PyTorch源码解读之torchvision.transforms

PyTorch框架中有一个非常重要且好用的包：torchvision，该包主要由3个子包组成，分别是：torchvision.datasets、torchvision.models、torchvision.transforms。这3个子包的具体介绍可以参考官网：http://pytorch.org/docs/master/torchvision/index.html。具体代码可以参考github：https://github.com/pytorch/vision/tree/master/torchvision。

这篇博客介绍torchvision.transformas。torchvision.transforms这个包中包含resize、crop等常见的data augmentation操作，基本上PyTorch中的data augmentation操作都可以通过该接口实现。该包主要包含两个脚本：transformas.py和functional.py，前者定义了各种data augmentation的类，在每个类中通过调用functional.py中对应的函数完成data augmentation操作。

使用例子：

import torchvision
import torch
train_augmentation = torchvision.transforms.Compose([torchvision.transforms.Resize(256),
                                                    torchvision.transforms.RandomCrop(224),                                                                            
                                                    torchvision.transofrms.RandomHorizontalFlip(),
                                                    torchvision.transforms.ToTensor(),
                                                    torch vision.Normalize([0.485, 0.456, -.406],[0.229, 0.224, 0.225])
                                                    ])

Class custom_dataread(torch.utils.data.Dataset):
    def __init__():
        ...
    def __getitem__():
        # use self.transform for input image
    def __len__():
        ...

train_loader = torch.utils.data.DataLoader(
    custom_dataread(transform=train_augmentation),
    batch_size = batch_size, shuffle = True,
    num_workers = workers, pin_memory = True)

这里定义了resize、crop、normalize等数据预处理操作，并最终作为数据读取类custom_dataread的一个参数传入，可以在内部方法__getitem__中实现数据增强操作。

主要代码在transformas.py脚本中，这里仅介绍常见的data augmentation操作，源码如下：
首先是导入必须的模型，这里比较重要的是from . import functional as F，也就是导入了functional.py脚本中具体的data augmentation函数。__all__列表定义了可以从外部import的函数名或类名。

from __future__ import division
import torch
import math
import random
from PIL import Image, ImageOps, ImageEnhance
try:
    import accimage
except ImportError:
    accimage = None
import numpy as np
import numbers
import types
import collections
import warnings

from . import functional as F

__all__ = ["Compose", "ToTensor", "ToPILImage", "Normalize", "Resize",
"Scale", "CenterCrop", "Pad", "Lambda", "RandomCrop", 
"RandomHorizontalFlip", "RandomVerticalFlip", "RandomResizedCrop", 
"RandomSizedCrop", "FiveCrop", "TenCrop","LinearTransformation", 
"ColorJitter", "RandomRotation", "Grayscale", "RandomGrayscale"]

Compose这个类是用来管理各个transform的，可以看到主要的__call__方法就是对输入图像img循环所有的transform操作

class Compose(object):
    """Composes several transforms together.

    Args:
        transforms (list of ``Transform`` objects): list of transforms to compose.

    Example:
        >>> transforms.Compose([
        >>>     transforms.CenterCrop(10),
        >>>     transforms.ToTensor(),
        >>> ])
    """

    def __init__(self, transforms):
        self.transforms = transforms

    def __call__(self, img):
        for t in self.transforms:
            img = t(img)
        return img

    def __repr__(self):
        format_string = self.__class__.__name__ + '('
        for t in self.transforms:
            format_string += '\n'
            format_string += '    {0}'.format(t)
        format_string += '\n)'
        return format_string

ToTensor类是实现：Convert a PIL Image or numpy.ndarray to tensor 的过程，在PyTorch中常用PIL库来读取图像数据，因此这个方法相当于搭建了PIL Image和Tensor的桥梁。另外要强调的是在做数据归一化之前必须要把PIL Image转成Tensor，而其他resize或crop操作则不需要。

class ToTensor(object):
    """Convert a ``PIL Image`` or ``numpy.ndarray`` to tensor.

    Converts a PIL Image or numpy.ndarray (H x W x C) in the range
    [0, 255] to a torch.FloatTensor of shape (C x H x W) in the range [0.0, 1.0].
    """

    def __call__(self, pic):
        """
        Args:
            pic (PIL Image or numpy.ndarray): Image to be converted to tensor.

        Returns:
            Tensor: Converted image.
        """
        return F.to_tensor(pic)

    def __repr__(self):
        return self.__class__.__name__ + '()'

ToPILImage顾名思义是从Tensor到PIL Image的过程，和前面ToTensor类的相反的操作。

class ToPILImage(object):
    """Convert a tensor or an ndarray to PIL Image.

    Converts a torch.*Tensor of shape C x H x W or a numpy ndarray of shape
    H x W x C to a PIL Image while preserving the value range.

    Args:
        mode (`PIL.Image mode`_): color space and pixel depth of input data (optional).
            If ``mode`` is ``None`` (default) there are some assumptions made about the input data:
            1. If the input has 3 channels, the ``mode`` is assumed to be ``RGB``.
            2. If the input has 4 channels, the ``mode`` is assumed to be ``RGBA``.
            3. If the input has 1 channel, the ``mode`` is determined by the data type (i,e,
            ``int``, ``float``, ``short``).

    .. _PIL.Image mode: http://pillow.readthedocs.io/en/3.4.x/handbook/concepts.html#modes
    """
    def __init__