【深度学习】torchvision.transforms对两张图片(一组图片/源图像与标签)进行完全相同的随机裁剪

参考https://blog.csdn.net/qq_34106090/article/details/124254586
对两张图片进行裁剪的关键在于为它们指定相同的随机数种子

import os.path

import torch
import matplotlib.pyplot as plt
from PIL import Image
from torchvision.transforms import transforms

transform = transforms.Compose([
    transforms.RandomCrop(size=256)
    # transforms.ToTensor()
])

render_img = Image.open(os.path.join('D:/test_datasets/B07263589Q/render/0/render_0.jpg'))
mask_img = Image.open(os.path.join('D:/test_datasets/B07263589Q/segmentation/segmentation_0.jpg'))

seed = torch.random.seed()

torch.random.manual_seed(seed)
render_img = transform(render_img)
torch.random.manual_seed(seed)
mask_img = transform(mask_img)

plt.imshow(render_img)
plt.show()
plt.imshow(mask_img)
plt.show()