This article is a learning record blog in 365-day deep learning training camp
Reference article address: 365 Days Deep Learning Training Camp – Week P3: Weather Recognition
Author: Classmate K
###This project comes from the online guidance of K students###
Data set download: https://pan.baidu.com/s/1Viq7s2FEtmcQQ3sRTMhszg
Extraction code: hqij
import torch import torch.nn as nn import torchvision.transforms as transforms from torchvision import datasets import pathlib ##Check if there is a cuda graphics card device=torch.device("cuda" if torch.cuda.is_available() else "cpu") print(device) data_dir='./weather_photos/' data_dir=pathlib.Path(data_dir) data_paths=list(data_dir.glob('*')) classNames=[str(path).split('\')[1] for path in data_paths] print(classNames) train_transforms = transforms. Compose([ transforms.Resize([224, 224]), # resize input image transforms.ToTensor(), # Convert PIL Image or numpy.ndarray to tensor transforms. Normalize( mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]) # Calculated by random sampling from the data set ]) total_data = datasets. ImageFolder(data_dir, transform=train_transforms) 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 = 32 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) break import torch.nn.functional as F num_classes = 4 # Number of categories of pictures class Network_bn(nn.Module): def __init__(self): super().__init__() # feature extraction network self.conv1 = nn.Conv2d(in_channels=3, out_channels=12, kernel_size=5, stride=1, padding=0) self.bn1 = nn.BatchNorm2d(12) self.conv2 = nn.Conv2d(in_channels=12, out_channels=12, kernel_size=5, stride=1, padding=0) self.bn2 = nn.BatchNorm2d(12) self.pool = nn.MaxPool2d(2, 2) self.conv3 = nn.Conv2d(in_channels=12, out_channels=24, kernel_size=5, stride=1, padding=0) self.bn3 = nn.BatchNorm2d(24) self.conv4 = nn.Conv2d(in_channels=24, out_channels=24, kernel_size=5, stride=1, padding=0) self.bn4 = nn.BatchNorm2d(24) # classification network self.fc1 = nn.Linear(24 * 50 * 50, num_classes) # forward pass def forward(self, x): x = F.relu(self.bn1(self.conv1(x))) x = F.relu(self.bn2(self.conv2(x))) x = self. pool(x) x = F.relu(self.bn3(self.conv3(x))) x = F.relu(self.bn4(self.conv4(x))) x = self. pool(x) x = x. view(-1, 24 * 50 * 50) x = self.fc1(x) return x model = Network_bn().to(device) loss_fn = nn.CrossEntropyLoss() # create loss function learn_rate = 1e-4 # learning rate opt = torch.optim.SGD(model.parameters(), lr=learn_rate) # training loop def train(dataloader, model, loss_fn, optimizer): size = len(dataloader.dataset) # The size of the training set, a total of 900 pictures num_batches = len(dataloader) # number of batches, 29 (900/32) train_loss, train_acc = 0, 0 # Initialize training loss and accuracy for X, y in dataloader: # Get pictures and their labels X, y = X.to(device), y.to(device) # Compute prediction error pred = model(X) # network output loss = loss_fn(pred, y) # Calculate the gap between the network output and the real value, targets is the real value, and the difference between the two is calculated as the loss # Backpropagation optimizer.zero_grad() # grad attribute is zeroed loss.backward() # backpropagation optimizer.step() # Automatically update each step # Record acc and 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) # The size of the test set, a total of 10,000 pictures num_batches = len(dataloader) # Number of batches, 8 (255/32=8, rounded up) test_loss, test_acc = 0, 0 # When training is not in progress, stop gradient update to save computing memory consumption with torch.no_grad(): for imgs, target in dataloader: imgs, target = imgs.to(device), target.to(device) # calculate 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 epochs = 100 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') import matplotlib.pyplot as plt # hide warnings import warnings warnings.filterwarnings("ignore") # ignore warnings plt.rcParams['font.sans-serif'] = ['SimHei'] # used to display Chinese labels normally plt.rcParams['axes.unicode_minus'] = False # used to display the negative sign normally plt.rcParams['figure.dpi'] = 100 # resolution 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()
Training process:
Epoch: 1, Train_acc: 65.3%, Train_loss: 1.396, Test_acc: 52.4%, Test_loss: 0.886
Epoch: 2, Train_acc: 85.1%, Train_loss: 0.381, Test_acc: 86.7%, Test_loss: 0.345
Epoch: 3, Train_acc: 91.7%, Train_loss: 0.252, Test_acc: 89.8%, Test_loss: 0.257
Epoch: 4, Train_acc: 92.1%, Train_loss: 0.250, Test_acc: 40.4%, Test_loss: 6.429
Epoch: 5, Train_acc: 91.1%, Train_loss: 0.342, Test_acc: 67.1%, Test_loss: 0.966
Epoch: 6, Train_acc: 93.9%, Train_loss: 0.150, Test_acc: 89.8%, Test_loss: 0.223
Epoch: 7, Train_acc: 95.8%, Train_loss: 0.136, Test_acc: 83.6%, Test_loss: 0.820
Epoch: 8, Train_acc: 95.8%, Train_loss: 0.149, Test_acc: 86.7%, Test_loss: 0.291
Epoch: 9, Train_acc: 94.2%, Train_loss: 0.143, Test_acc: 89.8%, Test_loss: 0.264
Epoch: 10, Train_acc: 98.1%, Train_loss: 0.083, Test_acc: 72.9%, Test_loss: 1.241
Epoch: 11, Train_acc: 96.9%, Train_loss: 0.145, Test_acc: 53.3%, Test_loss: 7.011
Epoch: 12, Train_acc: 88.3%, Train_loss: 0.469, Test_acc: 56.0%, Test_loss: 2.457
Epoch: 13, Train_acc: 94.0%, Train_loss: 0.204, Test_acc: 92.0%, Test_loss: 0.211
Epoch: 14, Train_acc: 98.0%, Train_loss: 0.096, Test_acc: 75.1%, Test_loss: 0.686
Epoch: 15, Train_acc: 95.3%, Train_loss: 0.121, Test_acc: 91.6%, Test_loss: 0.244
Epoch: 16, Train_acc: 97.3%, Train_loss: 0.083, Test_acc: 84.9%, Test_loss: 0.305
Epoch: 17, Train_acc: 98.8%, Train_loss: 0.047, Test_acc: 92.0%, Test_loss: 0.184
Epoch: 18, Train_acc: 99.1%, Train_loss: 0.034, Test_acc: 93.3%, Test_loss: 0.210
Epoch: 19, Train_acc: 99.8%, Train_loss: 0.027, Test_acc: 94.2%, Test_loss: 0.174
Epoch: 20, Train_acc: 99.7%, Train_loss: 0.033, Test_acc: 92.4%, Test_loss: 0.182
Done
experience:
(1) There is a certain degree of overfitting
(2) The training process has small fluctuations and is unstable
(3) The training method is relatively simple
Solution:
(1) Add L2 regularization or dropout
(2) Increase the data set and try to increase the number of channels in the second convolutional layer, resulting in a sharp drop in accuracy, and a slight increase in the third layer