- 🍨 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
本周任务：
1.阅读ResNeXt论文，了解作者的构建思路
2.对比之前介绍的ResNet50V2、DenseNet算法
3.复现ResNeXt-50算法
一、模型结构
ResNeXt由何凯明团队，2017年CVPR会议上提出新型图像分类网络。它是ResNet升级版，在ResNet的基础上，引入cardinality概念。
在论文中，作者提出当时普遍存在的一个问题，如果要提高模型准确率，往往采取加深网络或者加宽网络的方法。但网络设计的难度和计算开销也增加了。为了一点精度的提升往往付出更大的代价。因此，需要在不额外增加计算代价的情况下，提升网络精度。
左边--ResNet，输入的具有256个通道的特征经过1*1卷积压缩到64个通道，之后3*3的卷积核用于处理特征，经1*1卷积扩大通道数与原特征残差连接后输出。
右边--ResNeXt，输入的具有256个通道的特征被分为32个组，每组被压缩到4个通道后处理，32个组相加后与原特征残差连接后输出。cardinality指的是一个block中所具有相同的分支的数目。
二、分组卷积
1.ResNeXt采用分组卷积：将特征图分为不同的组，再对每组特征图分别进行卷积，有效降低计算量。
2.分组卷积中，每个卷积核只处理部分通道，如下图，红色卷积核只处理红色通道，绿色卷积核只处理绿色通道，黄色卷积核只处理黄色通道。此时，每个卷积核有2个通道，每个卷积核生成一张特征图。
三、代码
学习于深度学习第J6周：ResNeXt-50实战解析_resnext50-CSDN博客
1.前期准备
#配置GPU
import os, PIL, random, pathlib
import torch
import torch.nn as nn
import torchvision.transforms as transforms
import torchvision
from torchvision import transforms, datasets
import torch.nn.functional as Fdevice = torch.device("cuda" if torch.cuda.is_available() else "cpu")print(device)#导入数据集
data_dir = './data/'
data_dir = pathlib.Path(data_dir)data_paths = list(data_dir.glob('*'))
classeNames = [str(path).split("\\")[1] for path in data_paths]
print(classeNames)image_count = len(list(data_dir.glob('*/*')))
print("图片总数为：", image_count)#数据预处理+划分数据集
train_transforms = transforms.Compose([transforms.Resize([224, 224]), # 将输入图片resize成统一尺寸# transforms.RandomHorizontalFlip(), # 随机水平翻转transforms.ToTensor(), # 将PIL Image或numpy.ndarray转换为tensor，并归一化到[0,1]之间transforms.Normalize( # 标准化处理-->转换为标准正太分布（高斯分布），使模型更容易收敛mean=[0.485, 0.456, 0.406],std=[0.229, 0.224, 0.225]) # 其中 mean=[0.485,0.456,0.406]与std=[0.229,0.224,0.225] 从数据集中随机抽样计算得到的。
])test_transform = transforms.Compose([transforms.Resize([224, 224]), # 将输入图片resize成统一尺寸transforms.ToTensor(), # 将PIL Image或numpy.ndarray转换为tensor，并归一化到[0,1]之间transforms.Normalize( # 标准化处理-->转换为标准正太分布（高斯分布），使模型更容易收敛mean=[0.485, 0.456, 0.406],std=[0.229, 0.224, 0.225]) # 其中 mean=[0.485,0.456,0.406]与std=[0.229,0.224,0.225] 从数据集中随机抽样计算得到的。
])total_data = datasets.ImageFolder("./data/", transform=train_transforms)
print(total_data.class_to_idx)train_size = int(0.8 * len(total_data))
test_size = len(total_data) - train_size
train_dataset, test_dataset = torch.utils.data.random_split(total_data, [train_size, test_size])batch_size = 32
train_dl = torch.utils.data.DataLoader(train_dataset,batch_size=batch_size,shuffle=True,num_workers=0)
test_dl = torch.utils.data.DataLoader(test_dataset,batch_size=batch_size,shuffle=True,num_workers=0)
for X, y in test_dl:print("Shape of X [N, C, H, W]: ", X.shape)print("Shape of y: ", y.shape, y.dtype)break结果：
2.模型
class BN_Conv2d(nn.Module):"""BN_CONV_RELU"""def __init__(self, in_channels, out_channels, kernel_size, stride, padding, dilation=1, groups=1, bias=False):super(BN_Conv2d, self).__init__()self.seq = nn.Sequential(nn.Conv2d(in_channels, out_channels, kernel_size=kernel_size, stride=stride,padding=padding, dilation=dilation, groups=groups, bias=bias),nn.BatchNorm2d(out_channels))def forward(self, x):return F.relu(self.seq(x))class ResNeXt_Block(nn.Module):"""ResNeXt block with group convolutions"""def __init__(self, in_chnls, cardinality, group_depth, stride):super(ResNeXt_Block, self).__init__()self.group_chnls = cardinality * group_depthself.conv1 = BN_Conv2d(in_chnls, self.group_chnls, 1, stride=1, padding=0)self.conv2 = BN_Conv2d(self.group_chnls, self.group_chnls, 3, stride=stride, padding=1, groups=cardinality)self.conv3 = nn.Conv2d(self.group_chnls, self.group_chnls*2, 1, stride=1, padding=0)self.bn = nn.BatchNorm2d(self.group_chnls*2)self.short_cut = nn.Sequential(nn.Conv2d(in_chnls, self.group_chnls*2, 1, stride, 0, bias=False),nn.BatchNorm2d(self.group_chnls*2))def forward(self, x):out = self.conv1(x)out = self.conv2(out)out = self.bn(self.conv3(out))out += self.short_cut(x)return F.relu(out)class ResNeXt(nn.Module):"""ResNeXt builder"""def __init__(self, layers: object, cardinality, group_depth, num_classes) -> object:super(ResNeXt, self).__init__()self.cardinality = cardinalityself.channels = 64self.conv1 = BN_Conv2d(3, self.channels, 7, stride=2, padding=3)d1 = group_depthself.conv2 = self.___make_layers(d1, layers[0], stride=1)d2 = d1 * 2self.conv3 = self.___make_layers(d2, layers[1], stride=2)d3 = d2 * 2self.conv4 = self.___make_layers(d3, layers[2], stride=2)d4 = d3 * 2self.conv5 = self.___make_layers(d4, layers[3], stride=2)self.fc = nn.Linear(self.channels, num_classes) # 224x224 input sizedef ___make_layers(self, d, blocks, stride):strides = [stride] + [1] * (blocks-1)layers = []for stride in strides:layers.append(ResNeXt_Block(self.channels, self.cardinality, d, stride))self.channels = self.cardinality*d*2return nn.Sequential(*layers)def forward(self, x):out = self.conv1(x)out = F.max_pool2d(out, 3, 2, 1)out = self.conv2(out)out = self.conv3(out)out = self.conv4(out)out = self.conv5(out)out = F.avg_pool2d(out, 7)out = out.view(out.size(0), -1)out = F.softmax(self.fc(out),dim=1)return out# 定义完成，测试一下
model = ResNeXt([3, 4, 6, 3], 32, 4, 4)
model.to(device)# 统计模型参数量以及其他指标
import torchsummary as summary
summary.summary(model, (3, 224, 224))结果：
3.训练运行
# 训练循环
def train(dataloader, model, loss_fn, optimizer):size = len(dataloader.dataset) # 训练集的大小num_batches = len(dataloader) # 批次数目, (size/batch_size，向上取整)train_loss, train_acc = 0, 0 # 初始化训练损失和正确率for X, y in dataloader: # 获取图片及其标签X, y = X.to(device), y.to(device)# 计算预测误差pred = model(X) # 网络输出loss = loss_fn(pred, y) # 计算网络输出和真实值之间的差距，targets为真实值，计算二者差值即为损失# 反向传播optimizer.zero_grad() # grad属性归零loss.backward() # 反向传播optimizer.step() # 每一步自动更新# 记录acc与losstrain_acc += (pred.argmax(1) == y).type(torch.float).sum().item()train_loss += loss.item()train_acc /= sizetrain_loss /= num_batchesreturn train_acc, train_lossdef test(dataloader, model, loss_fn):size = len(dataloader.dataset) # 测试集的大小num_batches = len(dataloader) # 批次数目test_loss, test_acc = 0, 0# 当不进行训练时，停止梯度更新，节省计算内存消耗with torch.no_grad():for imgs, target in dataloader:imgs, target = imgs.to(device), target.to(device)# 计算losstarget_pred = model(imgs)loss = loss_fn(target_pred, target)test_loss += loss.item()test_acc += (target_pred.argmax(1) == target).type(torch.float).sum().item()test_acc /= sizetest_loss /= num_batchesreturn test_acc, test_loss
import copyoptimizer = torch.optim.Adam(model.parameters(), lr=1e-4)
loss_fn = nn.CrossEntropyLoss() # 创建损失函数epochs = 10train_loss = []
train_acc = []
test_loss = []
test_acc = []best_acc = 0 # 设置一个最佳准确率，作为最佳模型的判别指标for epoch in range(epochs):# 更新学习率（使用自定义学习率时使用）# adjust_learning_rate(optimizer, epoch, learn_rate)model.train()epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, optimizer)# scheduler.step() # 更新学习率（调用官方动态学习率接口时使用）model.eval()epoch_test_acc, epoch_test_loss = test(test_dl, model, loss_fn)# 保存最佳模型到 best_modelif epoch_test_acc > best_acc:best_acc = epoch_test_accbest_model = copy.deepcopy(model)train_acc.append(epoch_train_acc)train_loss.append(epoch_train_loss)test_acc.append(epoch_test_acc)test_loss.append(epoch_test_loss)# 获取当前的学习率lr = optimizer.state_dict()['param_groups'][0]['lr']template = ('Epoch:{:2d}, Train_acc:{:.1f}%, Train_loss:{:.3f}, Test_acc:{:.1f}%, Test_loss:{:.3f}, Lr:{:.2E}')print(template.format(epoch + 1, epoch_train_acc * 100, epoch_train_loss,epoch_test_acc * 100, epoch_test_loss, lr))# 保存最佳模型到文件中
PATH = './best_model.pth' # 保存的参数文件名
torch.save(model.state_dict(), PATH)print('Done')结果：
4.打印训练图
import matplotlib.pyplot as plt
# 隐藏警告
import warningswarnings.filterwarnings("ignore") # 忽略警告信息
plt.rcParams['font.sans-serif'] = ['SimHei'] # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False # 用来正常显示负号
plt.rcParams['figure.dpi'] = 100 # 分辨率epochs_range = range(epochs)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 and Validation 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 and Validation Loss')
plt.show()四、总结
1.读论文原文要花很长时间，但有讲义，就会快速知道论文的创新点是什么。
2.实验的流程已经很熟悉，现在就在慢慢学每一步的具体内容，争取下次能自己写出。