CNN算法（二）——稠密连接网络DenseNet

上一次的ResNet模型的核心是通过建立前面层与后面层之间的”短路连接“。而今天的DenseNet模型，他的基本思路与ResNet一致，但它建立的是前面所有层与后面层的密集连接。

一、使用pytorch实现DenseNet121完整代码

import torch
import torch.nn as nn
import torchvision.transforms as transforms
import torchvision
from torchvision import transforms, datasets
import os,PIL,pathlib,warnings
import matplotlib.pyplot as plt
import copy
import torch.nn.functional as F
from PIL import Image

#隐藏警告
import warnings

warnings.filterwarnings("ignore")

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(device)

#导入数据
import os,PIL,random,pathlib
# 支持中文
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

data_dir = r'D:\编程文件\数据库\J1ResNet-50\第8天\bird_photos'
data_dir = pathlib.Path(data_dir)


#查看数据
image_count=len(list(data_dir.glob('*/*')))
print("图片数为：",image_count)

data_paths  = list(data_dir.glob('*'))
classeNames = [str(path).split("\\")[6] for path in data_paths]
print(classeNames)


#加载数据
batch_size = 2

# 关于transforms.Compose的更多介绍可以参考：https://blog.csdn.net/qq_38251616/article/details/124878863
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("D:\编程文件\数据库\P2天气识别\weather_photos",transform=train_transforms)
print(total_data)

print(total_data.class_to_idx)


#划分数据集
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])
print(train_dataset, test_dataset)


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)

#检查数据
if __name__ == '__main__':
    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

image_folder = r'D:\编程文件\数据库\J1ResNet-50\第8天\bird_photos\Cockatoo'  # 指定图像文件夹路径
image_files = [f for f in os.listdir(image_folder) if f.endswith((".jpg", ".png", ".jpeg"))]
fig, axes = plt.subplots(2, 4, figsize=(16, 6))

for ax, img_file in zip(axes.flat, image_files):
    img_path = os.path.join(image_folder, img_file)
    img = Image.open(img_path)
    ax.imshow(img)
    ax.axis('off')
plt.tight_layout()
plt.show()

#构建残差模型
# 构造ResNet50模型
class DenseLayer(nn.Sequential):
    def __init__(self, num_input_features,growth_rate,bn_size,drop_rate):
        super(DenseLayer, self).__init__()
        self.add_module("norm1",nn.BatchNorm2d(num_input_features))
        self.add_module("relu1",nn.ReLU(inplace=True))
        self.add_module("conv1",nn.Conv2d(num_input_features,bn_size*growth_rate,kernel_size=1,stride=1,bias=False))

        self.add_module("norm2", nn.BatchNorm2d(bn_size*growth_rate))
        self.add_module("relu2", nn.ReLU(inplace=True))
        self.add_module("conv2",
                        nn.Conv2d(bn_size*growth_rate, growth_rate, kernel_size = 3, stride=1, padding=1,bias=False))
        self.drop_rate=drop_rate

    def forward(self, x):
        new_features=super(_DenseLayer,self).forward(x)
        if self.drop_rate>0:
            new_features=F.dropout(new_features,p=self.drop_rate,training=self.training)

        return torch.cat([x,new_features],1)


class _DenseBlock(nn.Sequential):
    def __init__(self, num_layers,num_input_features,bn_size,growth_rate,drop_rate):
        super(_DenseBlock, self).__init__()
        for i in range(num_layers):
            layer=_DenseLayer(num_input_features+i*growth_rate,growth_rate,bn,drop_rate)
            self.add_module("denselayer%d"%(i+1,),layer)


class _Transition(nn.Sequential):
    def __init__(self,num_input_feature,num_output_feature):
        super(_Transition,self).__init__()
        self.add_module("norm",nn.BatchNorm2d(num_input_feature))
        self.add_module("relu",nn.ReLU(inplace=True))
        self.add_module("conv",nn.Conv2d(num_input_feature,num_output_feature,kernel_size=1,stride=1,bias=False))
        self.add_module("pool",nn.AvgPool2d(2,stride=2))

class DenseNet(nn.Module):
    def __init__(self,growth=32,block_config=(6,12,24,16),num_init_feature=64,
                 bn_size=4,compression_rate=0.5,drop_rate=0,num_classes=1000):
        super(DenseNet, self).__init__()
        # first Conv2d
        self.features = nn.Sequential(OrderedDict([
            ("conv0", nn.Conv2d(3, num_init_features, kernel_size=7, stride=2, padding=3, bias=False)),
            ("norm0", nn.BatchNorm2d(num_init_features)),
            ("relu0", nn.ReLU(inplace=True)),
            ("pool0", nn.MaxPool2d(3, stride=2, padding=1))
        ]))

        # DenseBlock
        num_features = num_init_features
        for i, num_layers in enumerate(block_config):
            block = _DenseBlock(num_layers, num_features, bn_size, growth_rate, drop_rate)
            self.features.add_module("denseblock%d" % (i + 1), block)
            num_features += num_layers * growth_rate
            if i != len(block_config) - 1:
                transition = _Transition(num_features, int(num_features * compression_rate))
                self.features.add_module("transition%d" % (i + 1), transition)
                num_features = int(num_features * compression_rate)

        # final bn+ReLu
        self.features.add_module("norm5", nn.BatchNorm2d(num_features))
        self.features.add_module("relu5", nn.ReLU(inplace=True))

        # classification layer
        self.classifier = nn.Linear(num_features, num_classes)

        # params initialization
        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal(m.weight)
            elif isinstance(m, nn.BatchNorm2d):
                nn.init.constant_(m.bias, 0)
                nn.init.constant_(m.weight, 1)
            elif isinstance(m, nn.Linear):
                nn.init.constant_(m.bias, 0)

    def forward(self, x):
        features = self.features(x)
        out = F.avg_pool2d(features, 7, stride=1).view(features.size(0), -1)
        out = self.classifier(out)
        return out


def densenet121(pretrained=False, **kwargs):
    """DenseNet121"""
    model = DenseNet(num_init_features=64, growth_rate=32, block_config=(6, 12, 24, 16), **kwargs)
    if pretrained:
        # '.'s are no longer allowed in module names, but pervious _DenseLayer
        # has keys 'norm.1', 'relu.1', 'conv.1', 'norm.2', 'relu.2', 'conv.2'.
        # They are also in the checkpoints in model_urls. This pattern is used
        # to find such keys.
        pattern = re.compile(
            r'^(.*denselayer\d+\.(?:norm|relu|conv))\.((?:[12])\.(?:weight|bias|running_mean|running_var))$')
        state_dict = model_zoo.load_url(model_urls['densenet121'])
        for key in list(state_dict.keys()):
            res = pattern.match(key)
            if res:
                new_key = res.group(1) + res.group(2)
                state_dict[new_key] = state_dict[key]
                del state_dict[key]
        model.load_state_dict(state_dict)
    return model

import torchvision.models as models

model = models.densenet121(pretrained=True)


#编写训练函数
def train(dataloader, model, optimizer, loss_fn):
    size = len(dataloader.dataset)
    num_batches = len(dataloader)

    train_acc, train_loss = 0, 0

    for X, y in dataloader:
        X, y = X.to(device), y.to(device)

        pred = model(X)
        loss = loss_fn(pred, y)

        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        train_loss += loss.item()
        train_acc += (pred.argmax(1) == y).type(torch.float).sum().item()

    train_loss /= num_batches
    train_acc /= size

    return train_acc, train_loss


#编写测试函数

def test(dataloader, model, loss_fn):
    size = len(dataloader.dataset)  # 测试集的大小
    num_batches = len(dataloader)  # 批次数目, (size/batch_size，向上取整)
    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loss
            target_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 /= size
    test_loss /= num_batches

    return test_acc, test_loss



#设置损失函数和学习率
loss_fn = nn.CrossEntropyLoss()   #交叉熵函数
learn_rate = 1e-3
opt = torch.optim.Adam(model.parameters(), lr=learn_rate)

#正式训练
epochs = 8
train_loss = []
train_acc = []
test_loss = []
test_acc = []
best_acc = 0

# 开始训练
for epoch in range(epochs):

    model.train()
    epoch_train_acc, epoch_train_loss = train(train_dl, model, opt, loss_fn)

    model.eval()
    epoch_test_acc, epoch_test_loss = test(test_dl, model, loss_fn)

    if epoch_test_acc > best_acc:
        best_acc = epoch_test_acc
        best_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)

    lr = opt.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))

print('Done')

#结果可视化
import warnings

warnings.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()

二、理论知识储备

1.设计理念

• ResNet是每个层与前面的某层（一般是2~4层）短路连接在一起，连接的方式是通过元素相加。

• DenseNet中，每个层都会与前面所有层在channel维度连接(concat)在一起，是通过元素叠加，并作为下一层的输入。

• 

2.DenseNet与RestNet最主要区别

对于一个L层的网络，DenseNet共包含L(L+1)/2个连接，相比于RestNet，这是一种密集连接，而且DenseNet是直接concat来自不同层的特征图，进而可以实现特征重用，提升效率，这是两者最主要的区别。

3.网络结构

DenseNet网络中使用DenseBlock+Transition的结构。

• 其中DenseBlock是包含很多层的模块，每个层的特征图大小相同，层与层之间采用密集连接方式； 结构为BN+ReLU+3x3 Conv
• 而Transition层是连接两个相邻的DenseBlock，并且通过Pooling使特征图大小降低。
  结构为BN+ReLU+1x1Conv+2x2AvgPooling。

4.DenseNet的几个结构

• DenseNet-B结构：由于后面层的输入会非常大，DenseBlock内部可以采用bottleneck层来减少计算量，主要是原有的结构中增加1x1Conv，即BN+ReLU+1x1Conv+BN+ReLU+3x3Conv。
• DenseNet-C结构：假定Transition层的上接DenseBlock得到的特征图channels数为 m，Transition层可以产生[θm]个特征(通过卷积层)，其中θ∈(0,1]是压缩系数。当θ=1时，特征个数经过Transition层没有变化，即无压缩，而当压缩系数小于1时，为DenseNet-C结构。
• DenseNet-BC结构：对于使用bottleneck层的DenseBlock结构和压缩系数小于1的Transition组合结构。

5.DenseNet121

• DenseNet121模型中，每个Dense Block内部使用BN+ ReLU + 3x3
  Conv的组合函数，并且可能包含Bottleneck结构，即在3x3卷积前使用1x1卷积来减少特征图的维度，降低计算量 。即BN+ReLU+1x1Conv+BN+ReLU+3x3Conv。
• DenseNet121的一个关键参数是growth rate，它定义了每个Dense Layer输出的feature
  maps数量。在DenseNet121中，这个值被设置为32，意味着每个Dense Layer会为网络新增32个feature maps。
• 在Torchvision库中，DenseNet121可以通过torchvision.models.densenet121函数直接调用，并支持使用预训练权重。