- 🍨 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
🚀我的环境：
- 语言环境：python 3.12.6
- 编译器：jupyter lab
- 深度学习环境：Pytorch
前期准备
import torch
import torch.nn as nn
import torchvision.transforms as transforms
import torchvision
from torchvision import transforms, datasets
import os,PIL,pathlib,warningswarnings.filterwarnings("ignore") #忽略警告信息device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
device
device(type='cpu')
import os,PIL,random,pathlibdata_dir = 'd:/Users/yxy/Desktop/weather_photos'
data_dir = pathlib.Path(data_dir)data_paths = list(data_dir.glob('*'))
classeNames = [str(path).split("\\")[5] for path in data_paths]
classeNames
['cloudy', 'rain', 'shine', 'sunrise']
# 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:/Users/yxy/Desktop/weather_photos",transform=train_transforms)
total_data
Dataset ImageFolderNumber of datapoints: 1125Root location: d:/Users/yxy/Desktop/weather_photosStandardTransform
Transform: Compose(Resize(size=[224, 224], interpolation=bilinear, max_size=None, antialias=True)ToTensor()Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]))
total_data.class_to_idx
{'cloudy': 0, 'rain': 1, 'shine': 2, 'sunrise': 3}
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
(<torch.utils.data.dataset.Subset at 0x22b647f0230>,<torch.utils.data.dataset.Subset at 0x22b00648800>)
batch_size = 4train_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
Shape of X [N, C, H, W]: torch.Size([4, 3, 224, 224])
Shape of y: torch.Size([4]) torch.int64
搭建模型
import torch.nn.functional as Fdef 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-padreturn pclass Conv(nn.Module):# Standard convolutiondef __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True): # ch_in, ch_out, kernel, stride, padding, groupssuper().__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 bottleneckdef __init__(self, c1, c2, shortcut=True, g=1, e=0.5): # ch_in, ch_out, shortcut, groups, expansionsuper().__init__()c_ = int(c2 * e) # hidden channelsself.cv1 = Conv(c1, c_, 1, 1)self.cv2 = Conv(c_, c2, 3, 1, g=g)self.add = shortcut and c1 == c2def 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 convolutionsdef __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5): # ch_in, ch_out, number, shortcut, groups, expansionsuper().__init__()c_ = int(c2 * e) # hidden channelsself.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 Jocherdef __init__(self, c1, c2, k=5): # equivalent to SPP(k=(5, 9, 13))super().__init__()c_ = c1 // 2 # hidden channelsself.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() warningy1 = 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 xdevice = "cuda" if torch.cuda.is_available() else "cpu"
print("Using {} device".format(device))model = YOLOv5_backbone().to(device)
model
Using cpu deviceYOLOv5_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())(C3_3): C3((cv1): Conv((conv): Conv2d(128, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU())(cv2): Conv((conv): Conv2d(128, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU())(cv3): Conv((conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU())(m): Sequential((0): Bottleneck((cv1): Conv((conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU())(cv2): Conv((conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU()))))(Conv_4): Conv((conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU())(C3_5): C3((cv1): Conv((conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU())(cv2): Conv((conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU())(cv3): Conv((conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU())(m): Sequential((0): Bottleneck((cv1): Conv((conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU())(cv2): Conv((conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU()))))(Conv_6): Conv((conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU())(C3_7): C3((cv1): Conv((conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU())(cv2): Conv((conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU())(cv3): Conv((conv): Conv2d(512, 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())(m): Sequential((0): Bottleneck((cv1): Conv((conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU())(cv2): Conv((conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU()))))(Conv_8): Conv((conv): Conv2d(512, 1024, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)(bn): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(act): SiLU())(C3_9): C3((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(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())(cv3): Conv((conv): Conv2d(1024, 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): Sequential((0): Bottleneck((cv1): Conv((conv): Conv2d(512, 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(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn): BatchNorm2d(512, 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))
)
# 统计模型参数量以及其他指标
import torchsummary as summary
summary.summary(model, (3, 224, 224))
----------------------------------------------------------------Layer (type) Output Shape Param #
================================================================Conv2d-1 [-1, 64, 113, 113] 1,728BatchNorm2d-2 [-1, 64, 113, 113] 128SiLU-3 [-1, 64, 113, 113] 0Conv-4 [-1, 64, 113, 113] 0Conv2d-5 [-1, 128, 57, 57] 73,728BatchNorm2d-6 [-1, 128, 57, 57] 256SiLU-7 [-1, 128, 57, 57] 0Conv-8 [-1, 128, 57, 57] 0Conv2d-9 [-1, 64, 57, 57] 8,192BatchNorm2d-10 [-1, 64, 57, 57] 128SiLU-11 [-1, 64, 57, 57] 0Conv-12 [-1, 64, 57, 57] 0Conv2d-13 [-1, 64, 57, 57] 4,096BatchNorm2d-14 [-1, 64, 57, 57] 128SiLU-15 [-1, 64, 57, 57] 0Conv-16 [-1, 64, 57, 57] 0Conv2d-17 [-1, 64, 57, 57] 36,864BatchNorm2d-18 [-1, 64, 57, 57] 128SiLU-19 [-1, 64, 57, 57] 0Conv-20 [-1, 64, 57, 57] 0Bottleneck-21 [-1, 64, 57, 57] 0Conv2d-22 [-1, 64, 57, 57] 8,192BatchNorm2d-23 [-1, 64, 57, 57] 128SiLU-24 [-1, 64, 57, 57] 0Conv-25 [-1, 64, 57, 57] 0Conv2d-26 [-1, 128, 57, 57] 16,384BatchNorm2d-27 [-1, 128, 57, 57] 256SiLU-28 [-1, 128, 57, 57] 0Conv-29 [-1, 128, 57, 57] 0C3-30 [-1, 128, 57, 57] 0Conv2d-31 [-1, 256, 29, 29] 294,912BatchNorm2d-32 [-1, 256, 29, 29] 512SiLU-33 [-1, 256, 29, 29] 0Conv-34 [-1, 256, 29, 29] 0Conv2d-35 [-1, 128, 29, 29] 32,768BatchNorm2d-36 [-1, 128, 29, 29] 256SiLU-37 [-1, 128, 29, 29] 0Conv-38 [-1, 128, 29, 29] 0Conv2d-39 [-1, 128, 29, 29] 16,384BatchNorm2d-40 [-1, 128, 29, 29] 256SiLU-41 [-1, 128, 29, 29] 0Conv-42 [-1, 128, 29, 29] 0Conv2d-43 [-1, 128, 29, 29] 147,456BatchNorm2d-44 [-1, 128, 29, 29] 256SiLU-45 [-1, 128, 29, 29] 0Conv-46 [-1, 128, 29, 29] 0Bottleneck-47 [-1, 128, 29, 29] 0Conv2d-48 [-1, 128, 29, 29] 32,768BatchNorm2d-49 [-1, 128, 29, 29] 256SiLU-50 [-1, 128, 29, 29] 0Conv-51 [-1, 128, 29, 29] 0Conv2d-52 [-1, 256, 29, 29] 65,536BatchNorm2d-53 [-1, 256, 29, 29] 512SiLU-54 [-1, 256, 29, 29] 0Conv-55 [-1, 256, 29, 29] 0C3-56 [-1, 256, 29, 29] 0Conv2d-57 [-1, 512, 15, 15] 1,179,648BatchNorm2d-58 [-1, 512, 15, 15] 1,024SiLU-59 [-1, 512, 15, 15] 0Conv-60 [-1, 512, 15, 15] 0Conv2d-61 [-1, 256, 15, 15] 131,072BatchNorm2d-62 [-1, 256, 15, 15] 512SiLU-63 [-1, 256, 15, 15] 0Conv-64 [-1, 256, 15, 15] 0Conv2d-65 [-1, 256, 15, 15] 65,536BatchNorm2d-66 [-1, 256, 15, 15] 512SiLU-67 [-1, 256, 15, 15] 0Conv-68 [-1, 256, 15, 15] 0Conv2d-69 [-1, 256, 15, 15] 589,824BatchNorm2d-70 [-1, 256, 15, 15] 512SiLU-71 [-1, 256, 15, 15] 0Conv-72 [-1, 256, 15, 15] 0Bottleneck-73 [-1, 256, 15, 15] 0Conv2d-74 [-1, 256, 15, 15] 131,072BatchNorm2d-75 [-1, 256, 15, 15] 512SiLU-76 [-1, 256, 15, 15] 0Conv-77 [-1, 256, 15, 15] 0Conv2d-78 [-1, 512, 15, 15] 262,144BatchNorm2d-79 [-1, 512, 15, 15] 1,024SiLU-80 [-1, 512, 15, 15] 0Conv-81 [-1, 512, 15, 15] 0C3-82 [-1, 512, 15, 15] 0Conv2d-83 [-1, 1024, 8, 8] 4,718,592BatchNorm2d-84 [-1, 1024, 8, 8] 2,048SiLU-85 [-1, 1024, 8, 8] 0Conv-86 [-1, 1024, 8, 8] 0Conv2d-87 [-1, 512, 8, 8] 524,288BatchNorm2d-88 [-1, 512, 8, 8] 1,024SiLU-89 [-1, 512, 8, 8] 0Conv-90 [-1, 512, 8, 8] 0Conv2d-91 [-1, 512, 8, 8] 262,144BatchNorm2d-92 [-1, 512, 8, 8] 1,024SiLU-93 [-1, 512, 8, 8] 0Conv-94 [-1, 512, 8, 8] 0Conv2d-95 [-1, 512, 8, 8] 2,359,296BatchNorm2d-96 [-1, 512, 8, 8] 1,024SiLU-97 [-1, 512, 8, 8] 0Conv-98 [-1, 512, 8, 8] 0Bottleneck-99 [-1, 512, 8, 8] 0Conv2d-100 [-1, 512, 8, 8] 524,288BatchNorm2d-101 [-1, 512, 8, 8] 1,024SiLU-102 [-1, 512, 8, 8] 0Conv-103 [-1, 512, 8, 8] 0Conv2d-104 [-1, 1024, 8, 8] 1,048,576BatchNorm2d-105 [-1, 1024, 8, 8] 2,048SiLU-106 [-1, 1024, 8, 8] 0Conv-107 [-1, 1024, 8, 8] 0C3-108 [-1, 1024, 8, 8] 0Conv2d-109 [-1, 512, 8, 8] 524,288BatchNorm2d-110 [-1, 512, 8, 8] 1,024SiLU-111 [-1, 512, 8, 8] 0Conv-112 [-1, 512, 8, 8] 0MaxPool2d-113 [-1, 512, 8, 8] 0MaxPool2d-114 [-1, 512, 8, 8] 0MaxPool2d-115 [-1, 512, 8, 8] 0Conv2d-116 [-1, 1024, 8, 8] 2,097,152BatchNorm2d-117 [-1, 1024, 8, 8] 2,048SiLU-118 [-1, 1024, 8, 8] 0Conv-119 [-1, 1024, 8, 8] 0SPPF-120 [-1, 1024, 8, 8] 0Linear-121 [-1, 100] 6,553,700ReLU-122 [-1, 100] 0Linear-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
----------------------------------------------------------------
训练模型
# 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losstrain_acc += (pred.argmax(1) == y).type(torch.float).sum().item()train_loss += loss.item()train_acc /= sizetrain_loss /= num_batchesreturn 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, (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losstarget_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 /= sizetest_loss /= num_batchesreturn test_acc, test_loss
import copyoptimizer = torch.optim.Adam(model.parameters(), lr= 1e-4)
loss_fn = nn.CrossEntropyLoss() # 创建损失函数epochs = 60train_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_modelif epoch_test_acc > best_acc:best_acc = epoch_test_accbest_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:/Users/yxy/Desktop/best_model.pth' # 保存的参数文件名
torch.save(best_model.state_dict(), PATH)print('Done')
Epoch: 1, Train_acc:70.4%, Train_loss:0.795, Test_acc:64.9%, Test_loss:0.909, Lr:1.00E-04
Epoch: 2, Train_acc:74.2%, Train_loss:0.688, Test_acc:81.3%, Test_loss:0.448, Lr:1.00E-04
Epoch: 3, Train_acc:79.3%, Train_loss:0.565, Test_acc:78.7%, Test_loss:0.541, Lr:1.00E-04
Epoch: 4, Train_acc:81.4%, Train_loss:0.553, Test_acc:72.0%, Test_loss:0.741, Lr:1.00E-04
Epoch: 5, Train_acc:84.8%, Train_loss:0.430, Test_acc:86.7%, Test_loss:0.328, Lr:1.00E-04
Epoch: 6, Train_acc:85.9%, Train_loss:0.384, Test_acc:79.1%, Test_loss:0.560, Lr:1.00E-04
Epoch: 7, Train_acc:88.8%, Train_loss:0.330, Test_acc:85.3%, Test_loss:0.353, Lr:1.00E-04
Epoch: 8, Train_acc:89.8%, Train_loss:0.332, Test_acc:87.6%, Test_loss:0.345, Lr:1.00E-04
Epoch: 9, Train_acc:89.2%, Train_loss:0.297, Test_acc:87.1%, Test_loss:0.299, Lr:1.00E-04
Epoch:10, Train_acc:91.1%, Train_loss:0.264, Test_acc:88.0%, Test_loss:0.294, Lr:1.00E-04
Epoch:11, Train_acc:93.1%, Train_loss:0.205, Test_acc:91.1%, Test_loss:0.280, Lr:1.00E-04
Epoch:12, Train_acc:94.2%, Train_loss:0.153, Test_acc:90.2%, Test_loss:0.323, Lr:1.00E-04
Epoch:13, Train_acc:91.9%, Train_loss:0.198, Test_acc:85.8%, Test_loss:0.468, Lr:1.00E-04
Epoch:14, Train_acc:92.4%, Train_loss:0.197, Test_acc:88.0%, Test_loss:0.439, Lr:1.00E-04
Epoch:15, Train_acc:94.9%, Train_loss:0.149, Test_acc:90.7%, Test_loss:0.372, Lr:1.00E-04
Epoch:16, Train_acc:93.9%, Train_loss:0.161, Test_acc:87.1%, Test_loss:0.381, Lr:1.00E-04
Epoch:17, Train_acc:95.0%, Train_loss:0.140, Test_acc:92.0%, Test_loss:0.313, Lr:1.00E-04
Epoch:18, Train_acc:96.3%, Train_loss:0.113, Test_acc:94.2%, Test_loss:0.251, Lr:1.00E-04
Epoch:19, Train_acc:97.4%, Train_loss:0.059, Test_acc:90.2%, Test_loss:0.352, Lr:1.00E-04
Epoch:20, Train_acc:97.4%, Train_loss:0.075, Test_acc:91.1%, Test_loss:0.348, Lr:1.00E-04
Epoch:21, Train_acc:97.7%, Train_loss:0.079, Test_acc:93.8%, Test_loss:0.263, Lr:1.00E-04
Epoch:22, Train_acc:95.4%, Train_loss:0.121, Test_acc:83.1%, Test_loss:0.763, Lr:1.00E-04
Epoch:23, Train_acc:96.8%, Train_loss:0.095, Test_acc:92.0%, Test_loss:0.281, Lr:1.00E-04
Epoch:24, Train_acc:96.9%, Train_loss:0.079, Test_acc:88.4%, Test_loss:0.460, Lr:1.00E-04
Epoch:25, Train_acc:97.2%, Train_loss:0.084, Test_acc:84.9%, Test_loss:0.531, Lr:1.00E-04
Epoch:26, Train_acc:98.8%, Train_loss:0.045, Test_acc:90.2%, Test_loss:0.346, Lr:1.00E-04
Epoch:27, Train_acc:97.9%, Train_loss:0.048, Test_acc:90.2%, Test_loss:0.431, Lr:1.00E-04
Epoch:28, Train_acc:97.9%, Train_loss:0.066, Test_acc:88.9%, Test_loss:0.380, Lr:1.00E-04
Epoch:29, Train_acc:98.2%, Train_loss:0.046, Test_acc:85.8%, Test_loss:0.573, Lr:1.00E-04
Epoch:30, Train_acc:96.8%, Train_loss:0.083, Test_acc:88.0%, Test_loss:0.585, Lr:1.00E-04
Epoch:31, Train_acc:96.2%, Train_loss:0.116, Test_acc:92.0%, Test_loss:0.331, Lr:1.00E-04
Epoch:32, Train_acc:98.2%, Train_loss:0.045, Test_acc:84.0%, Test_loss:0.680, Lr:1.00E-04
Epoch:33, Train_acc:96.7%, Train_loss:0.083, Test_acc:93.3%, Test_loss:0.372, Lr:1.00E-04
Epoch:34, Train_acc:97.6%, Train_loss:0.070, Test_acc:88.9%, Test_loss:0.450, Lr:1.00E-04
Epoch:35, Train_acc:99.0%, Train_loss:0.026, Test_acc:88.0%, Test_loss:0.546, Lr:1.00E-04
Epoch:36, Train_acc:99.0%, Train_loss:0.033, Test_acc:92.4%, Test_loss:0.358, Lr:1.00E-04
Epoch:37, Train_acc:99.9%, Train_loss:0.005, Test_acc:92.0%, Test_loss:0.384, Lr:1.00E-04
Epoch:38, Train_acc:99.8%, Train_loss:0.010, Test_acc:90.7%, Test_loss:0.470, Lr:1.00E-04
Epoch:39, Train_acc:98.3%, Train_loss:0.051, Test_acc:92.4%, Test_loss:0.429, Lr:1.00E-04
Epoch:40, Train_acc:98.2%, Train_loss:0.051, Test_acc:87.6%, Test_loss:0.473, Lr:1.00E-04
Epoch:41, Train_acc:98.9%, Train_loss:0.023, Test_acc:85.3%, Test_loss:0.683, Lr:1.00E-04
Epoch:42, Train_acc:98.1%, Train_loss:0.059, Test_acc:89.8%, Test_loss:0.425, Lr:1.00E-04
Epoch:43, Train_acc:98.6%, Train_loss:0.035, Test_acc:92.4%, Test_loss:0.361, Lr:1.00E-04
Epoch:44, Train_acc:99.6%, Train_loss:0.009, Test_acc:91.1%, Test_loss:0.406, Lr:1.00E-04
Epoch:45, Train_acc:99.7%, Train_loss:0.016, Test_acc:89.8%, Test_loss:0.479, Lr:1.00E-04
Epoch:46, Train_acc:98.7%, Train_loss:0.058, Test_acc:89.3%, Test_loss:0.416, Lr:1.00E-04
Epoch:47, Train_acc:97.8%, Train_loss:0.067, Test_acc:88.4%, Test_loss:0.398, Lr:1.00E-04
Epoch:48, Train_acc:98.8%, Train_loss:0.043, Test_acc:86.7%, Test_loss:0.455, Lr:1.00E-04
Epoch:49, Train_acc:99.8%, Train_loss:0.007, Test_acc:92.0%, Test_loss:0.291, Lr:1.00E-04
Epoch:50, Train_acc:98.1%, Train_loss:0.057, Test_acc:91.6%, Test_loss:0.390, Lr:1.00E-04
Epoch:51, Train_acc:98.8%, Train_loss:0.050, Test_acc:89.3%, Test_loss:0.462, Lr:1.00E-04
Epoch:52, Train_acc:98.9%, Train_loss:0.035, Test_acc:88.9%, Test_loss:0.507, Lr:1.00E-04
Epoch:53, Train_acc:96.8%, Train_loss:0.082, Test_acc:82.2%, Test_loss:1.082, Lr:1.00E-04
Epoch:54, Train_acc:99.2%, Train_loss:0.018, Test_acc:91.6%, Test_loss:0.350, Lr:1.00E-04
Epoch:55, Train_acc:99.7%, Train_loss:0.010, Test_acc:89.3%, Test_loss:0.443, Lr:1.00E-04
Epoch:56, Train_acc:99.8%, Train_loss:0.007, Test_acc:88.4%, Test_loss:0.661, Lr:1.00E-04
Epoch:57, Train_acc:99.7%, Train_loss:0.009, Test_acc:91.6%, Test_loss:0.439, Lr:1.00E-04
Epoch:58, Train_acc:99.1%, Train_loss:0.035, Test_acc:86.7%, Test_loss:0.643, Lr:1.00E-04
Epoch:59, Train_acc:98.6%, Train_loss:0.046, Test_acc:89.8%, Test_loss:0.593, Lr:1.00E-04
Epoch:60, Train_acc:99.9%, Train_loss:0.006, Test_acc:88.4%, Test_loss:0.530, Lr:1.00E-04
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()
# 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
(0.9422222222222222, 0.25122066404212984)
总结
用YOLOv5算法中的Backbone模块搭建网络，在 YOLOv5 中，Backbone 模块主要负责提取图像的基础特征信息，为后续检测任务打好基础。YOLOv5 通常采用 CSPDarknet53 作为主干网络，该结构结合了跨阶段部分连接（CSP）与残差模块（ResBlock），有效增强了网络的表达能力与计算效率。Backbone 的主要构成包括：
1.Focus层：对输入图像进行切片和通道融合，提高感受野；
2.Conv + BatchNorm + LeakyReLU 组合：实现高效特征提取；
3.CSP模块：优化梯度流和参数量；
4.残差连接：防止梯度消失，增强深层学习能力。
这些设计使得 YOLOv5 的 Backbone 既高效又轻量，非常适合实时目标检测任务。
