整体架构
分为3部分
- Transforms
- 用来做数据预处理，例如图片的缩放、剪切、正则化。
torchvision.transforms有现成的实现torchvision.transforms.Compose可以把多个串起来，当成一个来用- 可以自定义，继承自
object
- Dataset：
- 用来设定如何读取数据源，整合 Transforms
- 可以自定义，继承自
torch.utils.data.Dataset - 也可以用
dataset = torch.utils.data.Dataset(x_train_tensor, y_train_tensor) torchvision.datasets有现成的实现。例如torchvision.datasets.ImageFolder()以及经典数据集
torch.utils.data.DataLoader(dataset=dataset, batch_size=batch_size, shuffle=True, num_workers=num_workers)- 是一个iterable对象
简单示例
import torch
from torchvision import transforms, datasets
my_transform = transforms.Compose([
transforms.RandomSizedCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
])
my_dataset = datasets.ImageFolder(root='hymenoptera_data/train',
transform=my_transform)
my_dataloader = torch.utils.data.DataLoader(
my_dataset, batch_size=4, shuffle=True,
num_workers=4)
# DataLoader也预提供了不同的用法：
torch.utils.data.DataLoader
torch.utils.data.RandomSampler
torch.utils.data.SequentialSampler
自定义
1. Transformer
写法1
# 例如，下面这个对图片做一些操作
import torch
from torch.utils.data import DataLoader
from torchvision import transforms, datasets
class MyTransformer(object):
def __init__(self, output_size):
assert isinstance(output_size, (int, tuple))
self.output_size = output_size
def __call__(self, sample):
image, landmarks = sample['image'], sample['landmarks']
pass # 可以有很复杂的数据预处理逻辑，省略不写了
return {'image': image, 'landmarks': landmarks}
my_transformer = MyTransformer(255)
写法2:
# 例如，skimage 中有很多现成的，并且可以用 Compose 连接起来，当成一个用
from skimage import io, transform
my_transformer = transforms.Compose([
MyTransformer(256),
transforms.RandomCrop(224),
transforms.RandomResizedCrop(224),
transforms.ToTensor()
])
2. Dataset
写法1:
class MyDataset(torch.utils.data.Dataset):
"""Face Landmarks dataset."""
def __init__(self, csv_file, root_dir, transform=None):
"""
Args:
csv_file (string): Path to the csv file with annotations.
root_dir (string): Directory with all the images.
transform (callable, optional): Optional transform to be applied
on a sample.
"""
self.landmarks_frame = pd.read_csv(csv_file)
self.root_dir = root_dir
self.transform = transform
def __len__(self):
return len(self.landmarks_frame)
def __getitem__(self, idx):
if torch.is_tensor(idx):
idx = idx.tolist()
img_name = os.path.join(self.root_dir,
self.landmarks_frame.iloc[idx, 0])
image = io.imread(img_name)
landmarks = self.landmarks_frame.iloc[idx, 1:]
landmarks = np.array([landmarks])
landmarks = landmarks.astype('float').reshape(-1, 2)
sample = {'image': image, 'landmarks': landmarks}
if self.transform:
sample = self.transform(sample)
return sample
# 另一个源码返回的是这个：
# img_data,target
my_dataset = MyDataset(
csv_file='data/faces/face_landmarks.csv',
root_dir='data/faces/',
transform=my_transformer)
简洁写法（通用）
my_dataset = TensorDataset(torch.tensor(data), torch.tensor(label))
简洁写法（图像类）
from torchvision import datasets
my_dataset = datasets.ImageFolder('./train', transform=my_transformer)
# 目录格式约定如下：
# 图片文件名随意，目录名字就是类名
'''
.
├── train
│ ├── ants
│ │ └── ant2.jpg
│ └── bees
│ ├── a.jpg
│ └── b.jpg
└── val
├── ants
│ ├── ant8.jpg
│ └── ant9.jpg
└── bees
└── bee3.jpg
'''
3. DataLoader
dataloader = torch.utils.data.DataLoader(dataset=dataset, batch_size=batch_size, shuffle=True, num_workers=num_workers)
dataloader 是一个可迭代对象:
import torch.optim as optim
dataiter = iter(dataloader)
imgs, labels = next(dataiter)
# 训练时，是这样做的：
model = model_ft.to(device)
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=0.001, momentum=0.9)
scheduler = optim.lr_scheduler.StepLR(optimizer, step_size=7, gamma=0.1) # 这是一个学习率下降的东西
for epoch in range(num_epochs):
for inputs, labels in dataloaders:
inputs, labels = inputs.to(device), labels.to(device)
optimizer.zero_grad()
outputs = model(inputs)
_, preds = torch.max(outputs, 1) # 如果
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
scheduler.step()
