SENet代码复现＋超详细注释（PyTorch）

在卷积网络中通道注意力经常用到SENet模块，来增强网络模型在通道权重的选择能力，进而提点。关于SENet的原理和具体细节，我们在上一篇已经详细的介绍了：经典神经网络论文超详细解读（七）——SENet（注意力机制）学习笔记（翻译＋精读＋代码复现）

接下来我们来复现一下代码。

因为SENet不是一个全新的网络模型，而是相当于提出了一个即插即用的高性能小插件，所以代码实现也是比较简单的。本文是在ResNet基础上加入SEblock模块进行实现ResNet_SE50。

一、SENet结构组成介绍

​

上图为一个SEblock，由SEblock块构成的网络叫做SENet；可以基于原生网络，添加SEblock块构成SE-NameNet，如基于AlexNet等添加SE结构，称作SE-AlexNet、SE-ResNet等

SE块与先进的架构Inception、ResNet的结合效果

​

原理：通过一个全局平均池化层加两个全连接层以及全连接层对应激活【ReLU和sigmoid】组成的结构输出和输入特征同样数目的权重值，也就是每个特征通道的权重系数，学习一个通道的注意力出来，用于决定哪些通道应该重点提取特征，哪些部分放弃。

SE块详细过程

1.首先由 Inception结构 或 ResNet结构处理后的C×W×H特征图开始，通过Squeeze操作对特征图进行全局平均池化（GAP），得到1×1×C 的特征向量

2.紧接着两个 FC 层组成一个 Bottleneck 结构去建模通道间的相关性：

（1）经过第一个FC层，将C个通道变成 C/ r​ ，减少参数量，然后通过ReLU的非线性激活，到达第二个FC层

（2）经过第二个FC层，再将特征通道数恢复到C个，得到带有注意力机制的权重参数

3.最后经过Sigmoid激活函数，最后通过一个 Scale 的操作来将归一化后的权重加权到每个通道的特征上。

二、SEblock的具体介绍

Sequeeze：Fsq操作就是使用通道的全局平均池化，将包含全局信息的W×H×C 的特征图直接压缩成一个1×1×C的特征向量，即将每个二维通道变成一个具有全局感受野的数值，此时1个像素表示1个通道，屏蔽掉空间上的分布信息，更好的利用通道间的相关性。
具体操作：对原特征图50×512×7×7进行全局平均池化,然后得到了一个50×512×1×1大小的特征图，这个特征图具有全局感受野。

Excitation ：基于特征通道间的相关性，每个特征通道生成一个权重，用来代表特征通道的重要程度。由原本全为白色的C个通道的特征，得到带有不同深浅程度的颜色的特征向量，也就是不同的重要程度。

具体操作：输出的50×512×1×1特征图，经过两个全连接层，最后用一 个类似于循环神经网络中门控机制，通过参数来为每个特征通道生成权重，参数被学习用来显式地建模特征通道间的相关性(论文中使用的是sigmoid)。50×512×1×1变成50×512 / 16×1×1，最后再还原回来：50×512×1×1

Reweight：将Excitation输出的权重看做每个特征通道的重要性，也就是对于U每个位置上的所有H×W上的值都乘上对应通道的权值，完成对原始特征的重校准。

具体操作：50×512×1×1通过expand_as得到50×512×7×7, 完成在通道维度上对原始特征的重标定，并作为下一级的输入数据。

三、PyTorch代码实现

（1）SEblock搭建

全局平均池化+1*1卷积核+ReLu+1*1卷积核+Sigmoid

'''-------------一、SE模块-----------------------------'''#全局平均池化+1*1卷积核+ReLu+1*1卷积核+Sigmoidclass SE_Block(nn.Module):def __init__(self, inchannel, ratio=16):super(SE_Block, self).__init__()# 全局平均池化(Fsq操作)self.gap = nn.AdaptiveAvgPool2d((1, 1))# 两个全连接层(Fex操作)self.fc = nn.Sequential(nn.Linear(inchannel, inchannel // ratio, bias=False),# 从 c -> c/rnn.ReLU(),nn.Linear(inchannel // ratio, inchannel, bias=False),# 从 c/r -> cnn.Sigmoid())def forward(self, x):# 读取批数据图片数量及通道数b, c, h, w = x.size()# Fsq操作：经池化后输出b*c的矩阵y = self.gap(x).view(b, c)# Fex操作：经全连接层输出（b，c，1，1）矩阵y = self.fc(y).view(b, c, 1, 1)# Fscale操作：将得到的权重乘以原来的特征图xreturn x * y.expand_as(x)

（2）将SEblock嵌入残差模块

SEblock可以灵活的加入到resnet等相关完整模型中，通常加在残差之前。【因为激活是sigmoid原因，存在梯度弥散问题，所以尽量不放到主信号通道去，即使本个残差模块有弥散问题，以不至于影响整个网络模型】

​

这里我们将SE模块分别嵌入ResNet的BasicBlock和Bottleneck中，得到 SEBasicBlock和SEBottleneck（具体解释可以看我之前写的ResNet代码复现＋超详细注释（PyTorch））

BasicBlock模块

'''-------------二、BasicBlock模块-----------------------------'''# 左侧的 residual block 结构（18-layer、34-layer）class BasicBlock(nn.Module):expansion = 1def __init__(self, inchannel, outchannel, stride=1):super(BasicBlock, self).__init__()self.conv1 = nn.Conv2d(inchannel, outchannel, kernel_size=3, stride=stride, padding=1, bias=False)self.bn1 = nn.BatchNorm2d(outchannel)self.conv2 = nn.Conv2d(outchannel, outchannel, kernel_size=3, stride=1, padding=1, bias=False)self.bn2 = nn.BatchNorm2d(outchannel)# SE_Block放在BN之后，shortcut之前self.SE = SE_Block(outchannel)self.shortcut = nn.Sequential()if stride != 1 or inchannel != self.expansion*outchannel:self.shortcut = nn.Sequential(nn.Conv2d(inchannel, self.expansion*outchannel,kernel_size=1, stride=stride, bias=False),nn.BatchNorm2d(self.expansion*outchannel))def forward(self, x):out = F.relu(self.bn1(self.conv1(x)))out = self.bn2(self.conv2(out))SE_out = self.SE(out)out = out * SE_outout += self.shortcut(x)out = F.relu(out)return out

Bottleneck模块

'''-------------三、Bottleneck模块-----------------------------'''# 右侧的 residual block 结构（50-layer、101-layer、152-layer）class Bottleneck(nn.Module):expansion = 4def __init__(self, inchannel, outchannel, stride=1):super(Bottleneck, self).__init__()self.conv1 = nn.Conv2d(inchannel, outchannel, kernel_size=1, bias=False)self.bn1 = nn.BatchNorm2d(outchannel)self.conv2 = nn.Conv2d(outchannel, outchannel, kernel_size=3, stride=stride, padding=1, bias=False)self.bn2 = nn.BatchNorm2d(outchannel)self.conv3 = nn.Conv2d(outchannel, self.expansion*outchannel, kernel_size=1, bias=False)self.bn3 = nn.BatchNorm2d(self.expansion*outchannel)# SE_Block放在BN之后，shortcut之前self.SE = SE_Block(self.expansion*outchannel)self.shortcut = nn.Sequential()if stride != 1 or inchannel != self.expansion*outchannel:self.shortcut = nn.Sequential(nn.Conv2d(inchannel, self.expansion*outchannel,kernel_size=1, stride=stride, bias=False),nn.BatchNorm2d(self.expansion*outchannel))def forward(self, x):out = F.relu(self.bn1(self.conv1(x)))out = F.relu(self.bn2(self.conv2(out)))out = self.bn3(self.conv3(out))SE_out = self.SE(out)out = out * SE_outout += self.shortcut(x)out = F.relu(out)return out

（3）搭建SE_ResNet结构

'''-------------四、搭建SE_ResNet结构-----------------------------'''class SE_ResNet(nn.Module):def __init__(self, block, num_blocks, num_classes=10):super(SE_ResNet, self).__init__()self.in_planes = 64self.conv1 = nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1, bias=False)# conv1self.bn1 = nn.BatchNorm2d(64)self.layer1 = self._make_layer(block, 64, num_blocks[0], stride=1) # conv2_xself.layer2 = self._make_layer(block, 128, num_blocks[1], stride=2)# conv3_xself.layer3 = self._make_layer(block, 256, num_blocks[2], stride=2)# conv4_xself.layer4 = self._make_layer(block, 512, num_blocks[3], stride=2)# conv5_xself.avgpool = nn.AdaptiveAvgPool2d((1, 1))self.linear = nn.Linear(512 * block.expansion, num_classes)def _make_layer(self, block, planes, num_blocks, stride):strides = [stride] + [1]*(num_blocks-1)layers = []for stride in strides:layers.append(block(self.in_planes, planes, stride))self.in_planes = planes * block.expansionreturn nn.Sequential(*layers)def forward(self, x):x = F.relu(self.bn1(self.conv1(x)))x = self.layer1(x)x = self.layer2(x)x = self.layer3(x)x = self.layer4(x)x = self.avgpool(x)x = torch.flatten(x, 1)out = self.linear(x)return out

（4）网络模型的创建和测试

网络模型创建打印SE_ResNet50

# test()if __name__ == '__main__':model = SE_ResNet50()print(model)input = torch.randn(1, 3, 224, 224)out = model(input)print(out.shape)

打印模型如下

SE_ResNet((conv1): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(layer1): Sequential((0): Bottleneck((conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=256, out_features=16, bias=False)(1): ReLU()(2): Linear(in_features=16, out_features=256, bias=False)(3): Sigmoid()))(shortcut): Sequential((0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(1): Bottleneck((conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=256, out_features=16, bias=False)(1): ReLU()(2): Linear(in_features=16, out_features=256, bias=False)(3): Sigmoid()))(shortcut): Sequential())(2): Bottleneck((conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=256, out_features=16, bias=False)(1): ReLU()(2): Linear(in_features=16, out_features=256, bias=False)(3): Sigmoid()))(shortcut): Sequential()))(layer2): Sequential((0): Bottleneck((conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=512, out_features=32, bias=False)(1): ReLU()(2): Linear(in_features=32, out_features=512, bias=False)(3): Sigmoid()))(shortcut): Sequential((0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(1): Bottleneck((conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=512, out_features=32, bias=False)(1): ReLU()(2): Linear(in_features=32, out_features=512, bias=False)(3): Sigmoid()))(shortcut): Sequential())(2): Bottleneck((conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=512, out_features=32, bias=False)(1): ReLU()(2): Linear(in_features=32, out_features=512, bias=False)(3): Sigmoid()))(shortcut): Sequential())(3): Bottleneck((conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=512, out_features=32, bias=False)(1): ReLU()(2): Linear(in_features=32, out_features=512, bias=False)(3): Sigmoid()))(shortcut): Sequential()))(layer3): Sequential((0): Bottleneck((conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=1024, out_features=64, bias=False)(1): ReLU()(2): Linear(in_features=64, out_features=1024, bias=False)(3): Sigmoid()))(shortcut): Sequential((0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)(1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(1): Bottleneck((conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=1024, out_features=64, bias=False)(1): ReLU()(2): Linear(in_features=64, out_features=1024, bias=False)(3): Sigmoid()))(shortcut): Sequential())(2): Bottleneck((conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=1024, out_features=64, bias=False)(1): ReLU()(2): Linear(in_features=64, out_features=1024, bias=False)(3): Sigmoid()))(shortcut): Sequential())(3): Bottleneck((conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=1024, out_features=64, bias=False)(1): ReLU()(2): Linear(in_features=64, out_features=1024, bias=False)(3): Sigmoid()))(shortcut): Sequential())(4): Bottleneck((conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=1024, out_features=64, bias=False)(1): ReLU()(2): Linear(in_features=64, out_features=1024, bias=False)(3): Sigmoid()))(shortcut): Sequential())(5): Bottleneck((conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=1024, out_features=64, bias=False)(1): ReLU()(2): Linear(in_features=64, out_features=1024, bias=False)(3): Sigmoid()))(shortcut): Sequential()))(layer4): Sequential((0): Bottleneck((conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=2048, out_features=128, bias=False)(1): ReLU()(2): Linear(in_features=128, out_features=2048, bias=False)(3): Sigmoid()))(shortcut): Sequential((0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)(1): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(1): Bottleneck((conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=2048, out_features=128, bias=False)(1): ReLU()(2): Linear(in_features=128, out_features=2048, bias=False)(3): Sigmoid()))(shortcut): Sequential())(2): Bottleneck((conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(SE): SE_Block((gap): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Sequential((0): Linear(in_features=2048, out_features=128, bias=False)(1): ReLU()(2): Linear(in_features=128, out_features=2048, bias=False)(3): Sigmoid()))(shortcut): Sequential()))(avgpool): AdaptiveAvgPool2d(output_size=(1, 1))(linear): Linear(in_features=2048, out_features=10, bias=True))torch.Size([1, 10])

使用torchsummary打印每个网络模型的详细信息

if __name__ == '__main__':net = SE_ResNet50().cuda()summary(net, (3, 224, 224))

打印模型如下

----------------------------------------------------------------Layer (type) Output Shape Param #================================================================Conv2d-1 [-1, 64, 224, 224] 1,728 BatchNorm2d-2 [-1, 64, 224, 224] 128Conv2d-3 [-1, 64, 224, 224] 4,096 BatchNorm2d-4 [-1, 64, 224, 224] 128Conv2d-5 [-1, 64, 224, 224]36,864 BatchNorm2d-6 [-1, 64, 224, 224] 128Conv2d-7[-1, 256, 224, 224]16,384 BatchNorm2d-8[-1, 256, 224, 224] 512 AdaptiveAvgPool2d-9[-1, 256, 1, 1] 0 Linear-10 [-1, 16] 4,096 ReLU-11 [-1, 16] 0 Linear-12[-1, 256] 4,096Sigmoid-13[-1, 256] 0 SE_Block-14[-1, 256, 224, 224] 0 Conv2d-15[-1, 256, 224, 224]16,384BatchNorm2d-16[-1, 256, 224, 224] 512 Bottleneck-17[-1, 256, 224, 224] 0 Conv2d-18 [-1, 64, 224, 224]16,384BatchNorm2d-19 [-1, 64, 224, 224] 128 Conv2d-20 [-1, 64, 224, 224]36,864BatchNorm2d-21 [-1, 64, 224, 224] 128 Conv2d-22[-1, 256, 224, 224]16,384BatchNorm2d-23[-1, 256, 224, 224] 512AdaptiveAvgPool2d-24[-1, 256, 1, 1] 0 Linear-25 [-1, 16] 4,096 ReLU-26 [-1, 16] 0 Linear-27[-1, 256] 4,096Sigmoid-28[-1, 256] 0 SE_Block-29[-1, 256, 224, 224] 0 Bottleneck-30[-1, 256, 224, 224] 0 Conv2d-31 [-1, 64, 224, 224]16,384BatchNorm2d-32 [-1, 64, 224, 224] 128 Conv2d-33 [-1, 64, 224, 224]36,864BatchNorm2d-34 [-1, 64, 224, 224] 128 Conv2d-35[-1, 256, 224, 224]16,384BatchNorm2d-36[-1, 256, 224, 224] 512AdaptiveAvgPool2d-37[-1, 256, 1, 1] 0 Linear-38 [-1, 16] 4,096 ReLU-39 [-1, 16] 0 Linear-40[-1, 256] 4,096Sigmoid-41[-1, 256] 0 SE_Block-42[-1, 256, 224, 224] 0 Bottleneck-43[-1, 256, 224, 224] 0 Conv2d-44[-1, 128, 224, 224]32,768BatchNorm2d-45[-1, 128, 224, 224] 256 Conv2d-46[-1, 128, 112, 112] 147,456BatchNorm2d-47[-1, 128, 112, 112] 256 Conv2d-48[-1, 512, 112, 112]65,536BatchNorm2d-49[-1, 512, 112, 112] 1,024AdaptiveAvgPool2d-50[-1, 512, 1, 1] 0 Linear-51 [-1, 32]16,384 ReLU-52 [-1, 32] 0 Linear-53[-1, 512]16,384Sigmoid-54[-1, 512] 0 SE_Block-55[-1, 512, 112, 112] 0 Conv2d-56[-1, 512, 112, 112] 131,072BatchNorm2d-57[-1, 512, 112, 112] 1,024 Bottleneck-58[-1, 512, 112, 112] 0 Conv2d-59[-1, 128, 112, 112]65,536BatchNorm2d-60[-1, 128, 112, 112] 256 Conv2d-61[-1, 128, 112, 112] 147,456BatchNorm2d-62[-1, 128, 112, 112] 256 Conv2d-63[-1, 512, 112, 112]65,536BatchNorm2d-64[-1, 512, 112, 112] 1,024AdaptiveAvgPool2d-65[-1, 512, 1, 1] 0 Linear-66 [-1, 32]16,384 ReLU-67 [-1, 32] 0 Linear-68[-1, 512]16,384Sigmoid-69[-1, 512] 0 SE_Block-70[-1, 512, 112, 112] 0 Bottleneck-71[-1, 512, 112, 112] 0 Conv2d-72[-1, 128, 112, 112]65,536BatchNorm2d-73[-1, 128, 112, 112] 256 Conv2d-74[-1, 128, 112, 112] 147,456BatchNorm2d-75[-1, 128, 112, 112] 256 Conv2d-76[-1, 512, 112, 112]65,536BatchNorm2d-77[-1, 512, 112, 112] 1,024AdaptiveAvgPool2d-78[-1, 512, 1, 1] 0 Linear-79 [-1, 32]16,384 ReLU-80 [-1, 32] 0 Linear-81[-1, 512]16,384Sigmoid-82[-1, 512] 0 SE_Block-83[-1, 512, 112, 112] 0 Bottleneck-84[-1, 512, 112, 112] 0 Conv2d-85[-1, 128, 112, 112]65,536BatchNorm2d-86[-1, 128, 112, 112] 256 Conv2d-87[-1, 128, 112, 112] 147,456BatchNorm2d-88[-1, 128, 112, 112] 256 Conv2d-89[-1, 512, 112, 112]65,536BatchNorm2d-90[-1, 512, 112, 112] 1,024AdaptiveAvgPool2d-91[-1, 512, 1, 1] 0 Linear-92 [-1, 32]16,384 ReLU-93 [-1, 32] 0 Linear-94[-1, 512]16,384Sigmoid-95[-1, 512] 0 SE_Block-96[-1, 512, 112, 112] 0 Bottleneck-97[-1, 512, 112, 112] 0 Conv2d-98[-1, 256, 112, 112] 131,072BatchNorm2d-99[-1, 256, 112, 112] 512Conv2d-100[-1, 256, 56, 56] 589,824 BatchNorm2d-101[-1, 256, 56, 56] 512Conv2d-102 [-1, 1024, 56, 56] 262,144 BatchNorm2d-103 [-1, 1024, 56, 56] 2,048AdaptiveAvgPool2d-104 [-1, 1024, 1, 1] 0Linear-105 [-1, 64]65,536ReLU-106 [-1, 64] 0Linear-107 [-1, 1024]65,536 Sigmoid-108 [-1, 1024] 0SE_Block-109 [-1, 1024, 56, 56] 0Conv2d-110 [-1, 1024, 56, 56] 524,288 BatchNorm2d-111 [-1, 1024, 56, 56] 2,048Bottleneck-112 [-1, 1024, 56, 56] 0Conv2d-113[-1, 256, 56, 56] 262,144 BatchNorm2d-114[-1, 256, 56, 56] 512Conv2d-115[-1, 256, 56, 56] 589,824 BatchNorm2d-116[-1, 256, 56, 56] 512Conv2d-117 [-1, 1024, 56, 56] 262,144 BatchNorm2d-118 [-1, 1024, 56, 56] 2,048AdaptiveAvgPool2d-119 [-1, 1024, 1, 1] 0Linear-120 [-1, 64]65,536ReLU-121 [-1, 64] 0Linear-122 [-1, 1024]65,536 Sigmoid-123 [-1, 1024] 0SE_Block-124 [-1, 1024, 56, 56] 0Bottleneck-125 [-1, 1024, 56, 56] 0Conv2d-126[-1, 256, 56, 56] 262,144 BatchNorm2d-127[-1, 256, 56, 56] 512Conv2d-128[-1, 256, 56, 56] 589,824 BatchNorm2d-129[-1, 256, 56, 56] 512Conv2d-130 [-1, 1024, 56, 56] 262,144 BatchNorm2d-131 [-1, 1024, 56, 56] 2,048AdaptiveAvgPool2d-132 [-1, 1024, 1, 1] 0Linear-133 [-1, 64]65,536ReLU-134 [-1, 64] 0Linear-135 [-1, 1024]65,536 Sigmoid-136 [-1, 1024] 0SE_Block-137 [-1, 1024, 56, 56] 0Bottleneck-138 [-1, 1024, 56, 56] 0Conv2d-139[-1, 256, 56, 56] 262,144 BatchNorm2d-140[-1, 256, 56, 56] 512Conv2d-141[-1, 256, 56, 56] 589,824 BatchNorm2d-142[-1, 256, 56, 56] 512Conv2d-143 [-1, 1024, 56, 56] 262,144 BatchNorm2d-144 [-1, 1024, 56, 56] 2,048AdaptiveAvgPool2d-145 [-1, 1024, 1, 1] 0Linear-146 [-1, 64]65,536ReLU-147 [-1, 64] 0Linear-148 [-1, 1024]65,536 Sigmoid-149 [-1, 1024] 0SE_Block-150 [-1, 1024, 56, 56] 0Bottleneck-151 [-1, 1024, 56, 56] 0Conv2d-152[-1, 256, 56, 56] 262,144 BatchNorm2d-153[-1, 256, 56, 56] 512Conv2d-154[-1, 256, 56, 56] 589,824 BatchNorm2d-155[-1, 256, 56, 56] 512Conv2d-156 [-1, 1024, 56, 56] 262,144 BatchNorm2d-157 [-1, 1024, 56, 56] 2,048AdaptiveAvgPool2d-158 [-1, 1024, 1, 1] 0Linear-159 [-1, 64]65,536ReLU-160 [-1, 64] 0Linear-161 [-1, 1024]65,536 Sigmoid-162 [-1, 1024] 0SE_Block-163 [-1, 1024, 56, 56] 0Bottleneck-164 [-1, 1024, 56, 56] 0Conv2d-165[-1, 256, 56, 56] 262,144 BatchNorm2d-166[-1, 256, 56, 56] 512Conv2d-167[-1, 256, 56, 56] 589,824 BatchNorm2d-168[-1, 256, 56, 56] 512Conv2d-169 [-1, 1024, 56, 56] 262,144 BatchNorm2d-170 [-1, 1024, 56, 56] 2,048AdaptiveAvgPool2d-171 [-1, 1024, 1, 1] 0Linear-172 [-1, 64]65,536ReLU-173 [-1, 64] 0Linear-174 [-1, 1024]65,536 Sigmoid-175 [-1, 1024] 0SE_Block-176 [-1, 1024, 56, 56] 0Bottleneck-177 [-1, 1024, 56, 56] 0Conv2d-178[-1, 512, 56, 56] 524,288 BatchNorm2d-179[-1, 512, 56, 56] 1,024Conv2d-180[-1, 512, 28, 28] 2,359,296 BatchNorm2d-181[-1, 512, 28, 28] 1,024Conv2d-182 [-1, 2048, 28, 28] 1,048,576 BatchNorm2d-183 [-1, 2048, 28, 28] 4,096AdaptiveAvgPool2d-184 [-1, 2048, 1, 1] 0Linear-185[-1, 128] 262,144ReLU-186[-1, 128] 0Linear-187 [-1, 2048] 262,144 Sigmoid-188 [-1, 2048] 0SE_Block-189 [-1, 2048, 28, 28] 0Conv2d-190 [-1, 2048, 28, 28] 2,097,152 BatchNorm2d-191 [-1, 2048, 28, 28] 4,096Bottleneck-192 [-1, 2048, 28, 28] 0Conv2d-193[-1, 512, 28, 28] 1,048,576 BatchNorm2d-194[-1, 512, 28, 28] 1,024Conv2d-195[-1, 512, 28, 28] 2,359,296 BatchNorm2d-196[-1, 512, 28, 28] 1,024Conv2d-197 [-1, 2048, 28, 28] 1,048,576 BatchNorm2d-198 [-1, 2048, 28, 28] 4,096AdaptiveAvgPool2d-199 [-1, 2048, 1, 1] 0Linear-200[-1, 128] 262,144ReLU-201[-1, 128] 0Linear-202 [-1, 2048] 262,144 Sigmoid-203 [-1, 2048] 0SE_Block-204 [-1, 2048, 28, 28] 0Bottleneck-205 [-1, 2048, 28, 28] 0Conv2d-206[-1, 512, 28, 28] 1,048,576 BatchNorm2d-207[-1, 512, 28, 28] 1,024Conv2d-208[-1, 512, 28, 28] 2,359,296 BatchNorm2d-209[-1, 512, 28, 28] 1,024Conv2d-210 [-1, 2048, 28, 28] 1,048,576 BatchNorm2d-211 [-1, 2048, 28, 28] 4,096AdaptiveAvgPool2d-212 [-1, 2048, 1, 1] 0Linear-213[-1, 128] 262,144ReLU-214[-1, 128] 0Linear-215 [-1, 2048] 262,144 Sigmoid-216 [-1, 2048] 0SE_Block-217 [-1, 2048, 28, 28] 0Bottleneck-218 [-1, 2048, 28, 28] 0AdaptiveAvgPool2d-219 [-1, 2048, 1, 1] 0Linear-220 [-1, 10]20,490================================================================Total params: 26,035,786Trainable params: 26,035,786Non-trainable params: 0----------------------------------------------------------------Input size (MB): 0.57Forward/backward pass size (MB): 3914.25Params size (MB): 99.32Estimated Total Size (MB): 4014.14----------------------------------------------------------------Process finished with exit code 0

（5）完整代码

import torchimport torch.nn as nnimport torch.nn.functional as Ffrom torchsummary import summary'''-------------一、SE模块-----------------------------'''#全局平均池化+1*1卷积核+ReLu+1*1卷积核+Sigmoidclass SE_Block(nn.Module):def __init__(self, inchannel, ratio=16):super(SE_Block, self).__init__()# 全局平均池化(Fsq操作)self.gap = nn.AdaptiveAvgPool2d((1, 1))# 两个全连接层(Fex操作)self.fc = nn.Sequential(nn.Linear(inchannel, inchannel // ratio, bias=False),# 从 c -> c/rnn.ReLU(),nn.Linear(inchannel // ratio, inchannel, bias=False),# 从 c/r -> cnn.Sigmoid())def forward(self, x):# 读取批数据图片数量及通道数b, c, h, w = x.size()# Fsq操作：经池化后输出b*c的矩阵y = self.gap(x).view(b, c)# Fex操作：经全连接层输出（b，c，1，1）矩阵y = self.fc(y).view(b, c, 1, 1)# Fscale操作：将得到的权重乘以原来的特征图xreturn x * y.expand_as(x)'''-------------二、BasicBlock模块-----------------------------'''# 左侧的 residual block 结构（18-layer、34-layer）class BasicBlock(nn.Module):expansion = 1def __init__(self, inchannel, outchannel, stride=1):super(BasicBlock, self).__init__()self.conv1 = nn.Conv2d(inchannel, outchannel, kernel_size=3, stride=stride, padding=1, bias=False)self.bn1 = nn.BatchNorm2d(outchannel)self.conv2 = nn.Conv2d(outchannel, outchannel, kernel_size=3, stride=1, padding=1, bias=False)self.bn2 = nn.BatchNorm2d(outchannel)# SE_Block放在BN之后，shortcut之前self.SE = SE_Block(outchannel)self.shortcut = nn.Sequential()if stride != 1 or inchannel != self.expansion*outchannel:self.shortcut = nn.Sequential(nn.Conv2d(inchannel, self.expansion*outchannel,kernel_size=1, stride=stride, bias=False),nn.BatchNorm2d(self.expansion*outchannel))def forward(self, x):out = F.relu(self.bn1(self.conv1(x)))out = self.bn2(self.conv2(out))SE_out = self.SE(out)out = out * SE_outout += self.shortcut(x)out = F.relu(out)return out'''-------------三、Bottleneck模块-----------------------------'''# 右侧的 residual block 结构（50-layer、101-layer、152-layer）class Bottleneck(nn.Module):expansion = 4def __init__(self, inchannel, outchannel, stride=1):super(Bottleneck, self).__init__()self.conv1 = nn.Conv2d(inchannel, outchannel, kernel_size=1, bias=False)self.bn1 = nn.BatchNorm2d(outchannel)self.conv2 = nn.Conv2d(outchannel, outchannel, kernel_size=3, stride=stride, padding=1, bias=False)self.bn2 = nn.BatchNorm2d(outchannel)self.conv3 = nn.Conv2d(outchannel, self.expansion*outchannel, kernel_size=1, bias=False)self.bn3 = nn.BatchNorm2d(self.expansion*outchannel)# SE_Block放在BN之后，shortcut之前self.SE = SE_Block(self.expansion*outchannel)self.shortcut = nn.Sequential()if stride != 1 or inchannel != self.expansion*outchannel:self.shortcut = nn.Sequential(nn.Conv2d(inchannel, self.expansion*outchannel,kernel_size=1, stride=stride, bias=False),nn.BatchNorm2d(self.expansion*outchannel))def forward(self, x):out = F.relu(self.bn1(self.conv1(x)))out = F.relu(self.bn2(self.conv2(out)))out = self.bn3(self.conv3(out))SE_out = self.SE(out)out = out * SE_outout += self.shortcut(x)out = F.relu(out)return out'''-------------四、搭建SE_ResNet结构-----------------------------'''class SE_ResNet(nn.Module):def __init__(self, block, num_blocks, num_classes=10):super(SE_ResNet, self).__init__()self.in_planes = 64self.conv1 = nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1, bias=False)# conv1self.bn1 = nn.BatchNorm2d(64)self.layer1 = self._make_layer(block, 64, num_blocks[0], stride=1) # conv2_xself.layer2 = self._make_layer(block, 128, num_blocks[1], stride=2)# conv3_xself.layer3 = self._make_layer(block, 256, num_blocks[2], stride=2)# conv4_xself.layer4 = self._make_layer(block, 512, num_blocks[3], stride=2)# conv5_xself.avgpool = nn.AdaptiveAvgPool2d((1, 1))self.linear = nn.Linear(512 * block.expansion, num_classes)def _make_layer(self, block, planes, num_blocks, stride):strides = [stride] + [1]*(num_blocks-1)layers = []for stride in strides:layers.append(block(self.in_planes, planes, stride))self.in_planes = planes * block.expansionreturn nn.Sequential(*layers)def forward(self, x):x = F.relu(self.bn1(self.conv1(x)))x = self.layer1(x)x = self.layer2(x)x = self.layer3(x)x = self.layer4(x)x = self.avgpool(x)x = torch.flatten(x, 1)out = self.linear(x)return outdef SE_ResNet18():return SE_ResNet(BasicBlock, [2, 2, 2, 2])def SE_ResNet34():return SE_ResNet(BasicBlock, [3, 4, 6, 3])def SE_ResNet50():return SE_ResNet(Bottleneck, [3, 4, 6, 3])def SE_ResNet101():return SE_ResNet(Bottleneck, [3, 4, 23, 3])def SE_ResNet152():return SE_ResNet(Bottleneck, [3, 8, 36, 3])'''if __name__ == '__main__':model = SE_ResNet50()print(model)input = torch.randn(1, 3, 224, 224)out = model(input)print(out.shape)# test()'''if __name__ == '__main__':net = SE_ResNet50().cuda()summary(net, (3, 224, 224))

本篇就结束了，欢迎大家留言讨论呀！