| 知乎专栏 |
Compose()： 用来管理所有的transforms操作。
ToTensor()： 把图片数据转换成张量并转化范围在[0,1]区间内。
Normalize(mean, std)： 归一化。
Resize(size)： 输入的PIL图像调整为指定的大小，参数可以为int或int元组。
CenterCrop(size)： 将给定的PIL Image进行中心切割，得到指定size的tuple。
RandomCrop(size, padding=0)：随机中心点切割。
RandomHorizontalFlip(size, interpolation=2)：将给定的PIL Image随机切割，再resize。
RandomHorizontalFlip()：随机水平翻转给定的PIL Image。
RandomVerticalFlip()： 随机垂直翻转给定的PIL Image。
ToPILImage()： 将Tensor或numpy.ndarray转换为PIL Image。
FiveCrop(size)： 将给定的PIL图像裁剪成4个角落区域和中心区域。
Pad(padding, fill=0, padding_mode=‘constant’)：对PIL边缘进行填充。
RandomAffine(degrees, translate=None, scale=None)：保持中心不变的图片进行随机仿射变化。
RandomApply(transforms, p=0.5)：随机选取变换。
from torchvision import transforms
# 数据预处理
transform = transforms.Compose([transforms.RandomCrop(32), transforms.Resize(256)])
# 对图像进行转换
img = Image.open('example.jpg')
img_transformed = transform(img)
将 PIL Image 或 numpy.ndarray 转为 tensor
import numpy as np
from torchvision import transforms
data = np.array([
[0, 5, 10, 20, 0],
[255, 125, 180, 255, 196]
], dtype=np.uint8)
tensor = transforms.ToTensor()(data)
print(tensor)
"""
tensor([[[0.0000, 0.0196, 0.0392, 0.0784, 0.0000],
[1.0000, 0.4902, 0.7059, 1.0000, 0.7686]]])
"""
from torchvision.transforms PIL_img = transforms.ToPILImage()(tensor_img)
import torchvision.models as models
resnet18 = models.resnet18(pretrained=True)
vgg16 = models.vgg16(pretrained=True)
alexnet = models.alexnet(pretrained=True)
squeezenet = models.squeezenet1_0(pretrained=True)
densenet = models.densenet_161()
from torchvision import models # 加载预训练的ResNet-50模型 model = models.resnet50(pretrained=True)
# 数据预处理 transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5,), (0.5,)]) # 加载数据集 trainset = datasets.CIFAR10(root='./data', train=True, download=True, transform=transform) testset = datasets.CIFAR10(root='./data', train=False, download=True, transform=transform)
datasets.ImageFolder方法可以实现数据导入。 ImageFolder(root,transform=None,target_transform=None,loader=default_loader)
from torchvision import datasets train_datasets = datasets.CIFAR10(root='data',train=False,download=True) print(train_datasets)
/Users/neo/PycharmProjects/netkiller/venv/bin/python /Users/neo/PycharmProjects/netkiller/test.py
Downloading https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz to data/cifar-10-python.tar.gz
100.0%
Extracting data/cifar-10-python.tar.gz to data
Dataset CIFAR10
Number of datapoints: 10000
Root location: data
Split: Test
from torchvision import datasets train_datasets = datasets.CIFAR10(root='data',train=False,download=True) print(train_datasets.data)
from torchvision import datasets train_datasets = datasets.CIFAR10(root='data',train=False,download=True) print(train_datasets.data.shape)
Files already downloaded and verified (10000, 32, 32, 3)
(图片数量, 高度, 宽度, 通道数)
from torchvision import datasets train_datasets = datasets.CIFAR10(root='data',train=False,download=True) print(train_datasets.data.min()) print(train_datasets.data.max())
Files already downloaded and verified 0 255
from torchvision import datasets train_datasets = datasets.CIFAR10(root='data',train=False,download=True) print(train_datasets.classes) print(train_datasets.class_to_idx)
Files already downloaded and verified
['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']
{'airplane': 0, 'automobile': 1, 'bird': 2, 'cat': 3, 'deer': 4, 'dog': 5, 'frog': 6, 'horse': 7, 'ship': 8, 'truck': 9}
单张图片
from torchvision import datasets from matplotlib import pyplot as plt train_datasets = datasets.CIFAR10(root='data',train=False,download=True) plt.imshow(train_datasets.data[0]) plt.show()
显示3x3九张图片
from torchvision import datasets
from matplotlib import pyplot as plt
test_datasets = datasets.CIFAR10(root='data',train=False,download=True)
fig,axes=plt.subplots(3,3,figsize=(4,4))
for i,ax in enumerate(axes.flat):
ax.imshow(test_datasets.data[i])
ax.axis("off")
ax.set_title(test_datasets.classes[test_datasets.targets[i]])
plt.show()