Resnet解读和实战

Resnet解读和实战

动机（灵感来源）

1.增加网络的层数之后，训练误差往往不降反升。这是因为梯度消失或是梯度爆炸导致的。

2.假设现有一个比较浅的网络已达到了饱和的准确率，这时在它后面再加上几个恒等映射层（Identity mapping，也即y=x，输出等于输入），这样就增加了网络的深度，并且起码误差不会增加，也即更深的网络不应该带来训练集上误差的上升。

3.输入是x，期望输出是H(x)，即H(x)是期望的复杂潜在映射，如果是要学习这样的模型，则训练难度会比较大；如果已经学习到较饱和的准确率（或者当发现下层的误差变大时），那么接下来的学习目标就转变为恒等映射的学习，也就是使输入x近似于输出H(x)，以保持在后面的层次中不会造成精度下降。

4.通过"shortcut connections（捷径连接）"的方式，直接把输入x传到输出作为初始结果，输出结果为H(x)=F(x)+x，当F(x)=0时，那么H(x)=x，也就是上面所提到的恒等映射。于是，ResNet相当于将学习目标改变了，不再是学习一个完整的输出，而是目标值H(X)和x的差值，也就是所谓的残差F(x) := H(x)-x，因此，后面的训练目标就是要将残差结果逼近于0，使到随着网络加深，准确率不下降。

残差结构的优点

①残差结构比传统结构更好优化

提升了对梯度与损失的相关性，进而使得网络的学习能力增强，解决退化问题。

比如把5映射到5.1，那么引入残差前是F'(5)=H(5)=5.1，引入残差后是H(5)=F(5)+5=5.1, F(5)=0.1。这里的F'和F都表示网络参数映射，引入残差后的映射对输出的变化更敏感。比如输出从5.1变到5.2，映射F'的输出增加了1/51=2%，而对于残差结构输出从5.1到5.2，映射F是从0.1到0.2，增加了100%。明显后者输出变化对权重的调整作用更大，所以效果更好。训练残差的思想都是去掉相同的主体部分，从而突出微小的变化，残差网络可以看做是差分放大器。

②缓解梯度消失问题

具体方法

• 输出y＝F(x)＋x。这时不是直接学习输出特征y的表示，而是学习y－x。
• 如果想学习出原模型的表示，只需将F(x)的参数 全部设置为0，则是恒等映射。

原文：The identity shortcuts (Eqn.(1)) can be directly used when the input and output are of the same dimensions (solid line shortcuts in Fig.3). When the dimensions increase (dotted line shortcuts in Fig.3), we consider two options: (A) The shortcut still performs identity mapping, with extra zero entries padded for increasing dimensions. This option introduces no extra parameter; (B) The projection shortcut in Eqn.(2) is used to match dimensions (done by 1×1 convolutions).

网络结构

右图使用11卷积先将输入进行降维，然后再经过33卷积后再用11卷积进行升维。使用11卷积能大大降低参数量计算量。首先看上图右边的bottleneck结构，对于256维的输入特征，参数数目：1x1x256x64+3x3x64x64+1x1x64x256=69632，如果同样的输入输出维度但不使用1x1卷积，而使用两个3x3卷积的话，参数数目为(3x3x256x256)x2=1179648。

Pytorch官方文档Resnet代码解读

import torch.nn as nn
import torch.utils.model_zoo as model_zoo#
 
 
__all__ = ['ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101',
           'resnet152']
 
 
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',
}
 
 
def conv3x3(in_planes, out_planes, stride=1):
    """3x3 convolution with padding"""#卷积后尺寸不变
    return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,
                     padding=1, bias=False)

#浅层resnet网络结构，例如resnet18等
class BasicBlock(nn.Module):
#3*3->3*3
    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):
        residual = 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:#下采样结构，虚线
            residual = self.downsample(x)
 
        out += residual
        out = self.relu(out)
 
        return out

#定义深层resnet网络残差结构，例如resnet50
class Bottleneck(nn.Module):
    expansion = 4
 
    def __init__(self, inplanes, planes, stride=1, downsample=None):
#3*3->1*1->3*3
        super(Bottleneck, self).__init__()
        self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=1, bias=False)
		self.bn1 = nn.BatchNorm2d(planes)
        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride,                        padding=1, bias=False)
        self.bn2 = nn.BatchNorm2d(planes)
        self.conv3 = nn.Conv2d(planes, planes*self.expansion,kernel_size=1,bias=False)
        self.bn3 = nn.BatchNorm2d(planes * self.expansion)
        self.relu = nn.ReLU(inplace=True)
        self.downsample = downsample
        self.stride = stride
 
    def forward(self, x):
        residual = 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:
            residual = self.downsample(x)
 
        out += residual
        out = self.relu(out)
 
        return out

#resnet网络框架
class ResNet(nn.Module):
 
    def __init__(self, block, layers, num_classes=1000):#block残差块类型
        self.inplanes = 64
        super(ResNet, self).__init__()
        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3,
                               bias=False)
        self.bn1 = nn.BatchNorm2d(64)
        self.relu = nn.ReLU(inplace=True)
        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
        self.layer1 = self._make_layer(block, 64, layers[0])
        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.AvgPool2d(7, stride=1)
        self.fc = nn.Linear(512 * block.expansion, num_classes)
 
        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
            elif isinstance(m, nn.BatchNorm2d):
                nn.init.constant_(m.weight, 1)
                nn.init.constant_(m.bias, 0)
 
    def _make_layer(self, block, planes, blocks, stride=1):
        downsample = None
        if stride != 1 or self.inplanes != planes * block.expansion:#进行下采样的条
#件，stride！=1满足网络普遍结构下采样条件，self.inplanes != planes * block.expansion
#满足深层网络64->256下采样的条件
          downsample = nn.Sequential(
                nn.Conv2d(self.inplanes, planes * block.expansion,
                          kernel_size=1, stride=stride, bias=False),
                nn.BatchNorm2d(planes * block.expansion),
            )
 
        layers = []
        layers.append(block(self.inplanes, planes, stride, downsample))
        self.inplanes = planes * block.expansion
        for i in range(1, blocks):#先搭建了一个残差结构，再搭建bloks-1个结构
            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

def resnet18(pretrained=False, **kwargs):
#**允许你将不定长度的键值对作为参数传递给一个函数。如果你想要在一个函数里处理带名字的参数，你应该使用**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:
        model.load_state_dict(model_zoo.load_url(model_urls['resnet18']))
    return model

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:
        model.load_state_dict(model_zoo.load_url(model_urls['resnet50']))
    return model

Resnet实战

import torch
import torch.nn as nn
from torch.nn import functional as F


class RestNetBasicBlock(nn.Module):
#实线模块
    def __init__(self, in_channels, out_channels, stride):
        super(RestNetBasicBlock, self).__init__()
        self.conv1 = nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=stride, padding=1)
        self.bn1 = nn.BatchNorm2d(out_channels)
        self.conv2 = nn.Conv2d(out_channels, out_channels, kernel_size=3, stride=stride, padding=1)
        self.bn2 = nn.BatchNorm2d(out_channels)

    def forward(self, x):
        output = self.conv1(x)
        output = F.relu(self.bn1(output))
        output = self.conv2(output)
        output = self.bn2(output)
        return F.relu(x + output)

class RestNetDownBlock(nn.Module):
#虚线，shortcuts部分快捷连接
    def __init__(self, in_channels, out_channels, stride):
        super(RestNetDownBlock, self).__init__()
        self.conv1 = nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=stride[0], padding=1)
        self.bn1 = nn.BatchNorm2d(out_channels)
        self.conv2 = nn.Conv2d(out_channels, out_channels, kernel_size=3, stride=stride[1], padding=1)
        self.bn2 = nn.BatchNorm2d(out_channels)
        self.extra = nn.Sequential(
            nn.Conv2d(in_channels, out_channels, kernel_size=1, stride=stride[0], padding=0),
            nn.BatchNorm2d(out_channels)
        )

    def forward(self, x):
        extra_x = self.extra(x)
        output = self.conv1(x)
        out = F.relu(self.bn1(output))

        out = self.conv2(out)
        out = self.bn2(out)
        return F.relu(extra_x + out)

class RestNet18(nn.Module):
    def __init__(self):
        super(RestNet18, self).__init__()
        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3)
        self.bn1 = nn.BatchNorm2d(64)
        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)

        self.layer1 = nn.Sequential(RestNetBasicBlock(64, 64, 1),
                                    RestNetBasicBlock(64, 64, 1))

        self.layer2 = nn.Sequential(RestNetDownBlock(64, 128, [2, 1]),
                                    RestNetBasicBlock(128, 128, 1))

        self.layer3 = nn.Sequential(RestNetDownBlock(128, 256, [2, 1]),
                                    RestNetBasicBlock(256, 256, 1))

        self.layer4 = nn.Sequential(RestNetDownBlock(256, 512, [2, 1]),
                                    RestNetBasicBlock(512, 512, 1))

        self.avgpool = nn.AdaptiveAvgPool2d(output_size=(1, 1))

        self.fc = nn.Linear(512, 10)

    def forward(self, x):
        out = self.conv1(x)
        out = self.layer1(out)
        out = self.layer2(out)
        out = self.layer3(out)
        out = self.layer4(out)
        out = self.avgpool(out)
        out = out.reshape(x.shape[0], -1)
        out = self.fc(out)
        return out

import torch
from torch import nn, optim
import torchvision.transforms as transforms
from torchvision import datasets
from torch.utils.data import DataLoader
from shiyan import RestNet18


#  用CIFAR-10 数据集进行训练

def main():
    batchsz = 128
#datasets.CIFAR10(root=data_path, train=True, download=False, transform=train_transform)
    cifar_train = datasets.CIFAR10('cifar', True, transform=transforms.Compose([
        transforms.Resize((32, 32)),
        transforms.ToTensor(),
        transforms.Normalize(mean=[0.485, 0.456, 0.406],
                             std=[0.229, 0.224, 0.225])
    ]), download=True)
    cifar_train = DataLoader(cifar_train, batch_size=batchsz, shuffle=True)

    cifar_test = datasets.CIFAR10('cifar', False, transform=transforms.Compose([
        transforms.Resize((32, 32)),
        transforms.ToTensor(),
        transforms.Normalize(mean=[0.485, 0.456, 0.406],
                             std=[0.229, 0.224, 0.225])
    ]), download=True)
    cifar_test = DataLoader(cifar_test, batch_size=batchsz, shuffle=True)

    x, label = iter(cifar_train).next()
    #list、tuple等都是可迭代对象，我们可以通过iter()函数获取这些可迭代对象的迭代器。然后我们可以对获取到的迭代器不断使⽤next()函数来获取下⼀条数据。iter()函数实际上就是调⽤了可迭代对象的 __ iter __ ⽅法
    print('x:', x.shape, 'label:', label.shape)

    device = torch.device('cuda')
    # model = Lenet5().to(device)
    model = RestNet18().to(device)

    criteon = nn.CrossEntropyLoss().to(device)
    optimizer = optim.Adam(model.parameters(), lr=1e-3)
    print(model)

    for epoch in range(254):

        model.train()
        for batchidx, (x, label) in enumerate(cifar_train):
            # [b, 3, 32, 32]
            # [b]
            x, label = x.to(device), label.to(device)

            logits = model(x)
            # logits: [b, 10]
            # label:  [b]
            # loss: tensor scalar
            loss = criteon(logits, label)#算损失

            # backprop
            optimizer.zero_grad()
            loss.backward()
            optimizer.step()
#loss：tensor(2.8426, device='cuda:0', grad_fn=<NllLossBackward>)
        print(epoch, 'loss:', loss.item())

        model.eval()#在 model(test) 之前，需要加上model.eval()，否则只要有输入数据，即使不训练，model 也会改变权值
        with torch.no_grad():
        #为验证集的时候，我们只是想看一下训练的效果，并不是想通过验证集来更新网络时，就可以使用with torch.no_grad()。
            # test
            total_correct = 0
            total_num = 0
            for x, label in cifar_test:
                # [b, 3, 32, 32]
                # [b]
                x, label = x.to(device), label.to(device)
               
                # [b, 10]
                logits = model(x)
                # [b]
                pred = logits.argmax(dim=1)
                # [b] vs [b] => scalar tensor
                correct = torch.eq(pred, label).float().sum().item()
                total_correct += correct
                total_num += x.size(0)
                # print(correct)

            acc = total_correct / total_num
            print(epoch, 'test acc:', acc)


if __name__ == '__main__':
    main()

结果