完整项目代码：https://github.com/SPECTRELWF/pytorch-cnn-study

ResNet网络结构

ResNet是何恺明大神在CVPR2016的工作，也拿到了当年的最佳论文。是为了解决深层网络的梯度消失的问题，引入了残差块连接。
论文地址：https://arxiv.org/pdf/1512.03385.pdf
其实我自己刚开始看这篇文章的时候不是很明白，有很多细节并不能很清楚，比如怎么去实现shortcut，每个阶段的输出到下一阶段时候的特征通道并不一致，在这里推荐大神李沐在哔哩哔哩的讲解，讲解得非常牛逼。
ResNet34的网络结构图：

图是来自论文中的，Pytorch中可以很方便的使用自带的model模块实现ResNet，但是有些细节是看不明白的，还是得自己去实现才能了解。

这是不同的ResNet的具体细节，本代码完全按照上面的细节来搭建的。

可以使用Pytorch中的torchvision模块的方法去初始化一个网络，查看一下网络结构，照着实现。代码如下：

# !/usr/bin/python3
# -*- coding:utf-8 -*-
# Author:WeiFeng Liu
# @Time: 2021/11/5 下午3:55
import torch
import torchvisionnet = torchvision.models.resnet34()
print(net)

会输出一个标准的网络信息，我在复现这些基本的backbone的时候喜欢这么做，直接看着代码敲可能会少一些思考和对细节的理解，得到的结果如下：

ResNet((conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=True)(maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)(layer1): Sequential((0): BasicBlock((conv1): Conv2d(64, 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)(relu): ReLU(inplace=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))(1): BasicBlock((conv1): Conv2d(64, 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)(relu): ReLU(inplace=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))(2): BasicBlock((conv1): Conv2d(64, 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)(relu): ReLU(inplace=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)))(layer2): Sequential((0): BasicBlock((conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=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)(downsample): Sequential((0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(1): BasicBlock((conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=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))(2): BasicBlock((conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=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))(3): BasicBlock((conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=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)))(layer3): Sequential((0): BasicBlock((conv1): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=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)(downsample): Sequential((0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(1): BasicBlock((conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=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))(2): BasicBlock((conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=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))(3): BasicBlock((conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=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))(4): BasicBlock((conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=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))(5): BasicBlock((conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=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)))(layer4): Sequential((0): BasicBlock((conv1): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=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)(downsample): 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): BasicBlock((conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=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))(2): BasicBlock((conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=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)))(avgpool): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Linear(in_features=512, out_features=1000, bias=True)
)Process finished with exit code 0

其实只能得到具体每个块和每层的具体参数，怎么去实现还是的自己去思考，可能我太菜了，搞了一早上。

数据集描述

数据集使用的网上一个公开的汽车数据集，包含十个类别的汽车。

包含大巴，面包车等一些类别。需要数据集的私聊我发你百度链接。

网络结构

# !/usr/bin/python3
# -*- coding:utf-8 -*-
# Author:WeiFeng Liu
# @Time: 2021/11/16 上午10:40import torch
import torchvision
import torch.nn as nn
import torch.nn.functional as Fclass BasicBlock(nn.Module):def __init__(self, in_channels,out_channels,stride=(1,1),downsample=None):super(BasicBlock, self).__init__()self.Block = nn.Sequential(nn.Conv2d(in_channels, out_channels,kernel_size=3,stride=stride,padding=1),nn.BatchNorm2d(out_channels),nn.ReLU(inplace=True),nn.Conv2d(out_channels,out_channels,kernel_size=3,stride=1,padding=1),nn.BatchNorm2d(out_channels),)self.downsample = downsampledef forward(self,x):out = self.Block(x)if self.downsample is not None:residual = self.downsample(x)else:residual = xout += residualreturn outclass ResNet34(nn.Module):def __init__(self,num_classes=1000):super(ResNet34, self).__init__()self.head = nn.Sequential(nn.Conv2d(3,64,kernel_size=7,stride=2,padding=3),nn.BatchNorm2d(64),nn.ReLU(inplace=True),nn.MaxPool2d(kernel_size=3,stride=2,padding=1))self.layer1 = self.make_layer(64,64,3)self.layer2 = self.make_layer(64,128,4,strides=2)self.layer3 = self.make_layer(128,256,6,strides=2)self.layer4 = self.make_layer(256,512,3,strides=2)self.avg_pool = nn.AdaptiveAvgPool2d(1)self.fc1 = nn.Linear(512,256)self.fc2 = nn.Linear(256,num_classes)def make_layer(self,in_channels,out_channels,block_num,strides=1):downsample = nn.Sequential(nn.Conv2d(in_channels, out_channels, kernel_size=1,stride=strides,bias=False),nn.BatchNorm2d(out_channels),)layer = []layer.append(BasicBlock(in_channels,out_channels,strides,downsample))for i in range(1,block_num):layer.append(BasicBlock(out_channels,out_channels))return nn.Sequential(* layer)def forward(self,x):x = self.head(x)x = self.layer1(x)x = self.layer2(x)x = self.layer3(x)x = self.layer4(x)x = self.avg_pool(x)x = x.view(x.size(0),-1)x = self.fc1(x)x = self.fc2(x)return x

train

使用Adam优化器来训练的网络

# !/usr/bin/python3
# -*- coding:utf-8 -*-
# Author:WeiFeng Liu
# @Time: 2021/11/5 下午4:37import torch
from torch.utils.data import DataLoader
from torchvision.transforms import transforms as transforms
import torch.optim as optim
from dataload.car_dataload import CAR_DATASET
from resnet34 import ResNet34
import torch.nn as nn
from utils import plot_curvedevice = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
epochs = 200
batch_size = 32
lr = 0.001
transform = transforms.Compose([transforms.Resize([224, 224]),transforms.ToTensor(),transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5]),
])
train_dataset = CAR_DATASET(r'dataset/train', transform=transform)
train_loader = DataLoader(train_dataset,batch_size=batch_size,shuffle=True,
)model = ResNet34(10).to(device)
opt = optim.Adam(model.parameters(),lr=lr,betas=(0.9,0.999))
cri = nn.CrossEntropyLoss()train_loss = []
for epoch in range(epochs):sum_loss = 0for batch_idx, (x, y) in enumerate(train_loader):x = x.to(device)y = y.to(device)pred = model(x)opt.zero_grad()loss = cri(pred, y)loss.backward()opt.step()train_loss.append(loss.item())print('[epoch : %d  ,batch : %d  ,loss : %.3f]' %(epoch,batch_idx,loss.item()))torch.save(model.state_dict(), 'model/model.pth')
plot_curve(train_loss)

但是在测试集上也没有特别好，也就是0.8这个样子的正确率。

Pytorch：手撕ResNet34实现汽车分类相关推荐

手撕python_GitHub - caishiqing/manual: 手撕机器学习
手撕机器学习用腻了开源框架,尝试下手撕机器学习模型?写这个手撕机器学习系列,旨在不使用任何开源框架的条件下手推实现各种模型,同时保证高性能. Requirements 适用于python2.7与py ...
手撕 CNN 经典网络之 VGGNet（PyTorch实战篇）
大家好,我是红色石头! 在上一篇文章: 手撕 CNN 经典网络之 VGGNet(理论篇) 详细介绍了 VGGNet 的网络结构,今天我们将使用 PyTorch 来复现VGGNet网络,并用VGGNet ...
手撕 CNN 之 AlexNet（PyTorch 实战篇）
大家好,我是红色石头! 在上一篇文章: 手撕 CNN 经典网络之 AlexNet(理论篇) 详细介绍了 AlexNet 的网络结构,今天我们将使用 PyTorch 来复现AlexNet网络,并用Ale ...
关键点检测之直接回归(逐行手撕Pytorch)
直接回归的整个流程如下: 用卷积提特征用全连接层进行回归得到关键点与标签做损失并优化损失当图像很大的时候,神经网络是在整张图像中去寻找一个像素点,所以直接回归这一解决办法让神经网络的工作是非常难 ...
手撕 CNN 经典网络之 VGGNet（理论篇）
2014年,牛津大学计算机视觉组(Visual Geometry Group)和Google DeepMind公司一起研发了新的卷积神经网络,并命名为VGGNet.VGGNet是比AlexNet更深的 ...
深度学习之手撕深度神经网络DNN代码（基于numpy）
声明 1)本文仅供学术交流,非商用.所以每一部分具体的参考资料并没有详细对应.如果某部分不小心侵犯了大家的利益,还望海涵,并联系博主删除. 2)博主才疏学浅,文中如有不当之处,请各位指出,共同进步,谢 ...
手撕图机器学习，图神经网络
手撕图机器学习,图神经网络写在前面 & 配套链接(访者必读) 图的基本表示图的基本参数图的类别节点连接数(Node degree) 图的矩阵表示(邻接矩阵) 连接列表和邻接列表其他图 ...
predict函数 R_RROC三剑客（一）使用R语言手撕ROC曲线
之前因工作需要绘制ROC曲线,所以对该曲线的计算细节进行了一番摸索.刚开始我搜索ROC曲线一般跟机器学习相关联,导致我对它的概念有了曲解,理所当然地以为它只是一个用于机器学习的分类器评估标准,所以在绘 ...
机器学习算法之手撕SVM-线性(理论)
感谢Jack-Cui大佬的知识分享机器学习专栏点击这里目录感谢Jack-Cui大佬的知识分享 0. 什么是SVM? 概述 1. 线性SVM 1.1 相关概念:决策面,分类间隔,最优决策面(最优解 ...

Pytorch：手撕ResNet34实现汽车分类

ResNet网络结构

数据集描述

网络结构

train

Pytorch：手撕ResNet34实现汽车分类相关推荐

最新文章

热门文章