pytorch--transforms.RandomChoice 使用code

import numpy as np
from sklearn.model_selection import StratifiedShuffleSplit
import  PIL
from  torchvision  import transforms
from transforms import *

from PIL import Image
import matplotlib.pyplot as plt

subpolicies =[]
sub=[]
subpolicy1 = transforms.Compose([
            ## baseline augmentation
            transforms.Resize([32, 32]),
            transforms.RandomCrop([120, 720]),
            transforms.RandomHorizontalFlip(),
            transforms.ToTensor()])

subpolicy2 = transforms.Compose([
            ## baseline augmentation
            transforms.Resize([32, 64]),
            transforms.RandomCrop([120, 720]),
            # transforms.Pad(4),
            # transforms.RandomCrop(32),
            # transforms.RandomCrop([140, 720]),
            transforms.RandomHorizontalFlip(),
            ## policy
            # *subpolicy,
            ## to tensor
            transforms.ToTensor()])
tran1 = transforms.Compose([transforms.Resize([64, 64]),transforms.CenterCrop((12, 12)) ])
tran2= transforms.Compose([transforms.Resize([32, 32]), transforms.CenterCrop((20, 12))])

sub=[tran1 , tran2]

print("subpolicy2  222", sub)
print("subpolicies", sub[0])

transform1 = transforms.RandomChoice(sub)
print("trans1 ", transform1)

img_rgb = Image.open("./cifar-10-images/temp/1_dog.jpg")

# img_store = t.tensor(img_rgb).permute(2,0, 1)
img_store = transform1(img_rgb)


img_store = np.uint8(img_store)
plt.figure("dog")
plt.imshow(img_store)
plt.show()

备注：tran1  和tran2 如果转换成tensor 则需要转换后，方可图像显示。