• 🍨 本文為🔗365天深度學習訓練營中的學習記錄博客
• 🍖 原作者：K同學啊

本周任務：

1.閱讀ResNeXt論文，了解作者的構建思路

2.對比之前介紹的ResNet50V2、DenseNet算法

3.復現ResNeXt-50算法

一、模型結構

ResNeXt由何凱明團隊，2017年CVPR會議上提出新型圖像分類網絡。它是ResNet升級版，在ResNet的基礎上，引入cardinality概念。

在論文中，作者提出當時普遍存在的一個問題，如果要提高模型準確率，往往采取加深網絡或者加寬網絡的方法。但網絡設計的難度和計算開銷也增加了。為了一點精度的提升往往付出更大的代價。因此，需要在不額外增加計算代價的情況下，提升網絡精度。

左邊--ResNet，輸入的具有256個通道的特征經過1*1卷積壓縮到64個通道，之后3*3的卷積核用于處理特征，經1*1卷積擴大通道數與原特征殘差連接后輸出。

右邊--ResNeXt，輸入的具有256個通道的特征被分為32個組，每組被壓縮到4個通道后處理，32個組相加后與原特征殘差連接后輸出。cardinality指的是一個block中所具有相同的分支的數目。

二、分組卷積

1.ResNeXt采用分組卷積：將特征圖分為不同的組，再對每組特征圖分別進行卷積，有效降低計算量。

2.分組卷積中，每個卷積核只處理部分通道，如下圖，紅色卷積核只處理紅色通道，綠色卷積核只處理綠色通道，黃色卷積核只處理黃色通道。此時，每個卷積核有2個通道，每個卷積核生成一張特征圖。

三、代碼

學習于深度學習第J6周：ResNeXt-50實戰解析_resnext50-CSDN博客

?1.前期準備

#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)#導入數據集
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("圖片總數為：", image_count)#數據預處理+劃分數據集
train_transforms = transforms.Compose([transforms.Resize([224, 224]),  # 將輸入圖片resize成統一尺寸# transforms.RandomHorizontalFlip(), # 隨機水平翻轉transforms.ToTensor(),  # 將PIL Image或numpy.ndarray轉換為tensor，并歸一化到[0,1]之間transforms.Normalize(  # 標準化處理-->轉換為標準正太分布（高斯分布），使模型更容易收斂mean=[0.485, 0.456, 0.406],std=[0.229, 0.224, 0.225])  # 其中 mean=[0.485,0.456,0.406]與std=[0.229,0.224,0.225] 從數據集中隨機抽樣計算得到的。
])test_transform = transforms.Compose([transforms.Resize([224, 224]),  # 將輸入圖片resize成統一尺寸transforms.ToTensor(),  # 將PIL Image或numpy.ndarray轉換為tensor，并歸一化到[0,1]之間transforms.Normalize(  # 標準化處理-->轉換為標準正太分布（高斯分布），使模型更容易收斂mean=[0.485, 0.456, 0.406],std=[0.229, 0.224, 0.225])  # 其中 mean=[0.485,0.456,0.406]與std=[0.229,0.224,0.225] 從數據集中隨機抽樣計算得到的。
])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

結果：

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

# 定義完成，測試一下
model = ResNeXt([3, 4, 6, 3], 32, 4, 4)
model.to(device)# 統計模型參數量以及其他指標
import torchsummary as summary
summary.summary(model, (3, 224, 224))

結果：

?

?

?3.訓練運行

 
# 訓練循環
def train(dataloader, model, loss_fn, optimizer):size = len(dataloader.dataset)  # 訓練集的大小num_batches = len(dataloader)  # 批次數目, (size/batch_size，向上取整)train_loss, train_acc = 0, 0  # 初始化訓練損失和正確率for X, y in dataloader:  # 獲取圖片及其標簽X, y = X.to(device), y.to(device)# 計算預測誤差pred = model(X)  # 網絡輸出loss = loss_fn(pred, y)  # 計算網絡輸出和真實值之間的差距，targets為真實值，計算二者差值即為損失# 反向傳播optimizer.zero_grad()  # grad屬性歸零loss.backward()  # 反向傳播optimizer.step()  # 每一步自動更新# 記錄acc與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)  # 測試集的大小num_batches = len(dataloader)  # 批次數目test_loss, test_acc = 0, 0# 當不進行訓練時，停止梯度更新，節省計算內存消耗with torch.no_grad():for imgs, target in dataloader:imgs, target = imgs.to(device), target.to(device)# 計算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()  # 創建損失函數epochs = 10train_loss = []
train_acc = []
test_loss = []
test_acc = []best_acc = 0  # 設置一個最佳準確率，作為最佳模型的判別指標for epoch in range(epochs):# 更新學習率（使用自定義學習率時使用）# adjust_learning_rate(optimizer, epoch, learn_rate)model.train()epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, optimizer)# scheduler.step() # 更新學習率（調用官方動態學習率接口時使用）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)# 獲取當前的學習率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'  # 保存的參數文件名
torch.save(model.state_dict(), PATH)print('Done')

結果：

?

4.打印訓練圖

import matplotlib.pyplot as plt
# 隱藏警告
import warningswarnings.filterwarnings("ignore")  # 忽略警告信息
plt.rcParams['font.sans-serif'] = ['SimHei']  # 用來正常顯示中文標簽
plt.rcParams['axes.unicode_minus'] = False  # 用來正常顯示負號
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、總結

1.讀論文原文要花很長時間，但有講義，就會快速知道論文的創新點是什么。

2.實驗的流程已經很熟悉，現在就在慢慢學每一步的具體內容，爭取下次能自己寫出。