深度学习第P9周：YOLOv5-Backbone模块实现
·
- 🍨 本文为🔗365天深度学习训练营中的学习记录博客
- 🍖 原作者：K同学啊
本周是学习深度学习的第9周。编译器使用的是vscode，安装的是CPU版PyTorch：
torch 2.12.0+cpu。
1 数据集
该过程与前边所学一样，不详细介绍。
1.1 导入数据
import os,PIL,random,pathlib
data_dir = './9-data/'
data_dir = pathlib.Path(data_dir)
data_paths = list(data_dir.glob('*'))
classeNames = [str(path).split("\\")[1] for path in data_paths]
classeNames
# 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("./9-data/",transform=train_transforms)
total_data结果显示为：
['cloudy', 'rain', 'shine', 'sunrise']
Dataset ImageFolder
Number of datapoints: 1125
Root location: ./9-data/
StandardTransform
Transform: Compose(
Resize(size=[224, 224], interpolation=bilinear, max_size=None, antialias=None)
ToTensor()
Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
)另：
total_data.class_to_idx结果：
{'cloudy': 0, 'rain': 1, 'shine': 2, 'sunrise': 3}可以看出该数据集共有4种类型。
1.2 数据集划分
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])
train_dataset, test_dataset结果：
(<torch.utils.data.dataset.Subset at 0x2bbb0779330>,
<torch.utils.data.dataset.Subset at 0x2bbb0779e10>)另：
batch_size = 4
train_dl = torch.utils.data.DataLoader(train_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=1)
test_dl = torch.utils.data.DataLoader(test_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=1)
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结果：
Shape of X [N, C, H, W]: torch.Size([4, 3, 224, 224])
Shape of y: torch.Size([4]) torch.int642 模型
本次利用YOLOv5算法中的Backbone模块搭建网络，YOLOv5是目标检测算法，是否可以尝试将其网络结构用在目标识别上，或进行改进形成一个全新的算法。
2.1 模型搭建
import torch.nn.functional as F
def autopad(k, p=None): # kernel, padding
# Pad to 'same'
if p is None:
p = k // 2 if isinstance(k, int) else [x // 2 for x in k] # auto-pad
return p
class Conv(nn.Module):
# Standard convolution
def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True): # ch_in, ch_out, kernel, stride, padding, groups
super().__init__()
self.conv = nn.Conv2d(c1, c2, k, s, autopad(k, p), groups=g, bias=False)
self.bn = nn.BatchNorm2d(c2)
self.act = nn.SiLU() if act is True else (act if isinstance(act, nn.Module) else nn.Identity())
def forward(self, x):
return self.act(self.bn(self.conv(x)))
class Bottleneck(nn.Module):
# Standard bottleneck
def __init__(self, c1, c2, shortcut=True, g=1, e=0.5): # ch_in, ch_out, shortcut, groups, expansion
super().__init__()
c_ = int(c2 * e) # hidden channels
self.cv1 = Conv(c1, c_, 1, 1)
self.cv2 = Conv(c_, c2, 3, 1, g=g)
self.add = shortcut and c1 == c2
def forward(self, x):
return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x))
class C3(nn.Module):
# CSP Bottleneck with 3 convolutions
def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5): # ch_in, ch_out, number, shortcut, groups, expansion
super().__init__()
c_ = int(c2 * e) # hidden channels
self.cv1 = Conv(c1, c_, 1, 1)
self.cv2 = Conv(c1, c_, 1, 1)
self.cv3 = Conv(2 * c_, c2, 1) # act=FReLU(c2)
self.m = nn.Sequential(*(Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)))
def forward(self, x):
return self.cv3(torch.cat((self.m(self.cv1(x)), self.cv2(x)), dim=1))
class SPPF(nn.Module):
# Spatial Pyramid Pooling - Fast (SPPF) layer for YOLOv5 by Glenn Jocher
def __init__(self, c1, c2, k=5): # equivalent to SPP(k=(5, 9, 13))
super().__init__()
c_ = c1 // 2 # hidden channels
self.cv1 = Conv(c1, c_, 1, 1)
self.cv2 = Conv(c_ * 4, c2, 1, 1)
self.m = nn.MaxPool2d(kernel_size=k, stride=1, padding=k // 2)
def forward(self, x):
x = self.cv1(x)
with warnings.catch_warnings():
warnings.simplefilter('ignore') # suppress torch 1.9.0 max_pool2d() warning
y1 = self.m(x)
y2 = self.m(y1)
return self.cv2(torch.cat([x, y1, y2, self.m(y2)], 1))
"""
这个是YOLOv5, 6.0版本的主干网络，这里进行复现
（注：有部分删改，详细讲解将在后续进行展开）
"""
class YOLOv5_backbone(nn.Module):
def __init__(self):
super(YOLOv5_backbone, self).__init__()
self.Conv_1 = Conv(3, 64, 3, 2, 2)
self.Conv_2 = Conv(64, 128, 3, 2)
self.C3_3 = C3(128,128)
self.Conv_4 = Conv(128, 256, 3, 2)
self.C3_5 = C3(256,256)
self.Conv_6 = Conv(256, 512, 3, 2)
self.C3_7 = C3(512,512)
self.Conv_8 = Conv(512, 1024, 3, 2)
self.C3_9 = C3(1024, 1024)
self.SPPF = SPPF(1024, 1024, 5)
# 全连接网络层，用于分类
self.classifier = nn.Sequential(
nn.Linear(in_features=65536, out_features=100),
nn.ReLU(),
nn.Linear(in_features=100, out_features=4)
)
def forward(self, x):
x = self.Conv_1(x)
x = self.Conv_2(x)
x = self.C3_3(x)
x = self.Conv_4(x)
x = self.C3_5(x)
x = self.Conv_6(x)
x = self.C3_7(x)
x = self.Conv_8(x)
x = self.C3_9(x)
x = self.SPPF(x)
x = torch.flatten(x, start_dim=1)
x = self.classifier(x)
return x
device = "cuda" if torch.cuda.is_available() else "cpu"
print("Using {} device".format(device))
model = YOLOv5_backbone().to(device)
model显示结果为：
Using cpu device
YOLOv5_backbone(
(Conv_1): Conv(
(conv): Conv2d(3, 64, kernel_size=(3, 3), stride=(2, 2), padding=(2, 2), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act): SiLU()
)
(Conv_2): Conv(
(conv): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act): SiLU()
)
......
(SPPF): SPPF(
(cv1): Conv(
(conv): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act): SiLU()
)
(cv2): Conv(
(conv): Conv2d(2048, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act): SiLU()
)
(m): MaxPool2d(kernel_size=5, stride=1, padding=2, dilation=1, ceil_mode=False)
)
(classifier): Sequential(
(0): Linear(in_features=65536, out_features=100, bias=True)
(1): ReLU()
(2): Linear(in_features=100, out_features=4, bias=True)
)
)2.2 查看模型
# 统计模型参数量以及其他指标
import torchsummary as summary
summary.summary(model, (3, 224, 224))此过程运行结果为：
----------------------------------------------------------------
Layer (type) Output Shape Param #
================================================================
Conv2d-1 [-1, 64, 113, 113] 1,728
BatchNorm2d-2 [-1, 64, 113, 113] 128
SiLU-3 [-1, 64, 113, 113] 0
Conv-4 [-1, 64, 113, 113] 0
Conv2d-5 [-1, 128, 57, 57] 73,728
BatchNorm2d-6 [-1, 128, 57, 57] 256
SiLU-7 [-1, 128, 57, 57] 0
Conv-8 [-1, 128, 57, 57] 0
Conv2d-9 [-1, 64, 57, 57] 8,192
......
BatchNorm2d-117 [-1, 1024, 8, 8] 2,048
SiLU-118 [-1, 1024, 8, 8] 0
Conv-119 [-1, 1024, 8, 8] 0
SPPF-120 [-1, 1024, 8, 8] 0
Linear-121 [-1, 100] 6,553,700
ReLU-122 [-1, 100] 0
Linear-123 [-1, 4] 404
================================================================
Total params: 21,729,592
Trainable params: 21,729,592
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.57
Forward/backward pass size (MB): 137.59
Params size (MB): 82.89
Estimated Total Size (MB): 221.06
----------------------------------------------------------------3 模型训练与可视化
3.1训练函数与测试函数
训练函数：
# 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_loss3.2 正式训练与可视化
import copy
optimizer = torch.optim.Adam(model.parameters(), lr= 1e-4)
loss_fn = nn.CrossEntropyLoss() # 创建损失函数
epochs = 60
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):
model.train()
epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, optimizer)
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 = './best_model.pth' # 保存的参数文件名
torch.save(best_model.state_dict(), PATH)
print('Done')可视化部分：
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
from datetime import datetime
current_time = datetime.now() # 获取当前时间
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.xlabel(current_time) # 打卡请带上时间戳，否则代码截图无效
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()3.3 结果显示
训练轮数结果：
Epoch: 1,Train_acc:54.1%,Train_loss:1.235,Test_test:65.2%,Test_loss:0.742,Lr:1.00E-04
Epoch: 2,Train_acc:63.0%,Train_loss:0.781,Test_test:78.6%,Test_loss:0.689,Lr:1.00E-04
Epoch: 3,Train_acc:71.8%,Train_loss:0.604,Test_test:81.4%,Test_loss:0.462,Lr:1.00E-04
Epoch: 4,Train_acc:74.2%,Train_loss:0.518,Test_test:77.0%,Test_loss:0.384,Lr:1.00E-04
Epoch: 5,Train_acc:78.9%,Train_loss:0.472,Test_test:84.1%,Test_loss:0.412,Lr:1.00E-04
Epoch: 6,Train_acc:80.1%,Train_loss:0.436,Test_test:82.7%,Test_loss:0.386,Lr:1.00E-04
Epoch: 7,Train_acc:82.6%,Train_loss:0.401,Test_test:85.2%,Test_loss:0.352,Lr:1.00E-04
Epoch: 8,Train_acc:83.7%,Train_loss:0.378,Test_test:83.6%,Test_loss:0.337,Lr:1.00E-04
Epoch: 9,Train_acc:85.4%,Train_loss:0.351,Test_test:86.8%,Test_loss:0.319,Lr:1.00E-04
Epoch:10,Train_acc:86.1%,Train_loss:0.329,Test_test:88.0%,Test_loss:0.281,Lr:1.00E-04
Epoch:11,Train_acc:87.2%,Train_loss:0.315,Test_test:76.4%,Test_loss:0.598,Lr:1.00E-04
Epoch:12,Train_acc:88.0%,Train_loss:0.294,Test_test:84.7%,Test_loss:0.417,Lr:1.00E-04
Epoch:13,Train_acc:88.9%,Train_loss:0.281,Test_test:86.2%,Test_loss:0.389,Lr:1.00E-04
Epoch:14,Train_acc:89.6%,Train_loss:0.263,Test_test:89.3%,Test_loss:0.355,Lr:1.00E-04
Epoch:15,Train_acc:91.3%,Train_loss:0.246,Test_test:77.8%,Test_loss:0.801,Lr:1.00E-04
Epoch:16,Train_acc:92.0%,Train_loss:0.231,Test_test:82.4%,Test_loss:0.604,Lr:1.00E-04
Epoch:17,Train_acc:91.8%,Train_loss:0.239,Test_test:87.5%,Test_loss:0.394,Lr:1.00E-04
Epoch:18,Train_acc:92.7%,Train_loss:0.218,Test_test:90.1%,Test_loss:0.356,Lr:1.00E-04
Epoch:19,Train_acc:93.4%,Train_loss:0.206,Test_test:88.7%,Test_loss:0.342,Lr:1.00E-04
Epoch:20,Train_acc:92.6%,Train_loss:0.225,Test_test:90.6%,Test_loss:0.301,Lr:1.00E-04
Epoch:21,Train_acc:93.8%,Train_loss:0.198,Test_test:89.9%,Test_loss:0.286,Lr:1.00E-04
Epoch:22,Train_acc:94.2%,Train_loss:0.186,Test_test:78.9%,Test_loss:0.564,Lr:1.00E-04
Epoch:23,Train_acc:93.7%,Train_loss:0.194,Test_test:87.6%,Test_loss:0.361,Lr:1.00E-04
Epoch:24,Train_acc:94.5%,Train_loss:0.177,Test_test:89.2%,Test_loss:0.335,Lr:1.00E-04
Epoch:25,Train_acc:95.0%,Train_loss:0.168,Test_test:90.4%,Test_loss:0.298,Lr:1.00E-04
Epoch:26,Train_acc:94.3%,Train_loss:0.179,Test_test:86.0%,Test_loss:0.422,Lr:1.00E-04
Epoch:27,Train_acc:94.8%,Train_loss:0.164,Test_test:88.1%,Test_loss:0.347,Lr:1.00E-04
Epoch:28,Train_acc:95.6%,Train_loss:0.149,Test_test:91.2%,Test_loss:0.278,Lr:1.00E-04
Epoch:29,Train_acc:95.2%,Train_loss:0.156,Test_test:92.0%,Test_loss:0.257,Lr:1.00E-04
Epoch:30,Train_acc:96.0%,Train_loss:0.143,Test_test:86.5%,Test_loss:0.425,Lr:1.00E-04
Epoch:31,Train_acc:95.8%,Train_loss:0.149,Test_test:89.8%,Test_loss:0.318,Lr:1.00E-04
Epoch:32,Train_acc:96.3%,Train_loss:0.134,Test_test:91.7%,Test_loss:0.264,Lr:1.00E-04
Epoch:33,Train_acc:96.1%,Train_loss:0.139,Test_test:92.2%,Test_loss:0.251,Lr:1.00E-04
Epoch:34,Train_acc:96.7%,Train_loss:0.126,Test_test:91.4%,Test_loss:0.271,Lr:1.00E-04
Epoch:35,Train_acc:97.0%,Train_loss:0.118,Test_test:93.1%,Test_loss:0.239,Lr:1.00E-04
Epoch:36,Train_acc:96.5%,Train_loss:0.127,Test_test:87.8%,Test_loss:0.408,Lr:1.00E-04
Epoch:37,Train_acc:96.8%,Train_loss:0.121,Test_test:90.6%,Test_loss:0.301,Lr:1.00E-04
Epoch:38,Train_acc:97.4%,Train_loss:0.109,Test_test:92.5%,Test_loss:0.243,Lr:1.00E-04
Epoch:39,Train_acc:97.2%,Train_loss:0.113,Test_test:91.1%,Test_loss:0.287,Lr:1.00E-04
Epoch:40,Train_acc:97.6%,Train_loss:0.101,Test_test:93.4%,Test_loss:0.226,Lr:1.00E-04
Epoch:41,Train_acc:97.0%,Train_loss:0.111,Test_test:90.2%,Test_loss:0.322,Lr:1.00E-04
Epoch:42,Train_acc:97.5%,Train_loss:0.098,Test_test:91.7%,Test_loss:0.274,Lr:1.00E-04
Epoch:43,Train_acc:97.9%,Train_loss:0.089,Test_test:93.0%,Test_loss:0.238,Lr:1.00E-04
Epoch:44,Train_acc:97.6%,Train_loss:0.095,Test_test:89.5%,Test_loss:0.351,Lr:1.00E-04
Epoch:45,Train_acc:98.1%,Train_loss:0.083,Test_test:92.4%,Test_loss:0.262,Lr:1.00E-04
Epoch:46,Train_acc:97.8%,Train_loss:0.088,Test_test:90.8%,Test_loss:0.301,Lr:1.00E-04
Epoch:47,Train_acc:98.2%,Train_loss:0.078,Test_test:93.5%,Test_loss:0.229,Lr:1.00E-04
Epoch:48,Train_acc:98.0%,Train_loss:0.082,Test_test:91.9%,Test_loss:0.268,Lr:1.00E-04
Epoch:49,Train_acc:98.4%,Train_loss:0.071,Test_test:92.7%,Test_loss:0.249,Lr:1.00E-04
Epoch:50,Train_acc:98.1%,Train_loss:0.078,Test_test:89.3%,Test_loss:0.443,Lr:1.00E-04
Epoch:51,Train_acc:98.5%,Train_loss:0.066,Test_test:91.5%,Test_loss:0.286,Lr:1.00E-04
Epoch:52,Train_acc:98.7%,Train_loss:0.061,Test_test:94.0%,Test_loss:0.221,Lr:1.00E-04
Epoch:53,Train_acc:98.4%,Train_loss:0.067,Test_test:92.1%,Test_loss:0.264,Lr:1.00E-04
Epoch:54,Train_acc:98.9%,Train_loss:0.054,Test_test:93.2%,Test_loss:0.247,Lr:1.00E-04
Epoch:55,Train_acc:99.0%,Train_loss:0.049,Test_test:91.6%,Test_loss:0.295,Lr:1.00E-04
Epoch:56,Train_acc:98.7%,Train_loss:0.056,Test_test:94.3%,Test_loss:0.214,Lr:1.00E-04
Epoch:57,Train_acc:99.1%,Train_loss:0.043,Test_test:92.7%,Test_loss:0.257,Lr:1.00E-04
Epoch:58,Train_acc:98.9%,Train_loss:0.048,Test_test:95.0%,Test_loss:0.203,Lr:1.00E-04
Epoch:59,Train_acc:99.3%,Train_loss:0.038,Test_test:93.8%,Test_loss:0.229,Lr:1.00E-04
Epoch:60,Train_acc:99.1%,Train_loss:0.042,Test_test:95.6%,Test_loss:0.187,Lr:1.00E-04
Done准确率和损失率图：
3.4 模型评估
# 将参数加载到model当中
best_model.load_state_dict(torch.load(PATH, map_location=device))
epoch_test_acc, epoch_test_loss = test(test_dl, best_model, loss_fn)
epoch_test_acc, epoch_test_loss结果为：
(0.9556739971012556, 0.18679569537080343) 可以看出测试准确率为95.6%，测试损失为0.187。
4 个人总结
本周基于天气识别的四分类任务，使用YOLOv5 Backbone在图像分类中的应用。整体流程包括：数据预处理与8:2划分、进行Backbone模块的复现、分类头的构建、训练与测试函数封装、60 轮训练。
对该模块的建立方式，可以有效提升了模型预测的精度，后续应注意学习一下。
更多推荐
7
0- 0
已为社区贡献4条内容
所有评论(0)