Python+pytorch实现天气识别

一、前期工作

环境：python3.6，1080ti，pytorch1.10（实验室服务器的环境）

1.设置GPU或者cpu

import torch
import torch.nn as nn
import matplotlib.pyplot as plt
import torchvision

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

device

2.导入数据

import os,PIL,random,pathlib

data_dir = 'weather_photos/'
data_dir = pathlib.Path(data_dir)
print(data_dir)

data_paths = list(data_dir.glob('*'))
print(data_paths)
classeNames = [str(path).split("/")[1] for path in data_paths]
classeNames

二、数据预处理

数据格式设置

total_datadir = 'weather_photos/'

# 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.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(total_datadir,transform=train_transforms)
total_data

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

设置dataset

batch_size = 32

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)

检查数据格式 

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

三、搭建网络

import torch
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential,ReLU

num_classes = 4

class Model(nn.Module):
   def __init__(self):
       super(Model,self).__init__()
       # 卷积层
       self.layers = Sequential(
           # 第一层
           nn.Conv2d(3, 24, kernel_size=5),
           nn.BatchNorm2d(24),
           nn.ReLU(),
           # 第二层
           nn.Conv2d(24,64 , kernel_size=5),
           nn.BatchNorm2d(64),
           nn.ReLU(),
           nn.MaxPool2d(2,2),
           nn.Conv2d(64, 128, kernel_size=5),
           nn.BatchNorm2d(128),
           nn.ReLU(),
           nn.Conv2d(128, 24, kernel_size=5),
           nn.BatchNorm2d(24),
           nn.ReLU(),
           nn.MaxPool2d(2,2),
           nn.Flatten(),
           nn.Linear(24*50*50, 516,bias=True),
           nn.ReLU(),
           nn.Dropout(0.5),
           nn.Linear(516, 215,bias=True),
           nn.ReLU(),
           nn.Dropout(0.5),
           nn.Linear(215, num_classes,bias=True),
       )

def forward(self, x):

x = self.layers(x)
       return x    

device = "cuda" if torch.cuda.is_available() else "cpu"
print("Using {} device".format(device))

model = Model().to(device)
model

打印网络结构

四、训练模型

1.设置学习率

loss_fn    = nn.CrossEntropyLoss() # 创建损失函数
learn_rate = 1e-3 # 学习率
opt        = torch.optim.SGD(model.parameters(),lr=learn_rate)

2.模型训练

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，一共60000张图片
   num_batches = len(dataloader)   # 批次数目，1875（60000/32）

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，一共10000张图片
   num_batches = len(dataloader)          # 批次数目，313（10000/32=312.5，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

具体训练代码 

epochs     = 30
train_loss = []
train_acc  = []
test_loss  = []
test_acc   = []

for epoch in range(epochs):
   model.train()
   epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, opt)

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

train_acc.append(epoch_train_acc)
   train_loss.append(epoch_train_loss)
   test_acc.append(epoch_test_acc)
   test_loss.append(epoch_test_loss)

template = ('Epoch:{:2d}, Train_acc:{:.1f}％, Train_loss:{:.3f}, Test_acc:{:.1f}％，Test_loss:{:.3f}')
   print(template.format(epoch+1, epoch_train_acc*100, epoch_train_loss, epoch_test_acc*100, epoch_test_loss))
print('Done')

五、模型评估

1.Loss和Accuracy图

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()

2.对结果进行预测

import os
import json

import torch
from PIL import Image
from torchvision import transforms
import matplotlib.pyplot as plt

img_path = "weather_photos/cloudy/cloudy1.jpg"
classes = ['cloudy', 'rain', 'shine', 'sunrise']
data_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。
])
def main():
   device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

img = Image.open(img_path)
   plt.imshow(img)
   # [N, C, H, W]
   img = data_transform(img)
   # expand batch dimension
   img = torch.unsqueeze(img, dim=0)
   model.eval()
   with torch.no_grad():
       # predict class
       output = torch.squeeze(model(img.to(device))).cpu()
       predict = torch.softmax(output, dim=0)
       predict_cla = torch.argmax(predict).numpy()
       print(classes[predict_cla])
   plt.show()

if __name__ == '__main__':
   main()

预测结果如下：

3.总结

1.本次能主要对以下函数进行了学习

来源：https://blog.csdn.net/m0_60524373/article/details/127195845