resnet代碼分析

 

1.

先導入使用的包，並聲明可用的網絡和預訓練好的模型

import torch.nn as nn
import torch.utils.model_zoo as model_zoo

#聲明可調用的網絡
__all__ = ['ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101',
           'resnet152']

#用於加載的預訓練好的模型
model_urls = {
    'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth',
    'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth',
    'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth',
    'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth',
    'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth',
}

 

 2.

定義要使用到的1*1和3*3的卷積層

#卷積核為3*3，padding=1，stride=1(默認,根據實際傳入參數設定),dilation=1,groups=1,bias=False的二維卷積
def conv3x3(in_planes, out_planes, stride=1):
    """3x3 convolution with padding"""
    return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,
                     padding=1, bias=False)

#卷積核為1*1，padding=1，stride=1(默認,根據實際傳入參數設定),dilation=1,groups=1,bias=False的二維卷積
def conv1x1(in_planes, out_planes, stride=1):
    """1x1 convolution"""
    return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)

注意：這里bias設置為False,原因是：

下面使用了Batch Normalization，而其對隱藏層  有去均值的操作，所以這里的常數項 可以消去

因為Batch Normalization有一個操作，所以上面的數值效果是能由所替代的

因此我們在使用Batch Norm的時候，可以忽略各隱藏層的常數項  。

這樣在使用梯度下降算法時，只用對  ， 和  進行迭代更新

 

3.

實現兩層的殘差塊

比如：

#這個實現的是兩層的殘差塊，用於resnet18/34
class BasicBlock(nn.Module):
    expansion = 1

    def __init__(self, inplanes, planes, stride=1, downsample=None):
        super(BasicBlock, self).__init__()
        self.conv1 = conv3x3(inplanes, planes, stride)
        self.bn1 = nn.BatchNorm2d(planes)
        self.relu = nn.ReLU(inplace=True)
        self.conv2 = conv3x3(planes, planes)
        self.bn2 = nn.BatchNorm2d(planes)
        self.downsample = downsample
        self.stride = stride

    def forward(self, x):
        identity = x

        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu(out)

        out = self.conv2(out)
        out = self.bn2(out)

        if self.downsample is not None: #當連接的維度不同時，使用1*1的卷積核將低維轉成高維，然后才能進行相加
            identity = self.downsample(x)

        out += identity #實現H(x)=F(x)+x或H(x)=F(x)+Wx
        out = self.relu(out)

        return out

 

4.實現3層的殘差塊

如圖：

#這個實現的是三層的殘差塊，用於resnet50/101/152
class Bottleneck(nn.Module):
    expansion = 4

    def __init__(self, inplanes, planes, stride=1, downsample=None):
        super(Bottleneck, self).__init__()
        self.conv1 = conv1x1(inplanes, planes)
        self.bn1 = nn.BatchNorm2d(planes)
        self.conv2 = conv3x3(planes, planes, stride)
        self.bn2 = nn.BatchNorm2d(planes)
        self.conv3 = conv1x1(planes, planes * self.expansion)
        self.bn3 = nn.BatchNorm2d(planes * self.expansion)
        self.relu = nn.ReLU(inplace=True)
        self.downsample = downsample
        self.stride = stride

    def forward(self, x):
        identity = x

        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu(out)

        out = self.conv2(out)
        out = self.bn2(out)
        out = self.relu(out)

        out = self.conv3(out)
        out = self.bn3(out)

        if self.downsample is not None:
            identity = self.downsample(x) #當連接的維度不同時，使用1*1的卷積核將低維轉成高維，然后才能進行相加

        out += identity #實現H(x)=F(x)+x或H(x)=F(x)+Wx
        out = self.relu(out)

        return out

 

5.整個網絡實現

class ResNet(nn.Module):
    #參數block指明殘差塊是兩層或三層，參數layers指明每個卷積層需要的殘差塊數量，num_classes指明分類數，zero_init_residual是否初始化為0
    def __init__(self, block, layers, num_classes=1000, zero_init_residual=False):
        super(ResNet, self).__init__()
        self.inplanes = 64 #一開始先使用64*7*7的卷積核,stride=2, padding=3
        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3,
                               bias=False) #3通道的輸入RGB圖像數據變為64通道的數據
        self.bn1 = nn.BatchNorm2d(64)
        self.relu = nn.ReLU(inplace=True) #以上是第一層卷積--1
        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) #然后進行最大值池化操作--2
        self.layer1 = self._make_layer(block, 64, layers[0])#下面就是所有的卷積層的設置--3
        self.layer2 = self._make_layer(block, 128, layers[1], stride=2)
        self.layer3 = self._make_layer(block, 256, layers[2], stride=2)
        self.layer4 = self._make_layer(block, 512, layers[3], stride=2)
        self.avgpool = nn.AdaptiveAvgPool2d((1, 1)) #進行自適應平均池化--4
        self.fc = nn.Linear(512 * block.expansion, num_classes)#全連接層--5

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                #kaiming高斯初始化，目的是使得Conv2d卷積層反向傳播的輸出的方差都為1
                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
            elif isinstance(m, nn.BatchNorm2d):
                #初始化m.weight，即gamma的值為1；m.bias即beta的值為0
                nn.init.constant_(m.weight, 1)
                nn.init.constant_(m.bias, 0)

        # 在每個殘差分支中初始化最后一個BN，即BatchNorm2d
        # 以便殘差分支以零開始，並且每個殘差塊的行為類似於一個恆等式。
        # This improves the model by 0.2~0.3% according to https://arxiv.org/abs/1706.02677
        if zero_init_residual:
            for m in self.modules():
                if isinstance(m, Bottleneck):#Bottleneck的最后一個BN是m.bn3
                    nn.init.constant_(m.bn3.weight, 0)
                elif isinstance(m, BasicBlock):#BasicBlock的最后一個BN是m.bn2
                    nn.init.constant_(m.bn2.weight, 0)

    #實現一層卷積，block參數指定是兩層殘差塊或三層殘差塊，planes參數為輸入的channel數，blocks說明該卷積有幾個殘差塊
    def _make_layer(self, block, planes, blocks, stride=1):
        downsample = None
        #即如果該層的輸入的channel數inplanes和其輸出的channel數planes * block.expansion不同，
        #那要使用1*1的卷積核將輸入x低維轉成高維，然后才能進行相加
        if stride != 1 or self.inplanes != planes * block.expansion:
            downsample = nn.Sequential(
                conv1x1(self.inplanes, planes * block.expansion, stride),
                nn.BatchNorm2d(planes * block.expansion),
            )

        layers = []
        #只有卷積和卷積直接的連接需要低維轉高維
        layers.append(block(self.inplanes, planes, stride, downsample))
        self.inplanes = planes * block.expansion
        for _ in range(1, blocks):
            layers.append(block(self.inplanes, planes))

        return nn.Sequential(*layers)

    def forward(self, x):
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.relu(x)
        x = self.maxpool(x)

        x = self.layer1(x)
        x = self.layer2(x)
        x = self.layer3(x)
        x = self.layer4(x)

        x = self.avgpool(x)
        x = x.view(x.size(0), -1)
        x = self.fc(x)

        return x

 

 6.不同層次網絡實現

 

#18層的resnet
def resnet18(pretrained=False, **kwargs):
    """Constructs a ResNet-18 model.

    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
    """
    model = ResNet(BasicBlock, [2, 2, 2, 2], **kwargs)
    if pretrained:#是否使用已經訓練好的預訓練模型，在此基礎上繼續訓練
        model.load_state_dict(model_zoo.load_url(model_urls['resnet18']))
    return model

#34層的resnet
def resnet34(pretrained=False, **kwargs):
    """Constructs a ResNet-34 model.

    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
    """
    model = ResNet(BasicBlock, [3, 4, 6, 3], **kwargs)
    if pretrained:#是否使用已經訓練好的預訓練模型，在此基礎上繼續訓練
        model.load_state_dict(model_zoo.load_url(model_urls['resnet34']))
    return model

#50層的resnet
def resnet50(pretrained=False, **kwargs):
    """Constructs a ResNet-50 model.

    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
    """
    model = ResNet(Bottleneck, [3, 4, 6, 3], **kwargs)
    if pretrained:#是否使用已經訓練好的預訓練模型，在此基礎上繼續訓練
        model.load_state_dict(model_zoo.load_url(model_urls['resnet50']))
    return model

#101層的resnet
def resnet101(pretrained=False, **kwargs):
    """Constructs a ResNet-101 model.

    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
    """
    model = ResNet(Bottleneck, [3, 4, 23, 3], **kwargs)
    if pretrained:#是否使用已經訓練好的預訓練模型，在此基礎上繼續訓練
        model.load_state_dict(model_zoo.load_url(model_urls['resnet101']))
    return model

#152層的resnet
def resnet152(pretrained=False, **kwargs):
    """Constructs a ResNet-152 model.

    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
    """
    model = ResNet(Bottleneck, [3, 8, 36, 3], **kwargs)
    if pretrained:#是否使用已經訓練好的預訓練模型，在此基礎上繼續訓練
        model.load_state_dict(model_zoo.load_url(model_urls['resnet152']))
    return model