sequential(
      (0): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (1): ReLU(inplace)
      (2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (3): ReLU(inplace)
      (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    )
    (3): Sequential(
      (0): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (1): ReLU(inplace)
      (2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (3): ReLU(inplace)
      (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    )
    (4): Sequential(
      (0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (1): ReLU(inplace)
      (2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (3): ReLU(inplace)
      (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    )
  )
  (fc): Sequential(
    (0): Linear(in_features=512, out_features=100, bias=True)
    (1): ReLU(inplace)
    (2): Linear(in_features=100, out_features=10, bias=True)
  )
)