数据集，数据加载（基于dataset，dataloader）

从文件夹中加载Dataset

import torch.utils.data as data
from PIL import Image
import os


def default_loader(path):
    return Image.open(path).convert('RGB')
##type(return):<class 'PIL.Image.Image'>

IMG_EXTENSIONS = [
    '.jpg', '.JPG', '.jpeg', '.JPEG',
    '.png', '.PNG', '.ppm', '.PPM', '.bmp', '.BMP',
]

def is_image_file(filename):
    return any(filename.endswith(extension) for extension in IMG_EXTENSIONS)


#返回每个图像文件的路径（list）
def make_dataset(dir):
    images = []         #images保存图片文件的路径（不只是名字）
    assert os.path.isdir(dir), '%s is not a valid direntory' % dir

    # os.wark()方法
    # https://blog.csdn.net/weixin_44887621/article/details/118810180
    # os.wark()实际返回产生生成器<class 'generator'>
    # sorted(os.wark())返回有若干三元组组成的list
    # sorted()保证每次数据顺序一样
    for root, dirs, files in sorted(os.walk(dir)):
        for fname in files:
            if is_image_file(fname):
                path = os.path.join(root,fname)
                images.append(path)

    return images



class ImageFolder(data.Dataset):

    def __init__(self, root='makeup', transform=None,return_path=False,
                 loder=default_loader):
        imgs = sorted(make_dataset(root))
        if len(imgs) == 0:
            raise(RuntimeError("Found 0 image in:" + root + "\n"
                               "Supported image extensions are:" +
                               ",".join(IMG_EXTENSIONS)))
        self.root = root
        self.imgs = imgs
        self.transform = transform
        self.return_path = return_path
        self.loader = default_loader

        def __len__(self):
            return len(self.imgs)

        def __getitem__(self,index):
            path = self.imgs[index]
            img = self.loader(path)
            if self.transform is not None:
                img = self.transform(img)
            if self.return_path:
                return img, path
            else:
                return img