深度学习基础--ResNet50V2网络的讲解，ResNet50V2的复现(pytorch)以及用复现的ResNet50做鸟类图像分类

• 🍨 本文为🔗365天深度学习训练营 中的学习记录博客
• 🍖 原作者：K同学啊

前言

• 如果说最经典的神经网络，ResNet肯定是一个，从ResNet发布后，作者又进行修改，命名为ResNe50v2，这篇文章是本人学习ResNe50v2的学习笔记，并且用pytorch复现了ResNet50V2，后面用它做了一个鸟类图像分类demo，与上一篇ResNet50相比，效果明显好了不少。
• ResNet讲解： https://blog.csdn.net/weixin_74085818/article/details/145786990?spm=1001.2014.3001.5501
• 欢迎收藏 + 关注，本人将会持续更新

文章目录

• 1、简介
• • 与ResNet对比
  • 不同残差结构
  • 激活函数的尝试
  • 小结
• 2、ResNet50V2搭建
• • 1、导入数据
  • • 1、导入库
    • 2、查看数据信息和导入数据
    • 3、展示数据
    • 4、数据导入
    • 5、数据划分
    • 6、动态加载数据
  • 2、构建ResNet-50V2网络
  • 3、模型训练
  • • 1、构建训练集
    • 2、构建测试集
    • 3、设置超参数
  • 4、模型训练
  • 5、结果可视化

1、简介

与ResNet对比

👀 改进点:

• 原始resnet结果：先进行卷积，在进行BN和激活函数，最后执行addtion与RelU
• 修改版本：先进行BN和激活函数，把addtion后的ReLU放到了残差内部，改进后残差内有两个ReLU

不同残差结构

何凯明大神产实力不同的残差结构，如下：

最后结果：

发现还是原始的残差结构效果最好

激活函数的尝试

这个部分主要是激活函数、BN层的位置。

结果：

发现最好的是**(e)**结果

小结

通过学习，发现可以从两个角度修改模型：

• 激活函数、BN层的位置，如：数据处理中的位置，不同位置效果也不同。
• 残差结构：原始版本是恒等映射，但是也有可能不同的残差也会有不同的效果。

2、ResNet50V2搭建

1、导入数据

1、导入库

import torch  
import torch.nn as nn
import torchvision 
import numpy as np 
import os, PIL, pathlib 

# 设置设备
device = "cuda" if torch.cuda.is_available() else "cpu"

device 

'cuda'

2、查看数据信息和导入数据

数据目录有两个文件：一个数据文件，一个权重。

data_dir = "./data/"

data_dir = pathlib.Path(data_dir)

# 类别数量
classnames = [str(path).split("\\")[0] for path in os.listdir(data_dir)]

classnames

['bird_photos', 'resnet50_weights_tf_dim_ordering_tf_kernels.h5']

3、展示数据

import matplotlib.pylab as plt  
from PIL import Image 

# 获取文件名称
data_path_name = "./data/bird_photos/Bananaquit/"
data_path_list = [f for f in os.listdir(data_path_name) if f.endswith(('jpg', 'png'))]

# 创建画板
fig, axes = plt.subplots(2, 8, figsize=(16, 6))

for ax, img_file in zip(axes.flat, data_path_list):
    path_name = os.path.join(data_path_name, img_file)
    img = Image.open(path_name) # 打开
    # 显示
    ax.imshow(img)
    ax.axis('off')
    
plt.show()

​

​

4、数据导入

from torchvision import transforms, datasets 

# 数据统一格式
img_height = 224
img_width = 224 

data_tranforms = transforms.Compose([
    transforms.Resize([img_height, img_width]),
    transforms.ToTensor(),
    transforms.Normalize(   # 归一化
        mean=[0.485, 0.456, 0.406],
        std=[0.229, 0.224, 0.225] 
    )
])

# 加载所有数据
total_data = datasets.ImageFolder(root="./data/", transform=data_tranforms)

5、数据划分

# 大小 8 : 2
train_size = int(len(total_data) * 0.8)
test_size = len(total_data) - train_size 

train_data, test_data = torch.utils.data.random_split(total_data, [train_size, test_size])

6、动态加载数据

batch_size = 32 

train_dl = torch.utils.data.DataLoader(
    train_data,
    batch_size=batch_size,
    shuffle=True
)

test_dl = torch.utils.data.DataLoader(
    test_data,
    batch_size=batch_size,
    shuffle=False
)

# 查看数据维度
for data, labels in train_dl:
    print("data shape[N, C, H, W]: ", data.shape)
    print("labels: ", labels)
    break

data shape[N, C, H, W]:  torch.Size([32, 3, 224, 224])
labels:  tensor([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0])

2、构建ResNet-50V2网络

上一篇文章中，本人搭建的网络有点啰嗦，很多都是一步一步的，但是这一篇，这一种搭建，比较优雅，因为这个利用了三个网络模块很多相同点，依据这个搭建而成。

ResNet50V2搭建方式和ResNet50一样，只是残差堆积不同。

注意：参数，这个神经网络有很多参数，注意别错了。

import torch.nn.functional as F


'''  
conv_shortcut: 采用什么样的残差连接，对应上面图的1、3模块
filters: 输出通道数
卷积核：默认为3
'''
class Block2(nn.Module):
    def __init__(self, in_channel, filters, kernel_size=3, stride=1, conv_shortcut=False):
        super().__init__()
        
        # 第一个，preact，对应上图的前两层，bn、relu
        self.preact = nn.Sequential(
            nn.BatchNorm2d(in_channel),
            nn.ReLU(True)
        )
        
        # 判断是否需要使用残差连接，上图展示的网络中，有3个模块，有两个有残差连接，有一个没有，没有的那一块卷积核为 1
        self.shortcut = conv_shortcut
        if self.shortcut:   # 对应上图的第一块网络结构残差连接
            self.short = nn.Conv2d(in_channel, 4 * filters, kernel_size=1, stride=stride, padding=0, bias=False)  # padding默认为0, 4 * filtersz看源码得出,  输出通道
        else:
            self.short = nn.MaxPool2d(kernel_size=1, stride=stride, padding=0) if stride > 1 else nn.Identity()  # nn.Identity() 对输入的数据X，不做任何操作
        
        # 后面结果，三个模块都一样，我把他分层三个模块
        # 模块一，看源码
        self.conv1 = nn.Sequential(
            nn.Conv2d(in_channel, filters, kernel_size=1, stride=1, bias=False),
            nn.BatchNorm2d(filters),
            nn.ReLU(True)
        )
        
        # 模块二
        self.conv2 = nn.Sequential(
            nn.Conv2d(filters, filters, kernel_size=kernel_size, stride=stride, padding=1, bias=False),
            nn.BatchNorm2d(filters),
            nn.ReLU(True)
        )
        
        # 模块三
        self.conv3 = nn.Conv2d(filters, 4 * filters, kernel_size=1, stride=1)
        
    def forward(self, x):
        # 数据
        x1 = self.preact(x)
        if self.shortcut:  # 这个时候，对应对一个模块
            x2 = self.short(x1)  # 这个时候输入的是 x1
        else:
            x2 = self.short(x)  # 这个对应上面网络图第三个， 用的输入 x 
            
        x1 = self.conv1(x1)
        x1 = self.conv2(x1)
        x1 = self.conv3(x1)
        
        x = x1 + x2  # 合并
        return x
    
# 堆积
class Stack2(nn.Module):
    def __init__(self, in_channel, filters, blocks, stride=2):  # blocks代表上图中最左网络图，残差堆积 中 层数
        super().__init__()
        self.conv = nn.Sequential()
        # 上面网络图中，最左部分，残差堆积是很相似的
        self.conv.add_module(str(0), Block2(in_channel, filters, conv_shortcut=True))   # 参数，名字 + 模块
        # 中间层
        for i in range(1, blocks - 1):  # 上面一层去除，中间剩下 blocks - 2
            self.add_module(str(i), Block2(4 * filters, filters))  # 上一层输出：4 * filters，这一层回归filters
        self.conv.add_module(str(blocks-1), Block2(4 * filters, filters, stride=stride))  # 这里的stride不一样
        
    def forward(self, x):
        x = self.conv(x)
        
        return x
    
class ResNet50V2(nn.Module):
    def __init__(self,
                 include_top=True, # 是否需要包含最定层
                 preact=True,  # 是否需要预激活
                 use_bias=True,  # 卷积层是否用偏置
                 input_shape=[224, 224, 3],
                 classes=1000,  # 类别数量
                 pooling=None
                 ):
        super().__init__()
        
        # 上图神经网络，最左边，最顶层, ZeroPad是感受野参数
        self.conv1 = nn.Sequential() 
        self.conv1.add_module('conv', nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=use_bias))  
        # 这里的标准化，激活函数是可选的
        if not preact:
            self.conv1.add_module('bn', nn.BatchNorm2d(64))
            self.conv1.add_module('relu', nn.ReLU())
        self.conv1.add_module('max_pool', nn.MaxPool2d(kernel_size=3, stride=2, padding=1))
        
        # 上图神经网络，最左边，中间层
        self.conv2 = Stack2(64, 64, 3)
        self.conv3 = Stack2(256, 128, 4)
        self.conv4 = Stack2(512, 256, 6)
        self.conv5 = Stack2(1024, 512, 3, stride=1)  # 这些层数量变换挺有意思的
        
        self.last = nn.Sequential()
        if preact:
            self.last.add_module('bn', nn.BatchNorm2d(2048))
            self.last.add_module('relu', nn.ReLU(True))
        if include_top:
            self.last.add_module('avg_pool', nn.AdaptiveAvgPool2d((1, 1)))
            self.last.add_module('flatten', nn.Flatten())
            self.last.add_module('fc', nn.Linear(2048, classes))
        else:
            if pooling=='avg':
                self.last.add_module('avg_pool', nn.AdaptiveAvgPool2d((1, 1)))
            elif pooling=='max':
                self.last.add_module('max_pool', nn.AdaptiveAMaxPool2d((1, 1)))
        
    def forward(self, x):
        x = self.conv1(x)
        x = self.conv2(x)
        x = self.conv3(x)
        x = self.conv4(x)
        x = self.conv5(x)
        x = self.last(x)
        return x
    
model = ResNet50V2(classes=len(classnames)).to(device)

model

ResNet50V2(
  (conv1): Sequential(
    (conv): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3))
    (max_pool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
  )
  (conv2): Stack2(
    (conv): Sequential(
      (0): Block2(
        (preact): Sequential(
          (0): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU(inplace=True)
        )
        (short): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (conv1): Sequential(
          (0): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
          (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
        )
        (conv2): Sequential(
          (0): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
          (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
        )
        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))
      )
      (2): Block2(
        (preact): Sequential(
          (0): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU(inplace=True)
        )
        (short): MaxPool2d(kernel_size=1, stride=2, padding=0, dilation=1, ceil_mode=False)
        (conv1): Sequential(
          (0): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
          (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
        )
        (conv2): Sequential(
          (0): Conv2d(64, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
          (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
        )
        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))
      )
    )
    (1): Block2(
      (preact): Sequential(
        (0): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
      )
      (short): Identity()
      (conv1): Sequential(
        (0): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): ReLU(inplace=True)
      )
      (conv2): Sequential(
        (0): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): ReLU(inplace=True)
      )
      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (conv3): Stack2(
    (conv): Sequential(
      (0): Block2(
        (preact): Sequential(
          (0): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU(inplace=True)
        )
        (short): Conv2d(256, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (conv1): Sequential(
          (0): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
          (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
        )
        (conv2): Sequential(
          (0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
          (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
        )
        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))
      )
      (3): Block2(
        (preact): Sequential(
          (0): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU(inplace=True)
        )
        (short): MaxPool2d(kernel_size=1, stride=2, padding=0, dilation=1, ceil_mode=False)
        (conv1): Sequential(
          (0): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
          (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
        )
        (conv2): Sequential(
          (0): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
          (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
        )
        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))
      )
    )
    (1): Block2(
      (preact): Sequential(
        (0): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
      )
      (short): Identity()
      (conv1): Sequential(
        (0): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): ReLU(inplace=True)
      )
      (conv2): Sequential(
        (0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): ReLU(inplace=True)
      )
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))
    )
    (2): Block2(
      (preact): Sequential(
        (0): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
      )
      (short): Identity()
      (conv1): Sequential(
        (0): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): ReLU(inplace=True)
      )
      (conv2): Sequential(
        (0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): ReLU(inplace=True)
      )
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (conv4): Stack2(
    (conv): Sequential(
      (0): Block2(
        (preact): Sequential(
          (0): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU(inplace=True)
        )
        (short): Conv2d(512, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (conv1): Sequential(
          (0): Conv2d(512, 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)
          (2): ReLU(inplace=True)
        )
        (conv2): Sequential(
          (0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
          (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
        )
        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))
      )
      (5): Block2(
        (preact): Sequential(
          (0): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU(inplace=True)
        )
        (short): MaxPool2d(kernel_size=1, stride=2, padding=0, dilation=1, ceil_mode=False)
        (conv1): Sequential(
          (0): Conv2d(1024, 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)
          (2): ReLU(inplace=True)
        )
        (conv2): Sequential(
          (0): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
          (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
        )
        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))
      )
    )
    (1): Block2(
      (preact): Sequential(
        (0): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
      )
      (short): Identity()
      (conv1): Sequential(
        (0): Conv2d(1024, 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)
        (2): ReLU(inplace=True)
      )
      (conv2): Sequential(
        (0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): ReLU(inplace=True)
      )
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))
    )
    (2): Block2(
      (preact): Sequential(
        (0): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
      )
      (short): Identity()
      (conv1): Sequential(
        (0): Conv2d(1024, 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)
        (2): ReLU(inplace=True)
      )
      (conv2): Sequential(
        (0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): ReLU(inplace=True)
      )
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))
    )
    (3): Block2(
      (preact): Sequential(
        (0): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
      )
      (short): Identity()
      (conv1): Sequential(
        (0): Conv2d(1024, 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)
        (2): ReLU(inplace=True)
      )
      (conv2): Sequential(
        (0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): ReLU(inplace=True)
      )
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))
    )
    (4): Block2(
      (preact): Sequential(
        (0): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
      )
      (short): Identity()
      (conv1): Sequential(
        (0): Conv2d(1024, 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)
        (2): ReLU(inplace=True)
      )
      (conv2): Sequential(
        (0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): ReLU(inplace=True)
      )
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (conv5): Stack2(
    (conv): Sequential(
      (0): Block2(
        (preact): Sequential(
          (0): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU(inplace=True)
        )
        (short): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (conv1): Sequential(
          (0): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
        )
        (conv2): Sequential(
          (0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
        )
        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1))
      )
      (2): Block2(
        (preact): Sequential(
          (0): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU(inplace=True)
        )
        (short): Identity()
        (conv1): Sequential(
          (0): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
        )
        (conv2): Sequential(
          (0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
        )
        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1))
      )
    )
    (1): Block2(
      (preact): Sequential(
        (0): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
      )
      (short): Identity()
      (conv1): Sequential(
        (0): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): ReLU(inplace=True)
      )
      (conv2): Sequential(
        (0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (2): ReLU(inplace=True)
      )
      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (last): Sequential(
    (bn): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (relu): ReLU(inplace=True)
    (avg_pool): AdaptiveAvgPool2d(output_size=(1, 1))
    (flatten): Flatten(start_dim=1, end_dim=-1)
    (fc): Linear(in_features=2048, out_features=2, bias=True)
  )
)

3、模型训练

1、构建训练集

def train(dataloader, model, loss_fn, optimizer):
    size = len(dataloader.dataset)
    batch_size = len(dataloader)
    
    train_acc, train_loss = 0, 0 
    
    for X, y in dataloader:
        X, y = X.to(device), y.to(device)
        
        # 训练
        pred = model(X)
        loss = loss_fn(pred, y)
        
        # 梯度下降法
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()
        
        # 记录
        train_loss += loss.item()
        train_acc += (pred.argmax(1) == y).type(torch.float).sum().item()
        
    train_acc /= size
    train_loss /= batch_size
    
    return train_acc, train_loss

2、构建测试集

def test(dataloader, model, loss_fn):
    size = len(dataloader.dataset)
    batch_size = len(dataloader)
    
    test_acc, test_loss = 0, 0 
    
    with torch.no_grad():
        for X, y in dataloader:
            X, y = X.to(device), y.to(device)
        
            pred = model(X)
            loss = loss_fn(pred, y)
        
            test_loss += loss.item()
            test_acc += (pred.argmax(1) == y).type(torch.float).sum().item()
        
    test_acc /= size
    test_loss /= batch_size
    
    return test_acc, test_loss

3、设置超参数

loss_fn = nn.CrossEntropyLoss()  # 损失函数     
learn_lr = 1e-4             # 超参数
optimizer = torch.optim.Adam(model.parameters(), lr=learn_lr)   # 优化器

4、模型训练

train_acc = []
train_loss = []
test_acc = []
test_loss = []

epoches = 10

for i in range(epoches):
    model.train()
    epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, optimizer)
    
    model.eval()
    epoch_test_acc, epoch_test_loss = test(test_dl, model, loss_fn)
    
    train_acc.append(epoch_train_acc)
    train_loss.append(epoch_train_loss)
    test_acc.append(epoch_test_acc)
    test_loss.append(epoch_test_loss)
    
    # 输出
    template = ('Epoch:{:2d}, Train_acc:{:.1f}%, Train_loss:{:.3f}, Test_acc:{:.1f}%, Test_loss:{:.3f}')
    print(template.format(i + 1, epoch_train_acc*100, epoch_train_loss, epoch_test_acc*100, epoch_test_loss))
    
print("Done")

Epoch: 1, Train_acc:96.9%, Train_loss:0.289, Test_acc:100.0%, Test_loss:0.117
Epoch: 2, Train_acc:100.0%, Train_loss:0.025, Test_acc:100.0%, Test_loss:0.011
Epoch: 3, Train_acc:100.0%, Train_loss:0.007, Test_acc:100.0%, Test_loss:0.006
Epoch: 4, Train_acc:100.0%, Train_loss:0.004, Test_acc:100.0%, Test_loss:0.003
Epoch: 5, Train_acc:100.0%, Train_loss:0.003, Test_acc:100.0%, Test_loss:0.003
Epoch: 6, Train_acc:100.0%, Train_loss:0.002, Test_acc:100.0%, Test_loss:0.002
Epoch: 7, Train_acc:100.0%, Train_loss:0.002, Test_acc:100.0%, Test_loss:0.002
Epoch: 8, Train_acc:100.0%, Train_loss:0.002, Test_acc:100.0%, Test_loss:0.001
Epoch: 9, Train_acc:100.0%, Train_loss:0.001, Test_acc:100.0%, Test_loss:0.001
Epoch:10, Train_acc:100.0%, Train_loss:0.001, Test_acc:100.0%, Test_loss:0.001
Done

5、结果可视化

import matplotlib.pyplot as plt
#隐藏警告
import warnings
warnings.filterwarnings("ignore")               #忽略警告信息

epochs_range = range(epoches)

plt.figure(figsize=(12, 3))
plt.subplot(1, 2, 1)

plt.plot(epochs_range, train_acc, label='Training Accuracy')
plt.plot(epochs_range, test_acc, label='Test Accuracy')
plt.legend(loc='lower right')
plt.title('Training Accuracy')

plt.subplot(1, 2, 2)
plt.plot(epochs_range, train_loss, label='Training Loss')
plt.plot(epochs_range, test_loss, label='Test Loss')
plt.legend(loc='upper right')
plt.title('Training= Loss')
plt.show()

​

​

效果比ResNet50好