惠勒提问法

问题1：如何在自己定义的类WineDateset(继承自Dateset)实现transform?

问题2：多个transform如何作用在dataset上？

代码

'''
Transforms can be applied to PIL images, tensors, ndarrays, or custom data
during creation of the DataSet

complete list of built-in transforms: 
https://pytorch.org/docs/stable/torchvision/transforms.html

'''

import torch
import torchvision
from torch.utils.data import Dataset
import numpy as np


class WineDataset(Dataset):

    def __init__(self, transform=None):
        xy = np.loadtxt('./data/wine/wine.csv', delimiter=',', dtype=np.float32, skiprows=1)
        self.n_samples = xy.shape[0]

        # note that we do not convert to tensor here
        self.x_data = xy[:, 1:]
        self.y_data = xy[:, [0]]

        self.transform = transform

    def __getitem__(self, index):
        sample = self.x_data[index], self.y_data[index]

        if self.transform:
            sample = self.transform(sample)

        return sample

    def __len__(self):
        return self.n_samples

# 答案1 在自定义的数据集类中加上transform参数，然后自定义类，实现对数据的变换，
# 利用__call__类函数实现调用。

# Custom Transforms
# implement __call__(self, sample)
class ToTensor:
    # Convert ndarrays to Tensors
    def __call__(self, sample):
        inputs, targets = sample
        return torch.from_numpy(inputs), torch.from_numpy(targets)

class MulTransform:
    # multiply inputs with a given factor
    def __init__(self, factor):
        self.factor = factor

    def __call__(self, sample):
        inputs, targets = sample
        inputs *= self.factor
        return inputs, targets

print('Without Transform')
dataset = WineDataset()
first_data = dataset[0]
features, labels = first_data
print(type(features), type(labels))
print(features, labels)

print('\nWith Tensor Transform')
dataset = WineDataset(transform=ToTensor())
first_data = dataset[0]
features, labels = first_data
print(type(features), type(labels))
print(features, labels)

# 答案2:借助torchvision.transforms.Compose实现。

print('\nWith Tensor and Multiplication Transform')
composed = torchvision.transforms.Compose([ToTensor(), MulTransform(4)])
dataset = WineDataset(transform=composed)
first_data = dataset[0]
features, labels = first_data
print(type(features), type(labels))
print(features, labels)

输出结果