Pytorch實戰2：ResNet-18實現Cifar-10影象分類（測試集分類準確率95.170%）（轉）

阿新 • • 發佈：2019-01-16

Pytorch實戰2：ResNet-18實現Cifar-10影象分類

實驗環境:

torchvision 0.2.1
Python 3.6
CUDA8+cuDNN v7 (可選)
Win10+Pycharm

整個專案程式碼：點選這裡

ResNet-18網路結構：

這裡寫圖片描述
ResNet全名Residual Network殘差網路。Kaiming He 的《Deep Residual Learning for Image Recognition》獲得了CVPR最佳論文。他提出的深度殘差網路在2015年可以說是洗刷了影象方面的各大比賽，以絕對優勢取得了多個比賽的冠軍。而且它在保證網路精度的前提下，將網路的深度達到了152層，後來又進一步加到1000的深度。論文的開篇先是說明了深度網路的好處：特徵等級隨著網路的加深而變高，網路的表達能力也會大大提高。因此論文中提出了一個問題：是否可以通過疊加網路層數來獲得一個更好的網路呢？作者經過實驗發現，單純的把網路疊起來的深層網路的效果反而不如合適層數的較淺的網路效果。因此何愷明等人在普通平原網路的基礎上增加了一個shortcut，構成一個residual block。此時擬合目標就變為F(x)，F(x)就是殘差：

Pytorch上搭建ResNet-18：

'''ResNet-18 Image classfication for cifar-10 with PyTorch 

Author 'Sun-qian'.

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

classResidualBlock(nn.Module):
    def__init__(self, inchannel, outchannel, stride=1):
        super(ResidualBlock, self).__init__()
        self.left = nn.Sequential(
            nn.Conv2d(inchannel, outchannel, kernel_size=3 
, stride=stride, padding=1, bias=False),
            nn.BatchNorm2d(outchannel),
            nn.ReLU(inplace=True),
            nn.Conv2d(outchannel, outchannel, kernel_size=3, stride=1, padding=1, bias=False),
            nn.BatchNorm2d(outchannel)
        )
        self.shortcut = nn.Sequential()
        if 
 stride != 1 or inchannel != outchannel:
            self.shortcut = nn.Sequential(
                nn.Conv2d(inchannel, outchannel, kernel_size=1, stride=stride, bias=False),
                nn.BatchNorm2d(outchannel)
            )

    defforward(self, x):
        out = self.left(x)
        out += self.shortcut(x)
        out = F.relu(out)
        return out

classResNet(nn.Module):
    def__init__(self, ResidualBlock, num_classes=10):
        super(ResNet, self).__init__()
        self.inchannel = 64
        self.conv1 = nn.Sequential(
            nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1, bias=False),
            nn.BatchNorm2d(64),
            nn.ReLU(),
        )
        self.layer1 = self.make_layer(ResidualBlock, 64,  2, stride=1)
        self.layer2 = self.make_layer(ResidualBlock, 128, 2, stride=2)
        self.layer3 = self.make_layer(ResidualBlock, 256, 2, stride=2)
        self.layer4 = self.make_layer(ResidualBlock, 512, 2, stride=2)
        self.fc = nn.Linear(512, num_classes)

    defmake_layer(self, block, channels, num_blocks, stride):
        strides = [stride] + [1] * (num_blocks - 1)   #strides=[1,1]
        layers = []
        for stride in strides:
            layers.append(block(self.inchannel, channels, stride))
            self.inchannel = channels
        return nn.Sequential(*layers)

    defforward(self, x):
        out = self.conv1(x)
        out = self.layer1(out)
        out = self.layer2(out)
        out = self.layer3(out)
        out = self.layer4(out)
        out = F.avg_pool2d(out, 4)
        out = out.view(out.size(0), -1)
        out = self.fc(out)
        return out


defResNet18():

    return ResNet(ResidualBlock)

Pytorch上訓練：

所選資料集為Cifar-10，該資料集共有60000張帶標籤的彩色影象，這些影象尺寸32*32，分為10個類，每類6000張圖。這裡面有50000張用於訓練，每個類5000張，另外10000用於測試，每個類1000張。訓練時人為修改學習率，當epoch:[1-135] ，lr=0.1；epoch:[136-185]， lr=0.01；epoch:[186-240] ，lr=0.001。訓練程式碼如下：

import torch
import torch.nn as nn
import torch.optim as optim
import torchvision
import torchvision.transforms as transforms
import argparse
from resnet import ResNet18

# 定義是否使用GPU
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# 引數設定,使得我們能夠手動輸入命令列引數，就是讓風格變得和Linux命令列差不多
parser = argparse.ArgumentParser(description='PyTorch CIFAR10 Training')
parser.add_argument('--outf', default='./model/', help='folder to output images and model checkpoints') #輸出結果儲存路徑
parser.add_argument('--net', default='./model/Resnet18.pth', help="path to net (to continue training)")  #恢復訓練時的模型路徑
args = parser.parse_args()

# 超引數設定
EPOCH = 135   #遍歷資料集次數
pre_epoch = 0  # 定義已經遍歷資料集的次數
BATCH_SIZE = 128      #批處理尺寸(batch_size)
LR = 0.1        #學習率

# 準備資料集並預處理
transform_train = transforms.Compose([
    transforms.RandomCrop(32, padding=4),  #先四周填充0，在吧影象隨機裁剪成32*32
    transforms.RandomHorizontalFlip(),  #影象一半的概率翻轉，一半的概率不翻轉
    transforms.ToTensor(),
    transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)), #R,G,B每層的歸一化用到的均值和方差
])

transform_test = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
])

trainset = torchvision.datasets.CIFAR10(root='./data', train=True, download=True, transform=transform_train) #訓練資料集
trainloader = torch.utils.data.DataLoader(trainset, batch_size=BATCH_SIZE, shuffle=True, num_workers=2)   #生成一個個batch進行批訓練，組成batch的時候順序打亂取

testset = torchvision.datasets.CIFAR10(root='./data', train=False, download=True, transform=transform_test)
testloader = torch.utils.data.DataLoader(testset, batch_size=100, shuffle=False, num_workers=2)
# Cifar-10的標籤
classes = ('plane', 'car', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck')

# 模型定義-ResNet
net = ResNet18().to(device)

# 定義損失函式和優化方式
criterion = nn.CrossEntropyLoss()  #損失函式為交叉熵，多用於多分類問題
optimizer = optim.SGD(net.parameters(), lr=LR, momentum=0.9, weight_decay=5e-4) #優化方式為mini-batch momentum-SGD，並採用L2正則化（權重衰減）

# 訓練
if __name__ == "__main__":
    best_acc = 85  #2 初始化best test accuracy
    print("Start Training, Resnet-18!")  # 定義遍歷資料集的次數
    with open("acc.txt", "w") as f:
        with open("log.txt", "w")as f2:
            for epoch in range(pre_epoch, EPOCH):
                print('\nEpoch: %d' % (epoch + 1))
                net.train()
                sum_loss = 0.0
                correct = 0.0
                total = 0.0
                for i, data in enumerate(trainloader, 0):
                    # 準備資料
                    length = len(trainloader)
                    inputs, labels = data
                    inputs, labels = inputs.to(device), labels.to(device)
                    optimizer.zero_grad()

                    # forward + backward
                    outputs = net(inputs)
                    loss = criterion(outputs, labels)
                    loss.backward()
                    optimizer.step()

                    # 每訓練1個batch列印一次loss和準確率
                    sum_loss += loss.item()
                    _, predicted = torch.max(outputs.data, 1)
                    total += labels.size(0)
                    correct += predicted.eq(labels.data).cpu().sum()
                    print('[epoch:%d, iter:%d] Loss: %.03f | Acc: %.3f%% '
                          % (epoch + 1, (i + 1 + epoch * length), sum_loss / (i + 1), 100. * correct / total))
                    f2.write('%03d  %05d |Loss: %.03f | Acc: %.3f%% '
                          % (epoch + 1, (i + 1 + epoch * length), sum_loss / (i + 1), 100. * correct / total))
                    f2.write('\n')
                    f2.flush()

                # 每訓練完一個epoch測試一下準確率
                print("Waiting Test!")
                with torch.no_grad():
                    correct = 0
                    total = 0
                    for data in testloader:
                        net.eval()
                        images, labels = data
                        images, labels = images.to(device), labels.to(device)
                        outputs = net(images)
                        # 取得分最高的那個類 (outputs.data的索引號)
                        _, predicted = torch.max(outputs.data, 1)
                        total += labels.size(0)
                        correct += (predicted == labels).sum()
                    print('測試分類準確率為：%.3f%%' % (100 * correct / total))
                    acc = 100. * correct / total
                    # 將每次測試結果實時寫入acc.txt檔案中
                    print('Saving model......')
                    torch.save(net.state_dict(), '%s/net_%03d.pth' % (args.outf, epoch + 1))
                    f.write("EPOCH=%03d,Accuracy= %.3f%%" % (epoch + 1, acc))
                    f.write('\n')
                    f.flush()
                    # 記錄最佳測試分類準確率並寫入best_acc.txt檔案中
                    if acc > best_acc:
                        f3 = open("best_acc.txt", "w")
                        f3.write("EPOCH=%d,best_acc= %.3f%%" % (epoch + 1, acc))
                        f3.close()
                        best_acc = acc
            print("Training Finished, TotalEPOCH=%d" % EPOCH)

實驗結果：best_acc= 95.170%

這裡寫圖片描述

Pytorch實戰2：ResNet-18實現Cifar-10影象分類（測試集分類準確率95.170%）（轉）