Inception块
1.四个路径从不同层面抽取信息,然后在输出通道维合并。
2.有更少的参数个数和计算复杂度(相比于3X3和5X5卷积层)
GoogLeNet
1.五个stages,九个inception块
Inception各种后续变种
1.Inception-BN(V2)——使用batch normalization
2.Inception V3——修改了Inception块
替换5X5->多个3X3卷积层
替换5X5->1X7和7X1卷积层
替换5X5->1X3和3X1卷积层
更深
3.Inception V4——使用残差连接
总结
1.Inception块用4条不同超出单数卷积层和池化层的路来抽取不同信息
它的主要优点是模型参数小,计算复杂度低
2.GoogLenet使用9个Inception块,是第一个达到上百层的网络,后续有改进
代码实现
Inception块由四条并行路径组成。 前三条路径使用窗口大小为1X1、3X3和5X5的卷积层,从不同空间大小中提取信息。 中间的两条路径在输入上执行1X1卷积,以减少通道数,从而降低模型的复杂性。 第四条路径使用3X3最大汇聚层,然后使用1X1卷积层来改变通道数。 这四条路径都使用合适的填充来使输入与输出的高和宽一致,最后我们将每条线路的输出在通道维度上连结,并构成Inception块的输出。在Inception块中,通常调整的超参数是每层输出通道数。
import torch
from torch import nn
from d2l import torch as d2l
from torch.nn import functional as Fclass 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)# 线路2,1x1卷积层后接3x3卷积层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)
数字来源于炼丹(from沐神lol)
b1 = nn.Sequential(nn.Conv2d(1, 64, kernel_size=7, stride=2, padding=3),nn.ReLU(),nn.MaxPool2d(kernel_size=3, stride=2, padding=1))
b2 = nn.Sequential(nn.Conv2d(64, 64, kernel_size=1),nn.ReLU(),nn.Conv2d(64, 192, kernel_size=3, padding=1),nn.ReLU(),nn.MaxPool2d(kernel_size=3, stride=2, padding=1))
b3 = nn.Sequential(Inception(192, 64, (96, 128), (16, 32), 32),Inception(256, 128, (128, 192), (32, 96), 64),nn.MaxPool2d(kernel_size=3, stride=2, padding=1))
b4 = nn.Sequential(Inception(480, 192, (96, 208), (16, 48), 64),Inception(512, 160, (112, 224), (24, 64), 64),Inception(512, 128, (128, 256), (24, 64), 64),Inception(512, 112, (144, 288), (32, 64), 64),Inception(528, 256, (160, 320), (32, 128), 128),nn.MaxPool2d(kernel_size=3, stride=2, padding=1))
b5 = nn.Sequential(Inception(832, 256, (160, 320), (32, 128), 128),Inception(832, 384, (192, 384), (48, 128), 128),nn.AdaptiveAvgPool2d((1,1)),nn.Flatten())net = nn.Sequential(b1, b2, b3, b4, b5, nn.Linear(1024, 10))
将输入的高和宽从224降到96。
下面演示各个模块输出的形状变化。
X = torch.rand(size=(1, 1, 96, 96))
for layer in net:X = layer(X)print(layer.__class__.__name__,'output shape:\t', X.shape)
Sequential output shape: torch.Size([1, 64, 24, 24])
Sequential output shape: torch.Size([1, 192, 12, 12])
Sequential output shape: torch.Size([1, 480, 6, 6])
Sequential output shape: torch.Size([1, 832, 3, 3])
Sequential output shape: torch.Size([1, 1024])
Linear output shape: torch.Size([1, 10])
在训练之前,我们将图片转换为96X96分辨率。
lr, num_epochs, batch_size = 0.1, 10, 128
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size, resize=96)
d2l.train_ch6(net, train_iter, test_iter, num_epochs, lr, d2l.try_gpu())
loss 0.247, train acc 0.907, test acc 0.872
1404.8 examples/sec on cuda:0