【Pytorch】常见的Transforms

Lucy@IshtarXu

已于 2022-02-06 22:21:34 修改

阅读量488

点赞数

于 2022-02-06 22:18:38 首次发布

本文链接：https://blog.csdn.net/FelicityXu/article/details/122799062

版权

深度学习同时被 3 个专栏收录

43 篇文章 15 订阅

订阅专栏

Python

32 篇文章 1 订阅

订阅专栏

Pytorch

21 篇文章 2 订阅

订阅专栏

1 - Python语法复习：call的用法

创建一个新的python文件，输入

class Person:
    def __call__(self, name):
        print("__call__"+"hello "+name)

    def hello(self, name):
        print("hello"+ name)

person = Person()
person("zhangsan")
person.hello("lisi")

输出结果为

__call__hello zhangsan
hellolisi

2 - Normalize的使用

Normalize的作用为归一化和标准化
在这里插入图片描述

输入下述代码

from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision import transforms

writer = SummaryWriter("logs")
img = Image.open("dataset/hymenoptera_data/train/ants_image/0013035.jpg")
print(img)

# ToTensor PIL改为Tensor类型
trans_totensor = transforms.ToTensor()
img_tensor = trans_totensor(img)
writer.add_image("ToTensor", img_tensor)

# Normalize 计算均值和标准差
# """Normalize a tensor image with mean and standard deviation.
#     This transform does not support PIL Image.
#     Given mean: ``(mean[1],...,mean[n])`` and std: ``(std[1],..,std[n])`` for ``n``
#     channels, this transform will normalize each channel of the input
#     ``torch.*Tensor`` i.e.,
#     ``output[channel] = (input[channel] - mean[channel]) / std[channel]``
# tensor (0,1) -> (-1,1)
print(img_tensor[0][0][0])
trans_norm = transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])
img_norm = trans_norm(img_tensor)
print(img_norm[0][0][0])
writer.add_image("Normalize", img_norm)

writer.close()

运行后，可以发现输出

D:\Anaconda3\envs\pytorch\python.exe D:/研究生/代码尝试/P10_UsefulTransforms.py
<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=768x512 at 0x20B126FA040>
tensor(0.3137)
tensor(-0.3725)

在终端输入

(pytorch) D:\研究生\代码尝试>tensorboard --logdir=logs

打开网址

TensorFlow installation not found - running with reduced feature set.
Serving TensorBoard on localhost; to expose to the network, use a proxy or pass --bind_all
TensorBoard 2.8.0 at http://localhost:6006/ (Press CTRL+C to quit)

在这里插入图片描述
阴间的蚂蚁就出现啦

3 - Resize的使用

Resize的作用为缩放

from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision import transforms

writer = SummaryWriter("logs")
img = Image.open("dataset/hymenoptera_data/train/ants_image/0013035.jpg")
print(img)

# ToTensor PIL改为Tensor类型
trans_totensor = transforms.ToTensor()
img_tensor = trans_totensor(img)
writer.add_image("ToTensor", img_tensor)

# Normalize 计算均值和标准差
# """Normalize a tensor image with mean and standard deviation.
#     This transform does not support PIL Image.
#     Given mean: ``(mean[1],...,mean[n])`` and std: ``(std[1],..,std[n])`` for ``n``
#     channels, this transform will normalize each channel of the input
#     ``torch.*Tensor`` i.e.,
#     ``output[channel] = (input[channel] - mean[channel]) / std[channel]``
# tensor (0,1) -> (-1,1)
print(img_tensor[0][0][0])
trans_norm = transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])
img_norm = trans_norm(img_tensor)
print(img_norm[0][0][0])
writer.add_image("Normalize", img_norm)

# Resize
"""Resize the input image to the given size.
   If the image is torch Tensor, it is expected
   to have [..., H, W] shape, where ... means an arbitrary number of leading dimensions"""
print(img.size)
trans_resize = transforms.Resize((512, 512))
# img PIL -> resize -> img_resize PIL
img_resize = trans_resize(img)
# img_resize PIL -> Totensor -> img_resize tensor
img_resize = trans_totensor(img_resize)
writer.add_image("Resize", img_resize, 0)
print(img_resize)

writer.close()

输出结果为

<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=768x512 at 0x18990DDA040>
tensor(0.3137)
tensor(-0.3725)
(768, 512)
tensor([[[0.3137, 0.3137, 0.3176,  ..., 0.3137, 0.3137, 0.3020],
         [0.3176, 0.3176, 0.3176,  ..., 0.3098, 0.3137, 0.3020],
         [0.3216, 0.3216, 0.3176,  ..., 0.3059, 0.3137, 0.3059],
         ...,
         [0.3412, 0.3373, 0.3373,  ..., 0.0196, 0.2196, 0.3608],
         [0.3412, 0.3373, 0.3373,  ..., 0.3490, 0.3373, 0.3373],
         [0.3412, 0.3373, 0.3373,  ..., 0.3529, 0.3137, 0.3216]],

        [[0.5922, 0.5922, 0.5961,  ..., 0.5922, 0.5922, 0.5804],
         [0.5961, 0.5961, 0.5961,  ..., 0.5882, 0.5922, 0.5804],
         [0.6000, 0.6000, 0.5961,  ..., 0.5843, 0.5922, 0.5843],
         ...,
         [0.6275, 0.6235, 0.6235,  ..., 0.1020, 0.4157, 0.6157],
         [0.6275, 0.6235, 0.6235,  ..., 0.5373, 0.5882, 0.6078],
         [0.6275, 0.6235, 0.6235,  ..., 0.6392, 0.6275, 0.6275]],

        [[0.9137, 0.9137, 0.9176,  ..., 0.9137, 0.9137, 0.9020],
         [0.9176, 0.9176, 0.9176,  ..., 0.9098, 0.9137, 0.9020],
         [0.9216, 0.9216, 0.9176,  ..., 0.9059, 0.9137, 0.9059],
         ...,
         [0.9294, 0.9255, 0.9255,  ..., 0.1961, 0.6353, 0.9059],
         [0.9294, 0.9255, 0.9255,  ..., 0.7922, 0.9098, 0.9451],
         [0.9294, 0.9255, 0.9255,  ..., 0.9412, 0.9569, 0.9373]]])

打开Tensorboard看看吧~
在这里插入图片描述

4 - Compose的使用

将几种transform进行组合
在这里插入图片描述

from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision import transforms

writer = SummaryWriter("logs")
img = Image.open("dataset/hymenoptera_data/train/ants_image/0013035.jpg")
print(img)

# ToTensor PIL改为Tensor类型
trans_totensor = transforms.ToTensor()
img_tensor = trans_totensor(img)
writer.add_image("ToTensor", img_tensor)

# Normalize 计算均值和标准差
# """Normalize a tensor image with mean and standard deviation.
#     This transform does not support PIL Image.
#     Given mean: ``(mean[1],...,mean[n])`` and std: ``(std[1],..,std[n])`` for ``n``
#     channels, this transform will normalize each channel of the input
#     ``torch.*Tensor`` i.e.,
#     ``output[channel] = (input[channel] - mean[channel]) / std[channel]``
# tensor (0,1) -> (-1,1)
print(img_tensor[0][0][0])
trans_norm = transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])
img_norm = trans_norm(img_tensor)
print(img_norm[0][0][0])
writer.add_image("Normalize", img_norm)

# Resize
"""Resize the input image to the given size.
   If the image is torch Tensor, it is expected
   to have [..., H, W] shape, where ... means an arbitrary number of leading dimensions"""
print(img.size)
trans_resize = transforms.Resize((512, 512))
# img PIL -> resize -> img_resize PIL
img_resize = trans_resize(img)
# img_resize PIL -> Totensor -> img_resize tensor
img_resize = trans_totensor(img_resize)
writer.add_image("Resize", img_resize, 0)
print(img_resize)

# Compose - resize - 2
# Compose就是将函数的功能进行整合，设定一个模板，按照模板中设定好的操作处理
    """Composes several transforms together. This transform does not support torchscript.
    Please, see the note below."""
trans_resize_2 = transforms.Resize(512)
# PIL -> PIL -> tensor
trans_compose = transforms.Compose([trans_resize_2, trans_totensor])
img_resize_2 = trans_compose(img)
writer.add_image("Resize", img_resize_2, 1)
writer.close()

输出结果为
在这里插入图片描述

5 - RandomCrop的用法

作用：随机裁剪图片

from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision import transforms

writer = SummaryWriter("logs")
img = Image.open("dataset/hymenoptera_data/train/ants_image/0013035.jpg")
print(img)

# ToTensor PIL改为Tensor类型
trans_totensor = transforms.ToTensor()
img_tensor = trans_totensor(img)
writer.add_image("ToTensor", img_tensor)

# Normalize 计算均值和标准差
# """Normalize a tensor image with mean and standard deviation.
#     This transform does not support PIL Image.
#     Given mean: ``(mean[1],...,mean[n])`` and std: ``(std[1],..,std[n])`` for ``n``
#     channels, this transform will normalize each channel of the input
#     ``torch.*Tensor`` i.e.,
#     ``output[channel] = (input[channel] - mean[channel]) / std[channel]``
# tensor (0,1) -> (-1,1)
print(img_tensor[0][0][0])
trans_norm = transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])
img_norm = trans_norm(img_tensor)
print(img_norm[0][0][0])
writer.add_image("Normalize", img_norm)

# Resize
"""Resize the input image to the given size.
   If the image is torch Tensor, it is expected
   to have [..., H, W] shape, where ... means an arbitrary number of leading dimensions"""
print(img.size)
trans_resize = transforms.Resize((512, 512))
# img PIL -> resize -> img_resize PIL
img_resize = trans_resize(img)
# img_resize PIL -> Totensor -> img_resize tensor
img_resize = trans_totensor(img_resize)
writer.add_image("Resize", img_resize, 0)
print(img_resize)

# Compose - resize - 2
# Compose就是将函数的功能进行整合，设定一个模板，按照模板中设定好的操作处理
trans_resize_2 = transforms.Resize(512)
# PIL -> PIL -> tensor
trans_compose = transforms.Compose([trans_resize_2, trans_totensor])
img_resize_2 = trans_compose(img)
writer.add_image("Resize", img_resize_2, 1)

# RandomCrop
 """Crop the given image at a random location.
    If the image is torch Tensor, it is expected
    to have [..., H, W] shape, where ... means an arbitrary number of leading dimensions,
    but if non-constant padding is used, the input is expected to have at most 2 leading dimensions"""
trans_random = transforms.RandomCrop(512)
trans_compose_2 = transforms.Compose([trans_random, trans_totensor])
for i in range(10):
    img_crop = trans_compose_2(img)
    writer.add_image("RandomCrop", img_crop, i)

writer.close()

不知道返回值的时候

*print
*print(type())
*debug

Lucy@IshtarXu

关注

0
点赞
踩
0

收藏

觉得还不错? 一键收藏
0
评论
【Pytorch】常见的Transforms

1 - Python语法复习：__call__的用法创建一个新的python文件，输入class Person: def __call__(self, name): print("__call__"+"hello "+name) def hello(self, name): print("hello"+ name)person = Person()person("zhangsan")person.hello("lisi")输出结果为__ca
复制链接

扫一扫