下图为2014年ImageNet-2014竞赛第一网络GoogleNet的V1版本原论文的结构图
Inception block主要有两个重点:1x1卷积与多尺度并行卷积组合
下面以结构里的第一个inception模块3a为例子搭建网络
输入为1x192x28x28
import torch
from torch import nn
from torch.nn import functional as F
import torchvision
#Inception块的搭建,总体网络搭建较为繁琐
class Inception(nn.Module):
# c1--c4是每条路径的输出通道数
def __init__(self, in_channels, c1, c2, c3, c4, **kwargs):
super(Inception, self).__init__(**kwargs)
# 线路1,单1x1卷积层
self.p1_1 = nn.Conv2d(in_channels, c1, kernel_size=1)#sride默认为1
# 线路2,1x1卷积层后接3x3卷积层,#2_1代表第二层的第一步
self.p2_1 = nn.Conv2d(in_channels, c2[0], kernel_size=1)
self.p2_2 = nn.Conv2d(c2[0], c2[1], kernel_size=3, padding=1)
# 线路3,1x1卷积层后接5x5卷积层
self.p3_1 = nn.Conv2d(in_channels, c3[0], kernel_size=1)
self.p3_2 = nn.Conv2d(c3[0], c3[1], kernel_size=5, padding=2)
# 线路4,3x3最大汇聚层后接1x1卷积层
self.p4_1 = nn.MaxPool2d(kernel_size=3, stride=1, padding=1)
self.p4_2 = nn.Conv2d(in_channels, c4, kernel_size=1)
def forward(self, x):
p1 = F.relu(self.p1_1(x))
p2 = F.relu(self.p2_2(F.relu(self.p2_1(x))))
p3 = F.relu(self.p3_2(F.relu(self.p3_1(x))))
p4 = F.relu(self.p4_2(self.p4_1(x)))
# 在通道维度上连结输出
return torch.cat((p1, p2, p3, p4), dim=1)
test1=Inception(192,c1=64,c2=[96,128],c3=[16,32],c4=32)
img=torch.randn(size=(1,192,28,28))
out=test1(img)
print(out.shape)
#
# model = torchvision.models.googlenet.GoogLeNet()
# # model = models.googlenet(weights='DEFAULT')
# print(model)
利用随机数构建相应维度的张量输入,可以得到输出与原论文256个通道一致