train_data = MyDataset(txt='train_list.txt', transform=transforms.ToTensor()) AttributeError: 'list' object has no attribute 'ToTensor'

这个错误提示很明显，是因为在传入数据集时将一个列表传入了 transform 参数，而列表对象没有 ToTensor() 属性。正确的做法是将数据集传入 DataReader 中，然后对数据集进行 ToTensor() 处理。代码示例：

import torch.utils.data as Data
import torchvision.transforms as transforms

class DataReader():
    def __init__(self, txt_path):
        self.txt_path = txt_path

    def get_dataset(self):
        with open(self.txt_path) as f:
            lines = f.readlines()

        dataset = []
        for line in lines:
            line = line.strip().split()
            img_path, label = line[0], line[1]
            dataset.append((img_path, label))

        transform = transforms.Compose([
            transforms.Resize((224, 224)),
            transforms.ToTensor(),
        ])

        dataset = MyDataset(dataset, transform=transform)

        return dataset

txt_path = "train_list.txt"
train_data = Data.DataLoader(
    dataset=DataReader(txt_path).get_dataset(),
    batch_size=32,
    shuffle=True,
    drop_last=True,
    num_workers=2,
)