>- **🍨 本文为[🔗365天深度学习训练营](https://mp.weixin.qq.com/s/0dvHCaOoFnW8SCp3JpzKxg) 中的学习记录博客**
>- **🍖 原作者:[K同学啊](https://mtyjkh.blog.csdn.net/)**
实战代码如下:
import pathlib
import torch
import torch.nn as nn
import torch.nn.functional as F
from torchvision import transforms, datasets
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(device)
data_dir = 'D:/研究生课题/深度学习-代码/4-date'
data_dir = pathlib.Path(data_dir)
data_paths = list(data_dir.glob('*'))
classeNames = [str(path).split("\\")[4] for path in data_paths]
print(classeNames)
image_count = len(list(data_dir.glob('*/*')))
print("图片总数为:", image_count)
train_transforms = 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。
])
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:/研究生课题/深度学习-代码/4-date", 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 = 4
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
from collections import OrderedDict
import torch.utils.checkpoint as cp
def _bn_function_factory(norm, relu, conv):
def bn_function(*inputs):
concated_features = torch.cat(inputs, 1)
bottleneck_output = conv(relu(norm(concated_features)))
return bottleneck_output
return bn_function
class _DenseLayer(nn.Module):
def __init__(self, num_input_features, growth_rate, bn_size, drop_rate, efficient=False):
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.add_module('SE_Block', SE_Block(growth_rate, reduction=16))
self.drop_rate = drop_rate
self.efficient = efficient
def forward(self, *prev_features):
bn_function = _bn_function_factory(self.norm1, self.relu1, self.conv1)
if self.efficient and any(prev_feature.requires_grad for prev_feature in prev_features):
bottleneck_output = cp.checkpoint(bn_function, *prev_features)
else:
bottleneck_output = bn_function(*prev_features)
new_features = self.SE_Block(self.conv2(self.relu2(self.norm2(bottleneck_output))))
if self.drop_rate > 0:
new_features = F.dropout(new_features, p=self.drop_rate, training=self.training)
return new_features
class _Transition(nn.Sequential):
def __init__(self, num_input_features, num_output_features):
super(_Transition, self).__init__()
self.add_module('norm', nn.BatchNorm2d(num_input_features))
self.add_module('relu', nn.ReLU(inplace=True))
self.add_module('conv', nn.Conv2d(num_input_features, num_output_features,
kernel_size=1, stride=1, bias=False))
self.add_module('pool', nn.AvgPool2d(kernel_size=2, stride=2))
class _DenseBlock(nn.Module):
def __init__(self, num_layers, num_input_features, bn_size, growth_rate, drop_rate, efficient=False):
super(_DenseBlock, self).__init__()
for i in range(num_layers):
layer = _DenseLayer(
num_input_features + i * growth_rate,
growth_rate=growth_rate,
bn_size=bn_size,
drop_rate=drop_rate,
efficient=efficient,
)
self.add_module('denselayer%d' % (i + 1), layer)
def forward(self, init_features):
features = [init_features]
for name, layer in self.named_children():
new_features = layer(*features)
features.append(new_features)
return torch.cat(features, 1)
class SE_Block(nn.Module):
def __init__(self, ch_in, reduction=16):
super(SE_Block, self).__init__()
self.avg_pool = nn.AdaptiveAvgPool2d(1) # 全局自适应池化
self.fc = nn.Sequential(
nn.Linear(ch_in, ch_in // reduction, bias=False),
nn.ReLU(inplace=True),
nn.Linear(ch_in // reduction, ch_in, bias=False),
nn.Sigmoid()
)
def forward(self, x):
b, c, _, _ = x.size()
y = self.avg_pool(x).view(b, c) # squeeze操作
y = self.fc(y).view(b, c, 1, 1) # FC获取通道注意力权重,是具有全局信息的
return x * y.expand_as(x) # 注意力作用每一个通道上
class DenseNet(nn.Module):
def __init__(self, growth_rate, block_config, num_init_features=24, compression=0.5, bn_size=4, drop_rate=0,
num_classes=10, small_inputs=True, efficient=False):
super(DenseNet, self).__init__()
assert 0 < compression <= 1, 'compression of densenet should be between 0 and 1'
# First convolution
if small_inputs:
self.features = nn.Sequential(OrderedDict([
('conv0', nn.Conv2d(3, num_init_features, kernel_size=3, stride=1, padding=1, bias=False)),
]))
else:
self.features = nn.Sequential(OrderedDict([
('conv0', nn.Conv2d(3, num_init_features, kernel_size=7, stride=2, padding=3, bias=False)),
]))
self.features.add_module('norm0', nn.BatchNorm2d(num_init_features))
self.features.add_module('relu0', nn.ReLU(inplace=True))
self.features.add_module('pool0', nn.MaxPool2d(kernel_size=3, stride=2, padding=1,
ceil_mode=False))
# Each denseblock
num_features = num_init_features
for i, num_layers in enumerate(block_config):
block = _DenseBlock(
num_layers=num_layers,
num_input_features=num_features,
bn_size=bn_size,
growth_rate=growth_rate,
drop_rate=drop_rate,
efficient=efficient,
)
self.features.add_module('denseblock%d' % (i + 1), block)
num_features = num_features + num_layers * growth_rate
if i != len(block_config) - 1:
trans = _Transition(num_input_features=num_features,
num_output_features=int(num_features * compression))
self.features.add_module('transition%d' % (i + 1), trans)
num_features = int(num_features * compression)
# self.features.add_module('SE_Block%d' % (i + 1),SE_Block(num_features, reduction=16))
# Final batch norm
self.features.add_module('norm_final', nn.BatchNorm2d(num_features))
# Linear layer
self.classifier = nn.Linear(num_features, num_classes)
def forward(self, x):
features = self.features(x)
out = F.relu(features, inplace=True)
out = F.adaptive_avg_pool2d(out, (1, 1))
out = torch.flatten(out, 1)
out = self.classifier(out)
return out
x = torch.randn(2, 3, 224, 224)
model = DenseNet(growth_rate=32, block_config=(6, 12, 24, 16), compression=0.5,
num_init_features=64, bn_size=4, drop_rate=0.2, num_classes=4, efficient=True)
out = model(x)
print('out.shape: ', out.shape)
print(out)
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))
# 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loss
train_acc += (pred.argmax(1) == y).type(torch.float).sum().item()
train_loss += loss.item()
train_acc /= size
train_loss /= num_batches
return train_acc, train_loss
def 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
import copy
optimizer = torch.optim.Adam(model.parameters(), lr=1e-4)
loss_fn = nn.CrossEntropyLoss() # 创建损失函数
epochs = 10
train_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_model
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)
# 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 = 'D:/pythonProject6/best_model.pth' # 保存的参数文件名
torch.save(model.state_dict(), PATH)
print('Done')
import matplotlib.pyplot as plt
# 隐藏警告
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()
from PIL import Image
classes = list(total_data.class_to_idx)
def predict_one_image(image_path, model, transform, classes):
test_img = Image.open(image_path).convert('RGB')
plt.imshow(test_img) # 展示预测的图片
test_img = transform(test_img)
img = test_img.to(device).unsqueeze(0)
model.eval()
output = model(img)
_, pred = torch.max(output, 1)
pred_class = classes[pred]
print(f'预测结果是:{pred_class}')
# 预测训练集中的某张照片
predict_one_image(image_path='D:/研究生课题/深度学习-代码/4-date/Monkeypox/M01_01_00.jpg',
model=model,
transform=train_transforms,
classes=classes)
运行结果如下:
C:\Users\dell\anaconda3\envs\pytorch-gpu\python.exe D:\pythonProject6\dates.py
cuda
['Monkeypox', 'Others']
图片总数为: 2142
{'Monkeypox': 0, 'Others': 1}
Shape of X [N, C, H, W]: torch.Size([4, 3, 224, 224])
Shape of y: torch.Size([4]) torch.int64
out.shape: torch.Size([2, 4])
tensor([[-0.0028, 0.5430, 0.2391, -0.3382],
[ 0.0040, 0.5535, 0.2722, -0.3856]], grad_fn=<AddmmBackward>)
----------------------------------------------------------------
Layer (type) Output Shape Param #
================================================================
Conv2d-1 [-1, 64, 224, 224] 1,728
BatchNorm2d-2 [-1, 64, 224, 224] 128
ReLU-3 [-1, 64, 224, 224] 0
Conv2d-4 [-1, 128, 224, 224] 8,192
BatchNorm2d-5 [-1, 128, 224, 224] 256
ReLU-6 [-1, 128, 224, 224] 0
Conv2d-7 [-1, 32, 224, 224] 36,864
AdaptiveAvgPool2d-8 [-1, 32, 1, 1] 0
Linear-9 [-1, 2] 64
ReLU-10 [-1, 2] 0
Linear-11 [-1, 32] 64
Sigmoid-12 [-1, 32] 0
SE_Block-13 [-1, 32, 224, 224] 0
_DenseLayer-14 [-1, 32, 224, 224] 0
BatchNorm2d-15 [-1, 96, 224, 224] 192
ReLU-16 [-1, 96, 224, 224] 0
Conv2d-17 [-1, 128, 224, 224] 12,288
BatchNorm2d-18 [-1, 128, 224, 224] 256
ReLU-19 [-1, 128, 224, 224] 0
Conv2d-20 [-1, 32, 224, 224] 36,864
AdaptiveAvgPool2d-21 [-1, 32, 1, 1] 0
Linear-22 [-1, 2] 64
ReLU-23 [-1, 2] 0
Linear-24 [-1, 32] 64
Sigmoid-25 [-1, 32] 0
SE_Block-26 [-1, 32, 224, 224] 0
_DenseLayer-27 [-1, 32, 224, 224] 0
BatchNorm2d-28 [-1, 128, 224, 224] 256
ReLU-29 [-1, 128, 224, 224] 0
Conv2d-30 [-1, 128, 224, 224] 16,384
BatchNorm2d-31 [-1, 128, 224, 224] 256
ReLU-32 [-1, 128, 224, 224] 0
Conv2d-33 [-1, 32, 224, 224] 36,864
AdaptiveAvgPool2d-34 [-1, 32, 1, 1] 0
Linear-35 [-1, 2] 64
ReLU-36 [-1, 2] 0
Linear-37 [-1, 32] 64
Sigmoid-38 [-1, 32] 0
SE_Block-39 [-1, 32, 224, 224] 0
_DenseLayer-40 [-1, 32, 224, 224] 0
BatchNorm2d-41 [-1, 160, 224, 224] 320
ReLU-42 [-1, 160, 224, 224] 0
Conv2d-43 [-1, 128, 224, 224] 20,480
BatchNorm2d-44 [-1, 128, 224, 224] 256
ReLU-45 [-1, 128, 224, 224] 0
Conv2d-46 [-1, 32, 224, 224] 36,864
AdaptiveAvgPool2d-47 [-1, 32, 1, 1] 0
Linear-48 [-1, 2] 64
ReLU-49 [-1, 2] 0
Linear-50 [-1, 32] 64
Sigmoid-51 [-1, 32] 0
SE_Block-52 [-1, 32, 224, 224] 0
_DenseLayer-53 [-1, 32, 224, 224] 0
BatchNorm2d-54 [-1, 192, 224, 224] 384
ReLU-55 [-1, 192, 224, 224] 0
Conv2d-56 [-1, 128, 224, 224] 24,576
BatchNorm2d-57 [-1, 128, 224, 224] 256
ReLU-58 [-1, 128, 224, 224] 0
Conv2d-59 [-1, 32, 224, 224] 36,864
AdaptiveAvgPool2d-60 [-1, 32, 1, 1] 0
Linear-61 [-1, 2] 64
ReLU-62 [-1, 2] 0
Linear-63 [-1, 32] 64
Sigmoid-64 [-1, 32] 0
SE_Block-65 [-1, 32, 224, 224] 0
_DenseLayer-66 [-1, 32, 224, 224] 0
BatchNorm2d-67 [-1, 224, 224, 224] 448
ReLU-68 [-1, 224, 224, 224] 0
Conv2d-69 [-1, 128, 224, 224] 28,672
BatchNorm2d-70 [-1, 128, 224, 224] 256
ReLU-71 [-1, 128, 224, 224] 0
Conv2d-72 [-1, 32, 224, 224] 36,864
AdaptiveAvgPool2d-73 [-1, 32, 1, 1] 0
Linear-74 [-1, 2] 64
ReLU-75 [-1, 2] 0
Linear-76 [-1, 32] 64
Sigmoid-77 [-1, 32] 0
SE_Block-78 [-1, 32, 224, 224] 0
_DenseLayer-79 [-1, 32, 224, 224] 0
_DenseBlock-80 [-1, 256, 224, 224] 0
BatchNorm2d-81 [-1, 256, 224, 224] 512
ReLU-82 [-1, 256, 224, 224] 0
Conv2d-83 [-1, 128, 224, 224] 32,768
AvgPool2d-84 [-1, 128, 112, 112] 0
BatchNorm2d-85 [-1, 128, 112, 112] 256
ReLU-86 [-1, 128, 112, 112] 0
Conv2d-87 [-1, 128, 112, 112] 16,384
BatchNorm2d-88 [-1, 128, 112, 112] 256
ReLU-89 [-1, 128, 112, 112] 0
Conv2d-90 [-1, 32, 112, 112] 36,864
AdaptiveAvgPool2d-91 [-1, 32, 1, 1] 0
Linear-92 [-1, 2] 64
ReLU-93 [-1, 2] 0
Linear-94 [-1, 32] 64
Sigmoid-95 [-1, 32] 0
SE_Block-96 [-1, 32, 112, 112] 0
_DenseLayer-97 [-1, 32, 112, 112] 0
BatchNorm2d-98 [-1, 160, 112, 112] 320
ReLU-99 [-1, 160, 112, 112] 0
Conv2d-100 [-1, 128, 112, 112] 20,480
BatchNorm2d-101 [-1, 128, 112, 112] 256
ReLU-102 [-1, 128, 112, 112] 0
Conv2d-103 [-1, 32, 112, 112] 36,864
AdaptiveAvgPool2d-104 [-1, 32, 1, 1] 0
Linear-105 [-1, 2] 64
ReLU-106 [-1, 2] 0
Linear-107 [-1, 32] 64
Sigmoid-108 [-1, 32] 0
SE_Block-109 [-1, 32, 112, 112] 0
_DenseLayer-110 [-1, 32, 112, 112] 0
BatchNorm2d-111 [-1, 192, 112, 112] 384
ReLU-112 [-1, 192, 112, 112] 0
Conv2d-113 [-1, 128, 112, 112] 24,576
BatchNorm2d-114 [-1, 128, 112, 112] 256
ReLU-115 [-1, 128, 112, 112] 0
Conv2d-116 [-1, 32, 112, 112] 36,864
AdaptiveAvgPool2d-117 [-1, 32, 1, 1] 0
Linear-118 [-1, 2] 64
ReLU-119 [-1, 2] 0
Linear-120 [-1, 32] 64
Sigmoid-121 [-1, 32] 0
SE_Block-122 [-1, 32, 112, 112] 0
_DenseLayer-123 [-1, 32, 112, 112] 0
BatchNorm2d-124 [-1, 224, 112, 112] 448
ReLU-125 [-1, 224, 112, 112] 0
Conv2d-126 [-1, 128, 112, 112] 28,672
BatchNorm2d-127 [-1, 128, 112, 112] 256
ReLU-128 [-1, 128, 112, 112] 0
Conv2d-129 [-1, 32, 112, 112] 36,864
AdaptiveAvgPool2d-130 [-1, 32, 1, 1] 0
Linear-131 [-1, 2] 64
ReLU-132 [-1, 2] 0
Linear-133 [-1, 32] 64
Sigmoid-134 [-1, 32] 0
SE_Block-135 [-1, 32, 112, 112] 0
_DenseLayer-136 [-1, 32, 112, 112] 0
BatchNorm2d-137 [-1, 256, 112, 112] 512
ReLU-138 [-1, 256, 112, 112] 0
Conv2d-139 [-1, 128, 112, 112] 32,768
BatchNorm2d-140 [-1, 128, 112, 112] 256
ReLU-141 [-1, 128, 112, 112] 0
Conv2d-142 [-1, 32, 112, 112] 36,864
AdaptiveAvgPool2d-143 [-1, 32, 1, 1] 0
Linear-144 [-1, 2] 64
ReLU-145 [-1, 2] 0
Linear-146 [-1, 32] 64
Sigmoid-147 [-1, 32] 0
SE_Block-148 [-1, 32, 112, 112] 0
_DenseLayer-149 [-1, 32, 112, 112] 0
BatchNorm2d-150 [-1, 288, 112, 112] 576
ReLU-151 [-1, 288, 112, 112] 0
Conv2d-152 [-1, 128, 112, 112] 36,864
BatchNorm2d-153 [-1, 128, 112, 112] 256
ReLU-154 [-1, 128, 112, 112] 0
Conv2d-155 [-1, 32, 112, 112] 36,864
AdaptiveAvgPool2d-156 [-1, 32, 1, 1] 0
Linear-157 [-1, 2] 64
ReLU-158 [-1, 2] 0
Linear-159 [-1, 32] 64
Sigmoid-160 [-1, 32] 0
SE_Block-161 [-1, 32, 112, 112] 0
_DenseLayer-162 [-1, 32, 112, 112] 0
BatchNorm2d-163 [-1, 320, 112, 112] 640
ReLU-164 [-1, 320, 112, 112] 0
Conv2d-165 [-1, 128, 112, 112] 40,960
BatchNorm2d-166 [-1, 128, 112, 112] 256
ReLU-167 [-1, 128, 112, 112] 0
Conv2d-168 [-1, 32, 112, 112] 36,864
AdaptiveAvgPool2d-169 [-1, 32, 1, 1] 0
Linear-170 [-1, 2] 64
ReLU-171 [-1, 2] 0
Linear-172 [-1, 32] 64
Sigmoid-173 [-1, 32] 0
SE_Block-174 [-1, 32, 112, 112] 0
_DenseLayer-175 [-1, 32, 112, 112] 0
BatchNorm2d-176 [-1, 352, 112, 112] 704
ReLU-177 [-1, 352, 112, 112] 0
Conv2d-178 [-1, 128, 112, 112] 45,056
BatchNorm2d-179 [-1, 128, 112, 112] 256
ReLU-180 [-1, 128, 112, 112] 0
Conv2d-181 [-1, 32, 112, 112] 36,864
AdaptiveAvgPool2d-182 [-1, 32, 1, 1] 0
Linear-183 [-1, 2] 64
ReLU-184 [-1, 2] 0
Linear-185 [-1, 32] 64
Sigmoid-186 [-1, 32] 0
SE_Block-187 [-1, 32, 112, 112] 0
_DenseLayer-188 [-1, 32, 112, 112] 0
BatchNorm2d-189 [-1, 384, 112, 112] 768
ReLU-190 [-1, 384, 112, 112] 0
Conv2d-191 [-1, 128, 112, 112] 49,152
BatchNorm2d-192 [-1, 128, 112, 112] 256
ReLU-193 [-1, 128, 112, 112] 0
Conv2d-194 [-1, 32, 112, 112] 36,864
AdaptiveAvgPool2d-195 [-1, 32, 1, 1] 0
Linear-196 [-1, 2] 64
ReLU-197 [-1, 2] 0
Linear-198 [-1, 32] 64
Sigmoid-199 [-1, 32] 0
SE_Block-200 [-1, 32, 112, 112] 0
_DenseLayer-201 [-1, 32, 112, 112] 0
BatchNorm2d-202 [-1, 416, 112, 112] 832
ReLU-203 [-1, 416, 112, 112] 0
Conv2d-204 [-1, 128, 112, 112] 53,248
BatchNorm2d-205 [-1, 128, 112, 112] 256
ReLU-206 [-1, 128, 112, 112] 0
Conv2d-207 [-1, 32, 112, 112] 36,864
AdaptiveAvgPool2d-208 [-1, 32, 1, 1] 0
Linear-209 [-1, 2] 64
ReLU-210 [-1, 2] 0
Linear-211 [-1, 32] 64
Sigmoid-212 [-1, 32] 0
SE_Block-213 [-1, 32, 112, 112] 0
_DenseLayer-214 [-1, 32, 112, 112] 0
BatchNorm2d-215 [-1, 448, 112, 112] 896
ReLU-216 [-1, 448, 112, 112] 0
Conv2d-217 [-1, 128, 112, 112] 57,344
BatchNorm2d-218 [-1, 128, 112, 112] 256
ReLU-219 [-1, 128, 112, 112] 0
Conv2d-220 [-1, 32, 112, 112] 36,864
AdaptiveAvgPool2d-221 [-1, 32, 1, 1] 0
Linear-222 [-1, 2] 64
ReLU-223 [-1, 2] 0
Linear-224 [-1, 32] 64
Sigmoid-225 [-1, 32] 0
SE_Block-226 [-1, 32, 112, 112] 0
_DenseLayer-227 [-1, 32, 112, 112] 0
BatchNorm2d-228 [-1, 480, 112, 112] 960
ReLU-229 [-1, 480, 112, 112] 0
Conv2d-230 [-1, 128, 112, 112] 61,440
BatchNorm2d-231 [-1, 128, 112, 112] 256
ReLU-232 [-1, 128, 112, 112] 0
Conv2d-233 [-1, 32, 112, 112] 36,864
AdaptiveAvgPool2d-234 [-1, 32, 1, 1] 0
Linear-235 [-1, 2] 64
ReLU-236 [-1, 2] 0
Linear-237 [-1, 32] 64
Sigmoid-238 [-1, 32] 0
SE_Block-239 [-1, 32, 112, 112] 0
_DenseLayer-240 [-1, 32, 112, 112] 0
_DenseBlock-241 [-1, 512, 112, 112] 0
BatchNorm2d-242 [-1, 512, 112, 112] 1,024
ReLU-243 [-1, 512, 112, 112] 0
Conv2d-244 [-1, 256, 112, 112] 131,072
AvgPool2d-245 [-1, 256, 56, 56] 0
BatchNorm2d-246 [-1, 256, 56, 56] 512
ReLU-247 [-1, 256, 56, 56] 0
Conv2d-248 [-1, 128, 56, 56] 32,768
BatchNorm2d-249 [-1, 128, 56, 56] 256
ReLU-250 [-1, 128, 56, 56] 0
Conv2d-251 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-252 [-1, 32, 1, 1] 0
Linear-253 [-1, 2] 64
ReLU-254 [-1, 2] 0
Linear-255 [-1, 32] 64
Sigmoid-256 [-1, 32] 0
SE_Block-257 [-1, 32, 56, 56] 0
_DenseLayer-258 [-1, 32, 56, 56] 0
BatchNorm2d-259 [-1, 288, 56, 56] 576
ReLU-260 [-1, 288, 56, 56] 0
Conv2d-261 [-1, 128, 56, 56] 36,864
BatchNorm2d-262 [-1, 128, 56, 56] 256
ReLU-263 [-1, 128, 56, 56] 0
Conv2d-264 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-265 [-1, 32, 1, 1] 0
Linear-266 [-1, 2] 64
ReLU-267 [-1, 2] 0
Linear-268 [-1, 32] 64
Sigmoid-269 [-1, 32] 0
SE_Block-270 [-1, 32, 56, 56] 0
_DenseLayer-271 [-1, 32, 56, 56] 0
BatchNorm2d-272 [-1, 320, 56, 56] 640
ReLU-273 [-1, 320, 56, 56] 0
Conv2d-274 [-1, 128, 56, 56] 40,960
BatchNorm2d-275 [-1, 128, 56, 56] 256
ReLU-276 [-1, 128, 56, 56] 0
Conv2d-277 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-278 [-1, 32, 1, 1] 0
Linear-279 [-1, 2] 64
ReLU-280 [-1, 2] 0
Linear-281 [-1, 32] 64
Sigmoid-282 [-1, 32] 0
SE_Block-283 [-1, 32, 56, 56] 0
_DenseLayer-284 [-1, 32, 56, 56] 0
BatchNorm2d-285 [-1, 352, 56, 56] 704
ReLU-286 [-1, 352, 56, 56] 0
Conv2d-287 [-1, 128, 56, 56] 45,056
BatchNorm2d-288 [-1, 128, 56, 56] 256
ReLU-289 [-1, 128, 56, 56] 0
Conv2d-290 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-291 [-1, 32, 1, 1] 0
Linear-292 [-1, 2] 64
ReLU-293 [-1, 2] 0
Linear-294 [-1, 32] 64
Sigmoid-295 [-1, 32] 0
SE_Block-296 [-1, 32, 56, 56] 0
_DenseLayer-297 [-1, 32, 56, 56] 0
BatchNorm2d-298 [-1, 384, 56, 56] 768
ReLU-299 [-1, 384, 56, 56] 0
Conv2d-300 [-1, 128, 56, 56] 49,152
BatchNorm2d-301 [-1, 128, 56, 56] 256
ReLU-302 [-1, 128, 56, 56] 0
Conv2d-303 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-304 [-1, 32, 1, 1] 0
Linear-305 [-1, 2] 64
ReLU-306 [-1, 2] 0
Linear-307 [-1, 32] 64
Sigmoid-308 [-1, 32] 0
SE_Block-309 [-1, 32, 56, 56] 0
_DenseLayer-310 [-1, 32, 56, 56] 0
BatchNorm2d-311 [-1, 416, 56, 56] 832
ReLU-312 [-1, 416, 56, 56] 0
Conv2d-313 [-1, 128, 56, 56] 53,248
BatchNorm2d-314 [-1, 128, 56, 56] 256
ReLU-315 [-1, 128, 56, 56] 0
Conv2d-316 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-317 [-1, 32, 1, 1] 0
Linear-318 [-1, 2] 64
ReLU-319 [-1, 2] 0
Linear-320 [-1, 32] 64
Sigmoid-321 [-1, 32] 0
SE_Block-322 [-1, 32, 56, 56] 0
_DenseLayer-323 [-1, 32, 56, 56] 0
BatchNorm2d-324 [-1, 448, 56, 56] 896
ReLU-325 [-1, 448, 56, 56] 0
Conv2d-326 [-1, 128, 56, 56] 57,344
BatchNorm2d-327 [-1, 128, 56, 56] 256
ReLU-328 [-1, 128, 56, 56] 0
Conv2d-329 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-330 [-1, 32, 1, 1] 0
Linear-331 [-1, 2] 64
ReLU-332 [-1, 2] 0
Linear-333 [-1, 32] 64
Sigmoid-334 [-1, 32] 0
SE_Block-335 [-1, 32, 56, 56] 0
_DenseLayer-336 [-1, 32, 56, 56] 0
BatchNorm2d-337 [-1, 480, 56, 56] 960
ReLU-338 [-1, 480, 56, 56] 0
Conv2d-339 [-1, 128, 56, 56] 61,440
BatchNorm2d-340 [-1, 128, 56, 56] 256
ReLU-341 [-1, 128, 56, 56] 0
Conv2d-342 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-343 [-1, 32, 1, 1] 0
Linear-344 [-1, 2] 64
ReLU-345 [-1, 2] 0
Linear-346 [-1, 32] 64
Sigmoid-347 [-1, 32] 0
SE_Block-348 [-1, 32, 56, 56] 0
_DenseLayer-349 [-1, 32, 56, 56] 0
BatchNorm2d-350 [-1, 512, 56, 56] 1,024
ReLU-351 [-1, 512, 56, 56] 0
Conv2d-352 [-1, 128, 56, 56] 65,536
BatchNorm2d-353 [-1, 128, 56, 56] 256
ReLU-354 [-1, 128, 56, 56] 0
Conv2d-355 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-356 [-1, 32, 1, 1] 0
Linear-357 [-1, 2] 64
ReLU-358 [-1, 2] 0
Linear-359 [-1, 32] 64
Sigmoid-360 [-1, 32] 0
SE_Block-361 [-1, 32, 56, 56] 0
_DenseLayer-362 [-1, 32, 56, 56] 0
BatchNorm2d-363 [-1, 544, 56, 56] 1,088
ReLU-364 [-1, 544, 56, 56] 0
Conv2d-365 [-1, 128, 56, 56] 69,632
BatchNorm2d-366 [-1, 128, 56, 56] 256
ReLU-367 [-1, 128, 56, 56] 0
Conv2d-368 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-369 [-1, 32, 1, 1] 0
Linear-370 [-1, 2] 64
ReLU-371 [-1, 2] 0
Linear-372 [-1, 32] 64
Sigmoid-373 [-1, 32] 0
SE_Block-374 [-1, 32, 56, 56] 0
_DenseLayer-375 [-1, 32, 56, 56] 0
BatchNorm2d-376 [-1, 576, 56, 56] 1,152
ReLU-377 [-1, 576, 56, 56] 0
Conv2d-378 [-1, 128, 56, 56] 73,728
BatchNorm2d-379 [-1, 128, 56, 56] 256
ReLU-380 [-1, 128, 56, 56] 0
Conv2d-381 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-382 [-1, 32, 1, 1] 0
Linear-383 [-1, 2] 64
ReLU-384 [-1, 2] 0
Linear-385 [-1, 32] 64
Sigmoid-386 [-1, 32] 0
SE_Block-387 [-1, 32, 56, 56] 0
_DenseLayer-388 [-1, 32, 56, 56] 0
BatchNorm2d-389 [-1, 608, 56, 56] 1,216
ReLU-390 [-1, 608, 56, 56] 0
Conv2d-391 [-1, 128, 56, 56] 77,824
BatchNorm2d-392 [-1, 128, 56, 56] 256
ReLU-393 [-1, 128, 56, 56] 0
Conv2d-394 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-395 [-1, 32, 1, 1] 0
Linear-396 [-1, 2] 64
ReLU-397 [-1, 2] 0
Linear-398 [-1, 32] 64
Sigmoid-399 [-1, 32] 0
SE_Block-400 [-1, 32, 56, 56] 0
_DenseLayer-401 [-1, 32, 56, 56] 0
BatchNorm2d-402 [-1, 640, 56, 56] 1,280
ReLU-403 [-1, 640, 56, 56] 0
Conv2d-404 [-1, 128, 56, 56] 81,920
BatchNorm2d-405 [-1, 128, 56, 56] 256
ReLU-406 [-1, 128, 56, 56] 0
Conv2d-407 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-408 [-1, 32, 1, 1] 0
Linear-409 [-1, 2] 64
ReLU-410 [-1, 2] 0
Linear-411 [-1, 32] 64
Sigmoid-412 [-1, 32] 0
SE_Block-413 [-1, 32, 56, 56] 0
_DenseLayer-414 [-1, 32, 56, 56] 0
BatchNorm2d-415 [-1, 672, 56, 56] 1,344
ReLU-416 [-1, 672, 56, 56] 0
Conv2d-417 [-1, 128, 56, 56] 86,016
BatchNorm2d-418 [-1, 128, 56, 56] 256
ReLU-419 [-1, 128, 56, 56] 0
Conv2d-420 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-421 [-1, 32, 1, 1] 0
Linear-422 [-1, 2] 64
ReLU-423 [-1, 2] 0
Linear-424 [-1, 32] 64
Sigmoid-425 [-1, 32] 0
SE_Block-426 [-1, 32, 56, 56] 0
_DenseLayer-427 [-1, 32, 56, 56] 0
BatchNorm2d-428 [-1, 704, 56, 56] 1,408
ReLU-429 [-1, 704, 56, 56] 0
Conv2d-430 [-1, 128, 56, 56] 90,112
BatchNorm2d-431 [-1, 128, 56, 56] 256
ReLU-432 [-1, 128, 56, 56] 0
Conv2d-433 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-434 [-1, 32, 1, 1] 0
Linear-435 [-1, 2] 64
ReLU-436 [-1, 2] 0
Linear-437 [-1, 32] 64
Sigmoid-438 [-1, 32] 0
SE_Block-439 [-1, 32, 56, 56] 0
_DenseLayer-440 [-1, 32, 56, 56] 0
BatchNorm2d-441 [-1, 736, 56, 56] 1,472
ReLU-442 [-1, 736, 56, 56] 0
Conv2d-443 [-1, 128, 56, 56] 94,208
BatchNorm2d-444 [-1, 128, 56, 56] 256
ReLU-445 [-1, 128, 56, 56] 0
Conv2d-446 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-447 [-1, 32, 1, 1] 0
Linear-448 [-1, 2] 64
ReLU-449 [-1, 2] 0
Linear-450 [-1, 32] 64
Sigmoid-451 [-1, 32] 0
SE_Block-452 [-1, 32, 56, 56] 0
_DenseLayer-453 [-1, 32, 56, 56] 0
BatchNorm2d-454 [-1, 768, 56, 56] 1,536
ReLU-455 [-1, 768, 56, 56] 0
Conv2d-456 [-1, 128, 56, 56] 98,304
BatchNorm2d-457 [-1, 128, 56, 56] 256
ReLU-458 [-1, 128, 56, 56] 0
Conv2d-459 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-460 [-1, 32, 1, 1] 0
Linear-461 [-1, 2] 64
ReLU-462 [-1, 2] 0
Linear-463 [-1, 32] 64
Sigmoid-464 [-1, 32] 0
SE_Block-465 [-1, 32, 56, 56] 0
_DenseLayer-466 [-1, 32, 56, 56] 0
BatchNorm2d-467 [-1, 800, 56, 56] 1,600
ReLU-468 [-1, 800, 56, 56] 0
Conv2d-469 [-1, 128, 56, 56] 102,400
BatchNorm2d-470 [-1, 128, 56, 56] 256
ReLU-471 [-1, 128, 56, 56] 0
Conv2d-472 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-473 [-1, 32, 1, 1] 0
Linear-474 [-1, 2] 64
ReLU-475 [-1, 2] 0
Linear-476 [-1, 32] 64
Sigmoid-477 [-1, 32] 0
SE_Block-478 [-1, 32, 56, 56] 0
_DenseLayer-479 [-1, 32, 56, 56] 0
BatchNorm2d-480 [-1, 832, 56, 56] 1,664
ReLU-481 [-1, 832, 56, 56] 0
Conv2d-482 [-1, 128, 56, 56] 106,496
BatchNorm2d-483 [-1, 128, 56, 56] 256
ReLU-484 [-1, 128, 56, 56] 0
Conv2d-485 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-486 [-1, 32, 1, 1] 0
Linear-487 [-1, 2] 64
ReLU-488 [-1, 2] 0
Linear-489 [-1, 32] 64
Sigmoid-490 [-1, 32] 0
SE_Block-491 [-1, 32, 56, 56] 0
_DenseLayer-492 [-1, 32, 56, 56] 0
BatchNorm2d-493 [-1, 864, 56, 56] 1,728
ReLU-494 [-1, 864, 56, 56] 0
Conv2d-495 [-1, 128, 56, 56] 110,592
BatchNorm2d-496 [-1, 128, 56, 56] 256
ReLU-497 [-1, 128, 56, 56] 0
Conv2d-498 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-499 [-1, 32, 1, 1] 0
Linear-500 [-1, 2] 64
ReLU-501 [-1, 2] 0
Linear-502 [-1, 32] 64
Sigmoid-503 [-1, 32] 0
SE_Block-504 [-1, 32, 56, 56] 0
_DenseLayer-505 [-1, 32, 56, 56] 0
BatchNorm2d-506 [-1, 896, 56, 56] 1,792
ReLU-507 [-1, 896, 56, 56] 0
Conv2d-508 [-1, 128, 56, 56] 114,688
BatchNorm2d-509 [-1, 128, 56, 56] 256
ReLU-510 [-1, 128, 56, 56] 0
Conv2d-511 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-512 [-1, 32, 1, 1] 0
Linear-513 [-1, 2] 64
ReLU-514 [-1, 2] 0
Linear-515 [-1, 32] 64
Sigmoid-516 [-1, 32] 0
SE_Block-517 [-1, 32, 56, 56] 0
_DenseLayer-518 [-1, 32, 56, 56] 0
BatchNorm2d-519 [-1, 928, 56, 56] 1,856
ReLU-520 [-1, 928, 56, 56] 0
Conv2d-521 [-1, 128, 56, 56] 118,784
BatchNorm2d-522 [-1, 128, 56, 56] 256
ReLU-523 [-1, 128, 56, 56] 0
Conv2d-524 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-525 [-1, 32, 1, 1] 0
Linear-526 [-1, 2] 64
ReLU-527 [-1, 2] 0
Linear-528 [-1, 32] 64
Sigmoid-529 [-1, 32] 0
SE_Block-530 [-1, 32, 56, 56] 0
_DenseLayer-531 [-1, 32, 56, 56] 0
BatchNorm2d-532 [-1, 960, 56, 56] 1,920
ReLU-533 [-1, 960, 56, 56] 0
Conv2d-534 [-1, 128, 56, 56] 122,880
BatchNorm2d-535 [-1, 128, 56, 56] 256
ReLU-536 [-1, 128, 56, 56] 0
Conv2d-537 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-538 [-1, 32, 1, 1] 0
Linear-539 [-1, 2] 64
ReLU-540 [-1, 2] 0
Linear-541 [-1, 32] 64
Sigmoid-542 [-1, 32] 0
SE_Block-543 [-1, 32, 56, 56] 0
_DenseLayer-544 [-1, 32, 56, 56] 0
BatchNorm2d-545 [-1, 992, 56, 56] 1,984
ReLU-546 [-1, 992, 56, 56] 0
Conv2d-547 [-1, 128, 56, 56] 126,976
BatchNorm2d-548 [-1, 128, 56, 56] 256
ReLU-549 [-1, 128, 56, 56] 0
Conv2d-550 [-1, 32, 56, 56] 36,864
AdaptiveAvgPool2d-551 [-1, 32, 1, 1] 0
Linear-552 [-1, 2] 64
ReLU-553 [-1, 2] 0
Linear-554 [-1, 32] 64
Sigmoid-555 [-1, 32] 0
SE_Block-556 [-1, 32, 56, 56] 0
_DenseLayer-557 [-1, 32, 56, 56] 0
_DenseBlock-558 [-1, 1024, 56, 56] 0
BatchNorm2d-559 [-1, 1024, 56, 56] 2,048
ReLU-560 [-1, 1024, 56, 56] 0
Conv2d-561 [-1, 512, 56, 56] 524,288
AvgPool2d-562 [-1, 512, 28, 28] 0
BatchNorm2d-563 [-1, 512, 28, 28] 1,024
ReLU-564 [-1, 512, 28, 28] 0
Conv2d-565 [-1, 128, 28, 28] 65,536
BatchNorm2d-566 [-1, 128, 28, 28] 256
ReLU-567 [-1, 128, 28, 28] 0
Conv2d-568 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-569 [-1, 32, 1, 1] 0
Linear-570 [-1, 2] 64
ReLU-571 [-1, 2] 0
Linear-572 [-1, 32] 64
Sigmoid-573 [-1, 32] 0
SE_Block-574 [-1, 32, 28, 28] 0
_DenseLayer-575 [-1, 32, 28, 28] 0
BatchNorm2d-576 [-1, 544, 28, 28] 1,088
ReLU-577 [-1, 544, 28, 28] 0
Conv2d-578 [-1, 128, 28, 28] 69,632
BatchNorm2d-579 [-1, 128, 28, 28] 256
ReLU-580 [-1, 128, 28, 28] 0
Conv2d-581 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-582 [-1, 32, 1, 1] 0
Linear-583 [-1, 2] 64
ReLU-584 [-1, 2] 0
Linear-585 [-1, 32] 64
Sigmoid-586 [-1, 32] 0
SE_Block-587 [-1, 32, 28, 28] 0
_DenseLayer-588 [-1, 32, 28, 28] 0
BatchNorm2d-589 [-1, 576, 28, 28] 1,152
ReLU-590 [-1, 576, 28, 28] 0
Conv2d-591 [-1, 128, 28, 28] 73,728
BatchNorm2d-592 [-1, 128, 28, 28] 256
ReLU-593 [-1, 128, 28, 28] 0
Conv2d-594 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-595 [-1, 32, 1, 1] 0
Linear-596 [-1, 2] 64
ReLU-597 [-1, 2] 0
Linear-598 [-1, 32] 64
Sigmoid-599 [-1, 32] 0
SE_Block-600 [-1, 32, 28, 28] 0
_DenseLayer-601 [-1, 32, 28, 28] 0
BatchNorm2d-602 [-1, 608, 28, 28] 1,216
ReLU-603 [-1, 608, 28, 28] 0
Conv2d-604 [-1, 128, 28, 28] 77,824
BatchNorm2d-605 [-1, 128, 28, 28] 256
ReLU-606 [-1, 128, 28, 28] 0
Conv2d-607 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-608 [-1, 32, 1, 1] 0
Linear-609 [-1, 2] 64
ReLU-610 [-1, 2] 0
Linear-611 [-1, 32] 64
Sigmoid-612 [-1, 32] 0
SE_Block-613 [-1, 32, 28, 28] 0
_DenseLayer-614 [-1, 32, 28, 28] 0
BatchNorm2d-615 [-1, 640, 28, 28] 1,280
ReLU-616 [-1, 640, 28, 28] 0
Conv2d-617 [-1, 128, 28, 28] 81,920
BatchNorm2d-618 [-1, 128, 28, 28] 256
ReLU-619 [-1, 128, 28, 28] 0
Conv2d-620 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-621 [-1, 32, 1, 1] 0
Linear-622 [-1, 2] 64
ReLU-623 [-1, 2] 0
Linear-624 [-1, 32] 64
Sigmoid-625 [-1, 32] 0
SE_Block-626 [-1, 32, 28, 28] 0
_DenseLayer-627 [-1, 32, 28, 28] 0
BatchNorm2d-628 [-1, 672, 28, 28] 1,344
ReLU-629 [-1, 672, 28, 28] 0
Conv2d-630 [-1, 128, 28, 28] 86,016
BatchNorm2d-631 [-1, 128, 28, 28] 256
ReLU-632 [-1, 128, 28, 28] 0
Conv2d-633 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-634 [-1, 32, 1, 1] 0
Linear-635 [-1, 2] 64
ReLU-636 [-1, 2] 0
Linear-637 [-1, 32] 64
Sigmoid-638 [-1, 32] 0
SE_Block-639 [-1, 32, 28, 28] 0
_DenseLayer-640 [-1, 32, 28, 28] 0
BatchNorm2d-641 [-1, 704, 28, 28] 1,408
ReLU-642 [-1, 704, 28, 28] 0
Conv2d-643 [-1, 128, 28, 28] 90,112
BatchNorm2d-644 [-1, 128, 28, 28] 256
ReLU-645 [-1, 128, 28, 28] 0
Conv2d-646 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-647 [-1, 32, 1, 1] 0
Linear-648 [-1, 2] 64
ReLU-649 [-1, 2] 0
Linear-650 [-1, 32] 64
Sigmoid-651 [-1, 32] 0
SE_Block-652 [-1, 32, 28, 28] 0
_DenseLayer-653 [-1, 32, 28, 28] 0
BatchNorm2d-654 [-1, 736, 28, 28] 1,472
ReLU-655 [-1, 736, 28, 28] 0
Conv2d-656 [-1, 128, 28, 28] 94,208
BatchNorm2d-657 [-1, 128, 28, 28] 256
ReLU-658 [-1, 128, 28, 28] 0
Conv2d-659 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-660 [-1, 32, 1, 1] 0
Linear-661 [-1, 2] 64
ReLU-662 [-1, 2] 0
Linear-663 [-1, 32] 64
Sigmoid-664 [-1, 32] 0
SE_Block-665 [-1, 32, 28, 28] 0
_DenseLayer-666 [-1, 32, 28, 28] 0
BatchNorm2d-667 [-1, 768, 28, 28] 1,536
ReLU-668 [-1, 768, 28, 28] 0
Conv2d-669 [-1, 128, 28, 28] 98,304
BatchNorm2d-670 [-1, 128, 28, 28] 256
ReLU-671 [-1, 128, 28, 28] 0
Conv2d-672 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-673 [-1, 32, 1, 1] 0
Linear-674 [-1, 2] 64
ReLU-675 [-1, 2] 0
Linear-676 [-1, 32] 64
Sigmoid-677 [-1, 32] 0
SE_Block-678 [-1, 32, 28, 28] 0
_DenseLayer-679 [-1, 32, 28, 28] 0
BatchNorm2d-680 [-1, 800, 28, 28] 1,600
ReLU-681 [-1, 800, 28, 28] 0
Conv2d-682 [-1, 128, 28, 28] 102,400
BatchNorm2d-683 [-1, 128, 28, 28] 256
ReLU-684 [-1, 128, 28, 28] 0
Conv2d-685 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-686 [-1, 32, 1, 1] 0
Linear-687 [-1, 2] 64
ReLU-688 [-1, 2] 0
Linear-689 [-1, 32] 64
Sigmoid-690 [-1, 32] 0
SE_Block-691 [-1, 32, 28, 28] 0
_DenseLayer-692 [-1, 32, 28, 28] 0
BatchNorm2d-693 [-1, 832, 28, 28] 1,664
ReLU-694 [-1, 832, 28, 28] 0
Conv2d-695 [-1, 128, 28, 28] 106,496
BatchNorm2d-696 [-1, 128, 28, 28] 256
ReLU-697 [-1, 128, 28, 28] 0
Conv2d-698 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-699 [-1, 32, 1, 1] 0
Linear-700 [-1, 2] 64
ReLU-701 [-1, 2] 0
Linear-702 [-1, 32] 64
Sigmoid-703 [-1, 32] 0
SE_Block-704 [-1, 32, 28, 28] 0
_DenseLayer-705 [-1, 32, 28, 28] 0
BatchNorm2d-706 [-1, 864, 28, 28] 1,728
ReLU-707 [-1, 864, 28, 28] 0
Conv2d-708 [-1, 128, 28, 28] 110,592
BatchNorm2d-709 [-1, 128, 28, 28] 256
ReLU-710 [-1, 128, 28, 28] 0
Conv2d-711 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-712 [-1, 32, 1, 1] 0
Linear-713 [-1, 2] 64
ReLU-714 [-1, 2] 0
Linear-715 [-1, 32] 64
Sigmoid-716 [-1, 32] 0
SE_Block-717 [-1, 32, 28, 28] 0
_DenseLayer-718 [-1, 32, 28, 28] 0
BatchNorm2d-719 [-1, 896, 28, 28] 1,792
ReLU-720 [-1, 896, 28, 28] 0
Conv2d-721 [-1, 128, 28, 28] 114,688
BatchNorm2d-722 [-1, 128, 28, 28] 256
ReLU-723 [-1, 128, 28, 28] 0
Conv2d-724 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-725 [-1, 32, 1, 1] 0
Linear-726 [-1, 2] 64
ReLU-727 [-1, 2] 0
Linear-728 [-1, 32] 64
Sigmoid-729 [-1, 32] 0
SE_Block-730 [-1, 32, 28, 28] 0
_DenseLayer-731 [-1, 32, 28, 28] 0
BatchNorm2d-732 [-1, 928, 28, 28] 1,856
ReLU-733 [-1, 928, 28, 28] 0
Conv2d-734 [-1, 128, 28, 28] 118,784
BatchNorm2d-735 [-1, 128, 28, 28] 256
ReLU-736 [-1, 128, 28, 28] 0
Conv2d-737 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-738 [-1, 32, 1, 1] 0
Linear-739 [-1, 2] 64
ReLU-740 [-1, 2] 0
Linear-741 [-1, 32] 64
Sigmoid-742 [-1, 32] 0
SE_Block-743 [-1, 32, 28, 28] 0
_DenseLayer-744 [-1, 32, 28, 28] 0
BatchNorm2d-745 [-1, 960, 28, 28] 1,920
ReLU-746 [-1, 960, 28, 28] 0
Conv2d-747 [-1, 128, 28, 28] 122,880
BatchNorm2d-748 [-1, 128, 28, 28] 256
ReLU-749 [-1, 128, 28, 28] 0
Conv2d-750 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-751 [-1, 32, 1, 1] 0
Linear-752 [-1, 2] 64
ReLU-753 [-1, 2] 0
Linear-754 [-1, 32] 64
Sigmoid-755 [-1, 32] 0
SE_Block-756 [-1, 32, 28, 28] 0
_DenseLayer-757 [-1, 32, 28, 28] 0
BatchNorm2d-758 [-1, 992, 28, 28] 1,984
ReLU-759 [-1, 992, 28, 28] 0
Conv2d-760 [-1, 128, 28, 28] 126,976
BatchNorm2d-761 [-1, 128, 28, 28] 256
ReLU-762 [-1, 128, 28, 28] 0
Conv2d-763 [-1, 32, 28, 28] 36,864
AdaptiveAvgPool2d-764 [-1, 32, 1, 1] 0
Linear-765 [-1, 2] 64
ReLU-766 [-1, 2] 0
Linear-767 [-1, 32] 64
Sigmoid-768 [-1, 32] 0
SE_Block-769 [-1, 32, 28, 28] 0
_DenseLayer-770 [-1, 32, 28, 28] 0
_DenseBlock-771 [-1, 1024, 28, 28] 0
BatchNorm2d-772 [-1, 1024, 28, 28] 2,048
Linear-773 [-1, 4] 4,100
================================================================
Total params: 6,957,572
Trainable params: 6,957,572
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.57
Forward/backward pass size (MB): 4854.11
Params size (MB): 26.54
Estimated Total Size (MB): 4881.22
----------------------------------------------------------------
Epoch: 1, Train_acc:57.5%, Train_loss:0.723, Test_acc:59.2%, Test_loss:0.682, Lr:1.00E-04
Epoch: 2, Train_acc:62.0%, Train_loss:0.664, Test_acc:67.4%, Test_loss:0.642, Lr:1.00E-04
Epoch: 3, Train_acc:61.9%, Train_loss:0.659, Test_acc:68.1%, Test_loss:0.662, Lr:1.00E-04
Epoch: 4, Train_acc:62.1%, Train_loss:0.656, Test_acc:61.1%, Test_loss:0.677, Lr:1.00E-04
Epoch: 5, Train_acc:64.6%, Train_loss:0.639, Test_acc:66.9%, Test_loss:0.613, Lr:1.00E-04
Epoch: 6, Train_acc:66.5%, Train_loss:0.626, Test_acc:68.1%, Test_loss:0.622, Lr:1.00E-04
Epoch: 7, Train_acc:65.8%, Train_loss:0.620, Test_acc:67.4%, Test_loss:0.642, Lr:1.00E-04
Epoch: 8, Train_acc:67.1%, Train_loss:0.606, Test_acc:68.3%, Test_loss:0.574, Lr:1.00E-04
Epoch: 9, Train_acc:67.3%, Train_loss:0.597, Test_acc:73.2%, Test_loss:0.523, Lr:1.00E-04
Epoch:10, Train_acc:70.2%, Train_loss:0.575, Test_acc:75.8%, Test_loss:0.496, Lr:1.00E-04
Done
预测结果是:Others
进程已结束,退出代码为 0
总结
结合DenseNet和SE-Net的实战经验显示,这种融合策略有效提升了深度学习模型在图像分类任务中的性能。DenseNet的密集连接促进了特征重用和梯度流动,而SE-Net的注意力机制则增强了对重要特征的建模能力。