首页 > itarticle > torchvision note

torchvision note

admin 1月 30, 2021 0

README

references

notes

torchvison is so fabulous, including popular datasets, model architectures and common image transformation.

datasets

load train and test

import torch
import torchvision
import torchvision.transforms as transforms


# (0.5, 0.5, 0.5), (0.5, 0.5, 0.5)分别是mean和std
transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
trainset = torchvision.datasets.CIFAR10(root='/home/carry/ifs/pytorch/cifar10', train=True, download=True, transform=transform)
trainloader = torch.utils.data.DataLoader(trainset, batch_size=4, shuffle=True, num_workers=2)
testset = torchvision.datasets.CIFAR10(root='/home/carry/ifs/pytorch/cifar10', train=False, download=True, transform=transform)
testloader = torch.utils.data.DataLoader(testset, batch_size=4, shuffle=False, num_workers=2)

classes = ('plane', 'car', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck')

cifar10

import torch
import torchvision
import torchvision.transforms as transforms
import matplotlib.pyplot as plt
import numpy as np
 = torchvision.datasets.CIFAR10(root='/home/carry/ifs/pytorch/cifar10', train=True,
                                        download=True, transform=transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))]))
print('total number:%d' % len(trainset))
trainloader = torch.utils.data.DataLoader(, batch_size=4,
                                          shuffle=True, num_workers=2)
print('total batches:%d' % len(trainloader))
dataiter = iter(trainloader)
images, labels = dataiter.next()
print(images.size())
plt.imshow(np.transpose((torchvision.utils.make_grid(images)/2+0.5).numpy(), (1, 2, 0)))

mnist

import torch
import torchvision
import torchvision.transforms as transforms
import matplotlib.pyplot as plt
import numpy as np
 = torchvision.datasets.MNIST(root='/media/data/hjr/pytorch/mnist', train=True,
                                        download=True, transform=transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))]))
print('total number:%d' % len(trainset))
trainloader = torch.utils.data.DataLoader(, batch_size=4,
                                          shuffle=True, num_workers=2)
print('total batches:%d' % len(trainloader))
dataiter = iter(trainloader)
images, labels = dataiter.next()
print(images.size())

plt.imshow((torchvision.utils.make_grid(images)/2+0.5).permute(1, 2, 0))

Fashionmnist

import torch
import torchvision
import torchvision.transforms as transforms
import matplotlib.pyplot as plt
import numpy as np
 = torchvision.datasets.FashionMNIST(root='/home/carry/ifs/pytorch/fashionmnist', train=True,
                                        download=True, transform=transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))]))
print('total number:%d' % len(trainset))
trainloader = torch.utils.data.DataLoader(trainset, batch_size=4,
                                          shuffle=True, num_workers=2)
print('total batches:%d' % len(trainloader))
dataiter = iter(trainloader)
images, labels = dataiter.next()
print(images.size())
plt.imshow(np.transpose((torchvision.utils.make_grid(images)/2+0.5).numpy(), (1, 2, 0)))