Pytorch学习（一）

pytorch Dataset

通过Dataset类自定义一个数据集类
该自定义数据集类可以通过路径创建一个训练集
通过图片名称创建Image对象

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

class Mydata(Dataset):
    '''
    自定义数据集类
    '''
    def __init__(self, root_dir, label_dir):
        '''
        输入：
        root_dir: 'datasets/hymenoptera_data/train'
        label_dir: 'ants'
        :param root_dir:
        :param label_dir:
        '''
        self.root_dir = root_dir
        self.label_dir = label_dir
        # 1) 获取label_path
        self.label_path = os.path.join('../',self.root_dir, self.label_dir)

        # 2) 获取image名列表
        self.image_names = os.listdir(self.label_path)

    def __getitem__(self, index):
        '''
        输入：
        index: img名字索引

        输出：
        Image对象
        :param index:
        :return:
        '''
        img_name = self.image_names[index]
        img_path = os.path.join(self.label_path, img_name)
        img = Image.open(img_path)
        label = self.label_dir
        return img, label

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

def test():
    root_dir = input("输入根目录：")
    data_label = input("输入数据集名称：")
    datasets = Mydata(root_dir, data_label)
    index = input("输入图片索引：")
    index = int(index)
    image, label = datasets.__getitem__(index)
    print('该数据集包括{}张图片'.format(datasets.__len__()))
    image.show()


if __name__ == '__main__':
    test()