Transforms的常见用法

一、封装函数与普通函数的用法区别

class Person:def __call__(self, name):print("__call__" + "Hello" + name)def hello(self,name):print("hello" + name)person = Person()
person("周杰伦")
person.hello("林俊杰")

运行结果：

注释：

• 双斜杠"__"封装的函数【 def __call__(self, name):】，在调用的时候直接加参数就可以使用了
  【person("周杰伦")】
• 而直接定义的函数，则需要通过 “.” +函数名来进行使用，如person.hello("林俊杰")

二、Image.open()打开图片的格式

from PIL import Image
from torch.utils.tensorboard import SummaryWriterwriter = SummaryWriter("logs")
img = Image.open('Dataset/train/ants/0013035.jpg')
print(img)

运行结果：

因此可以知道 Image.open（）打开图片的格式为PIL形

三、ToTensor打开图片格式

from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision.transforms import transformswriter = SummaryWriter("logs")
img = Image.open('Dataset/train/ants/0013035.jpg')trans_totensor = transforms.ToTensor()
img_tensor = trans_totensor(img)
print(img_tensor)

运行结果：

注：可以看到最终得到的是 tensor 形

四、ToTensor使用

from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision.transforms import transformswriter = SummaryWriter("logs")
img = Image.open('Dataset/train/ants/0013035.jpg')trans_totensor = transforms.ToTensor()
img_tensor = trans_totensor(img)
writer.add_image("ToTensor",img_tensor)
writer.close()

在终端输入 tensorboard --logdir=logs

得到链接

点开即可查看照片

五、Normalize归一化使用

归一化的计算方式：output[channel] = (input[channel] - mean[channel]) / std[channel]

from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision.transforms import transformswriter = SummaryWriter("logs")
img = Image.open('Dataset/train/ants/0013035.jpg')trans_totensor = transforms.ToTensor()
img_tensor = trans_totensor(img)
writer.add_image("ToTensor",img_tensor)
writer.close()#Normalize
print(img_tensor[0][0][0]) #将图片的第一层，第一行，第一列将他输出出来
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]) #经过归一化输出的第一层，第一行，第一列将他输出出来writer.close()

运行结果：

注：最开始是 0.31 ，经过变换之后就变成了 -0.37

将上面的代码进行输出一下

writer.add_image("Normalize",img_norm)

可以看出图片的改变

六、Resize的使用

先观察Resize的输出类型

from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision.transforms import transformswriter = SummaryWriter("logs")
img = Image.open('Dataset/train/ants/0013035.jpg')trans_totensor = transforms.ToTensor()
img_tensor = trans_totensor(img)
writer.add_image("ToTensor",img_tensor)
writer.close()#Resize
print(img.size)
trans_resize = transforms.Resize((512,512))
img_resize = trans_resize(img)
print(img_resize)writer.close()

运行结果：

可以知道Resize的输出结果是 PIL形，并且尺寸从（768,512） 变成了现在的 size = 512 x 512

观察图片

from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision.transforms import transformswriter = SummaryWriter("logs")
img = Image.open('Dataset/train/ants/0013035.jpg')trans_totensor = transforms.ToTensor()
img_tensor = trans_totensor(img)
writer.add_image("ToTensor",img_tensor)
writer.close()#Resize
print(img.size)
trans_resize = transforms.Resize((512,512))#img PIL形 --> 经过resize -> img_resize PIL形
img_resize = trans_resize(img)# img_resize PIL形 -> 经过totensor -> img_resize tensor形 ->进行图像的输出
img_resize = trans_totensor(img_resize)#图片写出
writer.add_image("Resize",img_resize,0)print(img_resize)writer.close()

运行结果：

注：可以看出图片经过了裁剪

七、Compose - Resize 使用

如果给Resize（）括号中写一个整数形int，那么图片不会改变高和宽，而是整体进行一个等比缩放
如：trans_resize_2 = transforms.Resize(512)

from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision.transforms import transformswriter = SummaryWriter("logs")
img = Image.open('Dataset/train/ants/0013035.jpg')trans_totensor = transforms.ToTensor()
img_tensor = trans_totensor(img)
writer.add_image("ToTensor",img_tensor)
writer.close()#Resize
print(img.size)
trans_resize = transforms.Resize((512,512))#img PIL形 --> 经过resize -> img_resize PIL形
img_resize = trans_resize(img)# img_resize PIL形 -> 经过totensor -> img_resize tensor形 ->进行图像的输出
img_resize = trans_totensor(img_resize)#图片写出
writer.add_image("Resize",img_resize,0)print(img_resize)writer.close()#Compose - Resize
trans_resize_2 = transforms.Resize(512)# 创建一个transforms.Compose对象，它将多个图像变换操作组合成一个序列
tran_compose = transforms.Compose([trans_resize_2,trans_totensor])img_resize_2 = tran_compose(img)# "Resize"是图像的标签，img_resize_2是经过变换的图像，1是图像在TensorBoard中的步数（step）
writer.add_image("Resize",img_resize_2,1)

注：Compose方法要求前面的输出做后面的输入，比方说tran_compose = transforms.Compose([trans_resize_2,trans_totensor])，前面的输出类型是PIL形，故要求后面的输入也要是PIL形，所以可以直接使用compose
但是如果tran_compose = transforms.Compose([trans_totensor，trans_resize_2])，也就是前面的输出是totensor形，但是后面的输入要求PIL形，这样子进行compose的话就会产生报错。

运行结果：

八、RandomCrop（） 随机裁剪用法

from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision.transforms import transformswriter = SummaryWriter("logs")
img = Image.open('Dataset/train/ants/0013035.jpg')trans_totensor = transforms.ToTensor()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()

运行结果：

自己设置高和宽：

**from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision.transforms import transformswriter = SummaryWriter("logs")
img = Image.open('Dataset/train/ants/0013035.jpg')trans_totensor = transforms.ToTensor()
#设置高为50，宽为100
trans_random = transforms.RandomCrop((50, 100))
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()**

运行结果：