第J6周：ResNeXt-50实战解析

FROM

• 🍨 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

---

我的环境

• 语言环境：Python 3.10.12
• 开发工具：Jupyter Lab
• 深度学习环境：
  • torch==2.3.1+cu121
  • torchvision==0.18.1+cu121

---

一、本周内容

1. ResNeXt 构建思路

ResNeXt 是由何凯明团队在 2017 年提出的一种新型图像分类网络，是 ResNet 的升级版。其核心思想是引入了 cardinality（基数） 的概念，即通过增加转换集合的大小来提升网络性能，而不是单纯地增加网络深度或宽度。ResNeXt 的设计基于以下几点：

• Split-Transform-Merge 策略：ResNeXt 采用了类似于 Inception 的分治思想，将输入特征拆分成多个组（group），每组独立进行卷积操作，最后将结果合并。这种设计在不增加计算复杂度的情况下，提高了特征的多样性。
• Cardinality：通过增加分组的数量（而不是增加卷积核的数量），ResNeXt 能够在保持计算量不变的情况下提升分类精度。
• 聚合残差结构：ResNeXt 的每个残差块聚合了一组具有相同拓扑结构的转换，这种设计使得网络在较深的结构下仍能保持高效的性能。

2.ResNeXt vs ResNet50V2 vs DenseNet

特性	ResNeXt	ResNet50V2	DenseNet
核心思想	引入 cardinality，通过分组卷积提升特征多样性	预激活设计，优化深层网络的训练	密集连接，特征重用
网络结构	聚合残差结构，分组卷积	改进的瓶颈结构，预激活	密集块 + 过渡层
计算量	在相同深度下计算量较低	参数量较少，适合超深网络	参数利用效率高，但计算量可能较大
优势	在较深网络下精度更高	适合超深网络，训练更稳定	特征重用，泛化能力强
适用场景	需要高精度且计算资源有限	构建非常深的网络	数据量较大，需要强特征提取

二、核心代码及运行截图

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


class ResNeXtBottleneck(nn.Module):
    expansion = 4

    def __init__(self, in_channels, out_channels, stride=1, cardinality=32, base_width=4):
        super().__init__()
        width_ratio = out_channels / (cardinality * base_width)
        D = cardinality * int(base_width * width_ratio)

        self.conv_reduce = nn.Conv2d(in_channels, D, kernel_size=1, bias=False)
        self.bn_reduce = nn.BatchNorm2d(D)

        self.conv_conv = nn.Conv2d(D, D, kernel_size=3, stride=stride, padding=1, groups=cardinality, bias=False)
        self.bn = nn.BatchNorm2d(D)

        self.conv_expand = nn.Conv2d(D, out_channels * self.expansion, kernel_size=1, bias=False)
        self.bn_expand = nn.BatchNorm2d(out_channels * self.expansion)

        self.shortcut = nn.Sequential()
        if stride != 1 or in_channels != out_channels * self.expansion:
            self.shortcut = nn.Sequential(
                nn.Conv2d(in_channels, out_channels * self.expansion, kernel_size=1, stride=stride, bias=False),
                nn.BatchNorm2d(out_channels * self.expansion)
            )

    def forward(self, x):
        out = F.relu(self.bn_reduce(self.conv_reduce(x)), inplace=True)
        out = F.relu(self.bn(self.conv_conv(out)), inplace=True)
        out = self.bn_expand(self.conv_expand(out))
        out += self.shortcut(x)
        return F.relu(out, inplace=True)


class ResNeXt(nn.Module):
    def __init__(self, block, layers, cardinality, num_classes=1000, base_width=4):
        super().__init__()
        self.in_channels = 64

        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], 1, cardinality, base_width)
        self.layer2 = self._make_layer(block, 128, layers[1], 2, cardinality, base_width)
        self.layer3 = self._make_layer(block, 256, layers[2], 2, cardinality, base_width)
        self.layer4 = self._make_layer(block, 512, layers[3], 2, cardinality, base_width)

        self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
        self.fc = nn.Linear(512 * block.expansion, num_classes)

    def _make_layer(self, block, out_channels, num_blocks, stride, cardinality, base_width):
        strides = [stride] + [1] * (num_blocks - 1)
        layers = []
        for stride in strides:
            layers.append(block(self.in_channels, out_channels, stride, cardinality, base_width))
            self.in_channels = out_channels * block.expansion
        return nn.Sequential(*layers)

    def forward(self, x):
        x = self.relu(self.bn1(self.conv1(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 = torch.flatten(x, 1)
        return self.fc(x)


def resnext50():
    return ResNeXt(ResNeXtBottleneck, [3, 4, 6, 3], cardinality=32, num_classes=1000, base_width=4)


# 创建模型实例
model = resnext50().cuda()
model

输出：

ResNeXt(
  (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): ResNeXtBottleneck(
      (conv_reduce): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential(
        (0): Conv2d(64, 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)
      )
    )
    (1): ResNeXtBottleneck(
      (conv_reduce): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential()
    )
    (2): ResNeXtBottleneck(
      (conv_reduce): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential()
    )
  )
  (layer2): Sequential(
    (0): ResNeXtBottleneck(
      (conv_reduce): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): 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): ResNeXtBottleneck(
      (conv_reduce): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential()
    )
    (2): ResNeXtBottleneck(
      (conv_reduce): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential()
    )
    (3): ResNeXtBottleneck(
      (conv_reduce): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential()
    )
  )
  (layer3): Sequential(
    (0): ResNeXtBottleneck(
      (conv_reduce): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential(
        (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): ResNeXtBottleneck(
      (conv_reduce): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential()
    )
    (2): ResNeXtBottleneck(
      (conv_reduce): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential()
    )
    (3): ResNeXtBottleneck(
      (conv_reduce): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential()
    )
    (4): ResNeXtBottleneck(
      (conv_reduce): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential()
    )
    (5): ResNeXtBottleneck(
      (conv_reduce): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential()
    )
  )
  (layer4): Sequential(
    (0): ResNeXtBottleneck(
      (conv_reduce): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential(
        (0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): ResNeXtBottleneck(
      (conv_reduce): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential()
    )
    (2): ResNeXtBottleneck(
      (conv_reduce): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_reduce): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
      (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv_expand): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn_expand): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (shortcut): Sequential()
    )
  )
  (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))
  (fc): Linear(in_features=2048, out_features=1000, bias=True)
)

训练结果：

绘制训练历史

输出：

验证

import torch
from torchvision import transforms
from PIL import Image

# 加载模型
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = resnext50().cuda()

# 加载保存的文件
checkpoint = torch.load('best_resnetxt_model.pth', map_location=device)

# 只加载模型的参数
model.load_state_dict(checkpoint['model_state_dict'])

# 设置为评估模式
model.eval()

# 数据预处理
transform = transforms.Compose([
    transforms.Resize(256),  # 将图像大小调整为256x256
    transforms.CenterCrop(224),  # 从中心裁剪出224x224的图像
    transforms.ToTensor(),  # 将图像转换为Tensor
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])  # 归一化
])

# 加载图像
img1 = Image.open('./houdou/Monkeypox/M01_03_08.jpg')  # 替换为图像路径
img2 = Image.open('./houdou/Others/NM01_01_06.jpg')
img1 = transform(img1).unsqueeze(0)  # 添加批次维度
img1 = img1.to(device)
img2 = transform(img2).unsqueeze(0)  # 添加批次维度
img2 = img2.to(device)

# 定义标签映射
label_mapping = {0: "猴痘", 1: "others"}  # 假设 0 对应猴痘，1 对应 others

# 预测
with torch.no_grad():
    outputs1 = model(img1)
    _, predicted1 = torch.max(outputs1, 1)
    outputs2 = model(img2)
    _, predicted2 = torch.max(outputs2, 1)

# 输出预测结果
predicted_class1 = predicted1.item()
predicted_label1 = label_mapping[predicted_class1]  # 将索引映射为标签

predicted_class2 = predicted2.item()
predicted_label2 = label_mapping[predicted_class2] 
print(f'img1预测结果：{predicted_label1}')
print(f'img2预测结果：{predicted_label2}')

输出：