TypeError: img should be PIL Image. Got class torch.Tensor

背景：

在pytorch中使用MNIST数据集，进行可视化，代码如下：

import torch
import torch.nn as nn
import torchvision
from torchvision import datasets, transforms
from torch.autograd import Variable
import matplotlib.pyplot as plt

# part 1: 数据集的获取，torch中提供了数据集的相关API
mnist_train_dataset = datasets.MNIST(root="./data/",
                                      train=True,
                                      download=True,
                                      transform=
                                        transforms.Compose([transforms.ToTensor(), transforms.Normalize(mean=[0.5],std=[0.5]),transforms.Resize((28,28))])
                                    )
                                        
mnist_test_dataset = datasets.MNIST(root="./data/",
                                      train=False,
                                      download=True,
                                      transform = transforms.Compose([transforms.ToTensor(), transforms.Resize((28,28)))
                  )

# part 2: 数据装载， dataloader
data_loader_train = torch.utils.data.DataLoader(
    dataset=mnist_train_dataset,
    batch_size=128,
    shuffle=True
)

data_loader_test = torch.utils.data.DataLoader(
    dataset=mnist_test_dataset,
    batch_size = 1,
    shuffle=True
)


# part 3: 数据可视化，检查数据
images,labels = next(iter(data_loader_train))
# TypeError: img should be PIL Image. Got <class 'torch.Tensor'>
img = torchvision.utils.make_grid(images)
img = img.numpy().transpose(1,2,0)
std=mean=[0.5,0.5,0.5]
img = img * std + mean
# 直接imshow会报错：Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
# 意思是需要归一化处理
print([int(labels[i].numpy()) for i,label in enumerate(labels)])
plt.imshow(img)
plt.show()

运行会出现以下报错：

Traceback (most recent call last):
  File "d:/GitHub/studyNote/pytorch基础/mnist.torch.py", line 45, in <module>
    images,labels = next(iter(data_loader_train))
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torch\utils\data\dataloader.py", line 560, in __next__
    batch = self.collate_fn([self.dataset[i] for i in indices])
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torch\utils\data\dataloader.py", line 560, in <listcomp>
    batch = self.collate_fn([self.dataset[i] for i in indices])
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torchvision\datasets\mnist.py", line 95, in __getitem__
    img = self.transform(img)
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torchvision\transforms\transforms.py", line 61, in __call__
    img = t(img)
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torchvision\transforms\transforms.py", line 196, in __call__
    return F.resize(img, self.size, self.interpolation)
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torchvision\transforms\functional.py", line 229, in resize
    raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
TypeError: img should be PIL Image. Got <class 'torch.Tensor'>

思考：

他需要PIL格式的图片，正好transforms中有一个方法为：transforms.ToPILImage(), 然后就变成了

transform=transforms.Compose([
    						transforms.ToTensor(),                               
                              transforms.Normalize(mean=[0.5],std=[0.5
                              transforms.Resize([28,28]),
                              transforms.ToPILImage()
                             ])

但是还是报错：

Traceback (most recent call last):
  File "d:/GitHub/studyNote/pytorch基础/mnist.torch.py", line 45, in <module>
    images,labels = next(iter(data_loader_train))
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torch\utils\data\dataloader.py", line 560, in __next__
    batch = self.collate_fn([self.dataset[i] for i in indices])
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torch\utils\data\dataloader.py", line 560, in <listcomp>
    batch = self.collate_fn([self.dataset[i] for i in indices])
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torchvision\datasets\mnist.py", line 95, in __getitem__
    img = self.transform(img)
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torchvision\transforms\transforms.py", line 61, in __call__
    img = t(img)
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torchvision\transforms\transforms.py", line 196, in __call__
    return F.resize(img, self.size, self.interpolation)
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torchvision\transforms\functional.py", line 229, in resize
    raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
TypeError: img should be PIL Image. Got <class 'torch.Tensor'>

去bing上查询到stackoverflow上发现一个类似的错误：

train_transforms = transforms.Compose(
[transforms.Resize(255), 
transforms.CenterCrop(224), 
transforms.ToTensor(), 
transforms.RandomHorizontalFlip(), 
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])])

TypeError: img should be PIL Image. Got <class ‘torch.Tensor’>

– from https://stackoverflow.com/questions/57079219/img-should-be-pil-image-got-class-torch-tensor

下边大神的解决方案是：

transforms.RandomHorizontalFlip() works on PIL.Images, not torch.Tensor. In your code above, you are applying transforms.ToTensor() prior to transforms.RandomHorizontalFlip(), which results in tensor.

transforms.RandomHorizontalFlip() works on PIL.Images, not torch.Tensor. In your code above, you are applying transforms.ToTensor() prior to transforms.RandomHorizontalFlip(), which results in tensor.

But, as per the official pytorch documentation here,

transforms.RandomHorizontalFlip() horizontally flip the given PIL Image randomly with a given probability.

So, just change the order of your transformation in above code, like below:

train_transforms = transforms.Compose([transforms.Resize(255), 
                                       transforms.CenterCrop(224),  
                                       transforms.RandomHorizontalFlip(),
                                       transforms.ToTensor(), 
                                       transforms.Normalize([0.485, 0.456, 0.406], 										[0.229, 0.224, 0.225])])

发现是顺序问题，需要调换，将ToTensor放在RandomHorizontalFlip之后。

解决：

这个问题我们也采用相同方法尝试。

从

transform=transforms.Compose([
    						transforms.ToTensor(),                               
                              transforms.Normalize(mean=[0.5],std=[0.5
                              transforms.Resize([28,28]),
                              transforms.ToPILImage()
                             ])

改为：

transform=transforms.Compose([
    						transforms.Resize([28,28]),
    						transforms.ToTensor(),                               
                              transforms.Normalize(mean=[0.5],std=[0.5])
                              # transforms.ToPILImage()
                             ])

对这个顺序比较好奇,又尝试了一下：

transform=transforms.Compose([
                              transforms.Scale([28,28]),
                              transforms.Normalize(mean=[0.5],std=[0.5])，
                              transforms.ToTensor()
                             ])

发现报错：

  File "d:/GitHub/studyNote/pytorch基础/mnist.torch.py", line 47, in <module>
    images,labels = next(iter(data_loader_train))
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torch\utils\data\dataloader.py", line 560, in __next__
    batch = self.collate_fn([self.dataset[i] for i in indices])
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torch\utils\data\dataloader.py", line 560, in <listcomp>
    batch = self.collate_fn([self.dataset[i] for i in indices])
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torchvision\datasets\mnist.py", line 95, in __getitem__
    img = self.transform(img)
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torchvision\transforms\transforms.py", line 61, in __call__
    img = t(img)
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torchvision\transforms\transforms.py", line 164, in __call__
    return F.normalize(tensor, self.mean, self.std, self.inplace)
  File "E:\ProgramData\Miniconda3\envs\pytorch\lib\site-packages\torchvision\transforms\functional.py", line 201, in normalize
    raise TypeError('tensor is not a torch image.')
TypeError: tensor is not a torch image.

看来ToTensor需要在Normalize之前才行。

大家如果有新的发现可以在评论补充