VGG的网络结构
如下图所示
使用Pytorch框架构建VGG16
# 定义VGG16网络
import torch
import pdb
import torch.nn as nn
import torch.nn.functional as F
chanels=[64,64,'MP',128,128,'MP',256,256,256,'MP',512,512,512,'MP',512,512,512,'MP3','FC','FC1','FC2']
class VGGNet(nn.Module):
#定义构造函数
def __init__(self,Chanels,num_class):
super(VGGNet,self).__init__()
self.num_class=num_class
layers=[]
input_chanel=3
for ar in Chanels:
if(ar=='MP'):
layers.append(nn.MaxPool2d(kernel_size=2,stride=2))
elif(ar=='MP3'):
layers.append(nn.MaxPool2d(3,stride=1,padding=1))
elif(ar=='FC'):
layers.append(nn.Conv2d(512,1024,3,padding=6,dilation=6))
layers.append(nn.ReLU(inplace=True))
elif(ar=='FC1'):
layers.append(nn.Conv2d(1024,1024,1))
layers.append(nn.ReLU(inplace=True))
elif(ar=='FC2'):
layers.append(nn.Conv2d(1024,self.num_class,1))
else:
layers.append(nn.Conv2d(in_channels=input_chanel,out_channels=ar,kernel_size=3,padding=1))
layers.append(nn.ReLU(inplace=True))
input_chanel=ar
self.Layers=nn.ModuleList(layers)
def forward(self,x):
for lays in self.Layers:
x=lays(x)
out=x
return out
inputs=torch.randn(1,3,300,300)
vgg16=VGGNet(chanels,10)
output=vgg16(inputs)
print(output.shape)
