CIFAR10数据集实战-ResNet网络构建（中）

用户6719124

发布于 2020-02-24 18:06:55

6690

发布于 2020-02-24 18:06:55

再定义一个ResNet网络

我们本次准备构建ResNet-18层结构

class ResNet(nn.Module):

    def __init__(self):
        super(ResNet, self).__init__()

        self.conv1 = nn.Sequential(
            nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1),
            nn.BatchNorm2d(64)
        )
        # 紧跟着要进行四次这样的单元
        # 构建辅助函数，使[b, 64, h, w] => [b, 128, h, w]
        self.blk1 = ResBlk(64, 128)
        # 构建辅助函数，使[b, 128, h, w] = > [b, 256, h, w]
        self.blk2 = ResBlk(128, 256)
        # 构建辅助函数，使[b, 256, h, w] = > [b, 512, h, w]
        self.blk3 = ResBlk(256, 512)
        # 构建辅助函数，使[b, 512, h, w] = > [b, 1024, h, w]
        self.blk4 = ResBlk(512, 1024)

接下来构建ResNet-18的forward函数

def forward(self, x):
    x = F.relu(self.conv1(x))
    # [b, 64, h, w] => [b, 1024, h, w]
    x = self.blk1(x)
    x = self.blk2(x)
    x = self.blk3(x)
    x = self.blk4(x)

由于我们要进行10分类问题，要将添加代码

self.outlayer = nn.Linear(1024, 10)

和

x = self.outlayer(x)
return x

为确定具体维度大小，我们先构建假数据

def main():
    blk = ResBlk(64, 128)
    tmp = torch.randn(2, 3, 32, 32)
    out = blk(tmp)
    print(out.shape)

if __name__ == "__main__":
    main()

此时代码为

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

class ResBlk(nn.Module):
    # 与上节一样，同样resnet的block单元，继承nn模块
    def __init__(self, ch_in, ch_out):
        super(ResBlk, self).__init__()
        # 完成初始化

        self.conv1 = nn.Conv2d(ch_in, ch_out, kernel_size=3, stride=1, padding=1)
        self.bn1 = nn.BatchNorm2d(ch_out)
        # 进行正则化处理,以使train过程更快更稳定
        self.conv2 = nn.Conv2d(ch_out, ch_out, kernel_size=3, stride=1, padding=1)
        self.bn2 = nn.BatchNorm2d(ch_out)

        self.extra = nn.Sequential()

        if ch_out != ch_in:
            self.extra = nn.Sequential(
                nn.Conv2d(ch_in, ch_out, kernel_size=1, stride=1),
                nn.BatchNorm2d(ch_out),
            )



    def forward(self, x):
        # 这里输入的是[b, ch, h, w]
        out = F.relu(self.bn1(self.conv1(x)))
        out = F.relu(self.bn2(self.conv2(out)))


        out = self.extra(x) + out
        # 这便是element.wise add，实现了[b, ch_in, h, w] 和 [b, ch_out, h, w]两个的相加

        return out


class ResNet(nn.Module):

    def __init__(self):
        super(ResNet, self).__init__()

        self.conv1 = nn.Sequential(
            nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1),
            nn.BatchNorm2d(64)
        )
        # 紧跟着要进行四次这样的单元
        # 构建辅助函数，使[b, 64, h, w] => [b, 128, h, w]
        self.blk1 = ResBlk(64, 128)
        # 构建辅助函数，使[b, 128, h, w] = > [b, 256, h, w]
        self.blk2 = ResBlk(128, 256)
        # 构建辅助函数，使[b, 256, h, w] = > [b, 512, h, w]
        self.blk3 = ResBlk(256, 512)
        # 构建辅助函数，使[b, 512, h, w] = > [b, 1024, h, w]
        self.blk4 = ResBlk(512, 1024)

        self.outlayer = nn.Linear(1024, 10)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        # [b, 64, h, w] => [b, 1024, h, w]
        x = self.blk1(x)
        x = self.blk2(x)
        x = self.blk3(x)
        x = self.blk4(x)

        x = self.outlayer(x)
        return x

def main():
    blk = ResBlk(64, 128)
    tmp = torch.randn(2, 3, 32, 32)
    out = blk(tmp)
    print(out.shape)

if __name__ == "__main__":
    main()

本文参与腾讯云自媒体同步曝光计划，分享自微信公众号。

原始发表：2020-01-21，如有侵权请联系 cloudcommunity@tencent.com 删除

函数