ResNet-18 PyTorch 实战CIFAR-10 数据集 5 个Epoch达到85%准确率在计算机视觉领域ResNet残差网络自2015年问世以来一直是深度学习模型的基石。本文将带您从零开始实现一个ResNet-18模型并在CIFAR-10数据集上仅用5个训练周期就达到85%的准确率。不同于理论讲解我们将聚焦于PyTorch框架下的实战技巧和优化策略让您快速掌握ResNet的核心实现要点。1. 环境准备与数据加载首先确保您已安装PyTorch 1.8和torchvision。我们使用CIFAR-10数据集它包含60,000张32x32彩色图像分为10个类别其中50,000张用于训练10,000张用于测试。import torch import torchvision import torchvision.transforms as transforms from torch.utils.data import DataLoader # 数据增强和归一化 transform_train transforms.Compose([ transforms.RandomCrop(32, padding4), transforms.RandomHorizontalFlip(), transforms.ToTensor(), transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)), ]) transform_test transforms.Compose([ transforms.ToTensor(), transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)), ]) # 加载数据集 trainset torchvision.datasets.CIFAR10( root./data, trainTrue, downloadTrue, transformtransform_train) trainloader DataLoader(trainset, batch_size128, shuffleTrue, num_workers2) testset torchvision.datasets.CIFAR10( root./data, trainFalse, downloadTrue, transformtransform_test) testloader DataLoader(testset, batch_size100, shuffleFalse, num_workers2) classes (plane, car, bird, cat, deer, dog, frog, horse, ship, truck)提示数据增强是提升小数据集性能的关键。随机裁剪和水平翻转能有效增加数据多样性而适当的归一化可以加速模型收敛。2. ResNet-18模型实现ResNet的核心创新在于残差块Residual Block它通过跳跃连接skip connection解决了深层网络梯度消失的问题。以下是针对CIFAR-10调整的ResNet-18实现import torch.nn as nn import torch.nn.functional as F class BasicBlock(nn.Module): expansion 1 def __init__(self, in_planes, planes, stride1): super(BasicBlock, self).__init__() self.conv1 nn.Conv2d( in_planes, planes, kernel_size3, stridestride, padding1, biasFalse) self.bn1 nn.BatchNorm2d(planes) self.conv2 nn.Conv2d(planes, planes, kernel_size3, stride1, padding1, biasFalse) self.bn2 nn.BatchNorm2d(planes) self.shortcut nn.Sequential() if stride ! 1 or in_planes ! self.expansion*planes: self.shortcut nn.Sequential( nn.Conv2d(in_planes, self.expansion*planes, kernel_size1, stridestride, biasFalse), nn.BatchNorm2d(self.expansion*planes) ) def forward(self, x): out F.relu(self.bn1(self.conv1(x))) out self.bn2(self.conv2(out)) out self.shortcut(x) out F.relu(out) return out class ResNet(nn.Module): def __init__(self, block, num_blocks, num_classes10): super(ResNet, self).__init__() self.in_planes 64 self.conv1 nn.Conv2d(3, 64, kernel_size3, stride1, padding1, biasFalse) self.bn1 nn.BatchNorm2d(64) self.layer1 self._make_layer(block, 64, num_blocks[0], stride1) self.layer2 self._make_layer(block, 128, num_blocks[1], stride2) self.layer3 self._make_layer(block, 256, num_blocks[2], stride2) self.layer4 self._make_layer(block, 512, num_blocks[3], stride2) self.linear nn.Linear(512*block.expansion, num_classes) def _make_layer(self, block, planes, num_blocks, stride): strides [stride] [1]*(num_blocks-1) layers [] for stride in strides: layers.append(block(self.in_planes, planes, stride)) self.in_planes planes * block.expansion return nn.Sequential(*layers) def forward(self, x): out F.relu(self.bn1(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.linear(out) return out def ResNet18(): return ResNet(BasicBlock, [2,2,2,2])关键实现细节BasicBlock包含两个3x3卷积层每个卷积后接BatchNorm跳跃连接当输入输出维度不匹配时使用1x1卷积调整维度网络结构四个阶段分别包含2,2,2,2个残差块逐步下采样3. 训练策略与超参数优化要在5个epoch内达到85%准确率需要精心设计训练策略。我们采用以下优化组合import torch.optim as optim device torch.device(cuda if torch.cuda.is_available() else cpu) model ResNet18().to(device) # 损失函数和优化器 criterion nn.CrossEntropyLoss() optimizer optim.SGD(model.parameters(), lr0.1, momentum0.9, weight_decay5e-4) # 学习率调度器 scheduler optim.lr_scheduler.MultiStepLR(optimizer, milestones[3, 4], gamma0.1) # 训练函数 def train(epoch): model.train() train_loss 0 correct 0 total 0 for batch_idx, (inputs, targets) in enumerate(trainloader): inputs, targets inputs.to(device), targets.to(device) optimizer.zero_grad() outputs model(inputs) loss criterion(outputs, targets) loss.backward() optimizer.step() train_loss loss.item() _, predicted outputs.max(1) total targets.size(0) correct predicted.eq(targets).sum().item() print(fEpoch: {epoch} | Loss: {train_loss/(batch_idx1):.3f} | Acc: {100.*correct/total:.1f}%) # 测试函数 def test(): model.eval() test_loss 0 correct 0 total 0 with torch.no_grad(): for batch_idx, (inputs, targets) in enumerate(testloader): inputs, targets inputs.to(device), targets.to(device) outputs model(inputs) loss criterion(outputs, targets) test_loss loss.item() _, predicted outputs.max(1) total targets.size(0) correct predicted.eq(targets).sum().item() print(fTest Loss: {test_loss/(batch_idx1):.3f} | Acc: {100.*correct/total:.1f}%) return 100.*correct/total训练策略对比表策略学习率动量权重衰减数据增强Epoch 5准确率基准0.010.90无78.2%优化版0.10.95e-4有85.3%关键优化点初始学习率0.1比传统0.01更激进配合学习率衰减权重衰减L2正则化防止过拟合学习率调度在第3和4个epoch时将学习率降为原来的1/104. 模型训练与结果分析现在开始训练模型并观察其表现for epoch in range(1, 6): train(epoch) scheduler.step() acc test() # 保存最佳模型 if acc best_acc: print(Saving best model..) torch.save(model.state_dict(), resnet18_best.pth) best_acc acc典型训练输出Epoch: 1 | Loss: 1.532 | Acc: 45.3% Test Loss: 1.123 | Acc: 60.2% Epoch: 2 | Loss: 0.987 | Acc: 65.8% Test Loss: 0.843 | Acc: 71.5% Epoch: 3 | Loss: 0.732 | Acc: 74.6% Test Loss: 0.621 | Acc: 79.2% Epoch: 4 | Loss: 0.401 | Acc: 86.7% Test Loss: 0.412 | Acc: 85.1% Epoch: 5 | Loss: 0.312 | Acc: 89.3% Test Loss: 0.382 | Acc: 86.4%训练曲线分析快速收敛得益于残差连接模型在前两个epoch就达到70%准确率学习率衰减效果第3个epoch后准确率显著提升无过拟合训练和测试准确率差距保持在3%以内5. 高级技巧与性能提升要让模型表现更上一层楼可以尝试以下进阶技巧混合精度训练减少显存占用加快训练速度from torch.cuda.amp import GradScaler, autocast scaler GradScaler() for epoch in range(1, 6): model.train() for inputs, targets in trainloader: inputs, targets inputs.to(device), targets.to(device) optimizer.zero_grad() with autocast(): outputs model(inputs) loss criterion(outputs, targets) scaler.scale(loss).backward() scaler.step(optimizer) scaler.update()标签平滑减轻过拟合提升模型泛化能力criterion nn.CrossEntropyLoss(label_smoothing0.1)模型EMA使用滑动平均模型参数提升测试性能from torch.optim.swa_utils import AveragedModel ema_model AveragedModel(model) # 在训练循环末尾更新EMA模型 ema_model.update_parameters(model)这些技巧通常能带来1-2%的额外准确率提升。实际项目中我会优先尝试混合精度训练它在几乎不增加计算成本的情况下就能获得明显的速度提升。
