1.官方的數據集 MNIST

使用torchvision.datasets 里面有很多數據集供選擇

import torch
import torchvision
from torchvision import transforms, models
batch_size = 32 
transform = transforms.Compose([transforms.ToTensor(),transforms.Normalize(mean=(0.5),std=(0.5)),
])
train_data = torchvision.datasets.MNIST('./mn',train=True,download=True,transform=transform)data_loader_train = torch.utils.data.DataLoader(dataset=train_data,batch_size= batch_size,shuffle=True)test_data = torchvision.datasets.MNIST('./mn',train=False,download=True,transform=transform)data_loader_test = torch.utils.data.DataLoader(dataset=test_data,batch_size= batch_size,shuffle=True)

next(iter(data_loader_train))  # 用于查看數據

2.自建的數據集

讀取單個數據文件

device=('cuda' if torch.cuda.is_available() else 'cpu')def load_img(image_path,transform=None,max_size=None,shape=None):image=Image.open(image_path)if max_size:scale=max_size/max(image.size)size=np.array(image.size)*scaleimage=image.resize(size.astype(int),Image.ANTIALIAS)if shape:image=image.resize(shape)if transform:image=transform(image).unsqueeze(0)return image.to(device)transform = transforms.Compose([transforms.ToTensor(),transforms.Normalize(mean=[0.485,0.456,0.406],std=[0.229,0.224,0.225]),
])content = load_img('image/content.jpg',transform,max_size=400)
style = load_img('image/style.jpg',transform,max_size=400)

多張圖片的情況 ImageFloder

這個時候需要把不同label 的數據放到不同的文件夾，ImageFolder 會自動加上標簽，

from torchvison import datasets
data_dir = './data'
all_imgs=datasets.ImageFolder(os.path.join(data_dir,"train"),transforms.Compose([transforms.RandomResizedCrop(input_size),transforms.RandomHorizontalFlip(),transforms.ToTensor(),
]))
loader = torch.utils.data.DataLoader(all_imgs,batch_size=batch_size,shuffle=True)

img=next(iter(loader))[0]unloader=transforms.ToPILImage()def imshow(tensor,title=None):image=tensor.cpu().clone()image=image.squeeze(0)image=unloader(image)plt.imshow(image)if title is not None:plt.title(title)plt.pause(0.001)plt.figure()
imshow(img[31],title='image')