PIL Image格式转Tensor

Image格式是由PIL库读入的图片格式  from PIL import Image

torch.Tensor是用于深度学习计算的张量格式 import torch

1 Image格式转Tensor

先转numpy 再转tensor
torch.from_numpy()
np.asarray()

image = torch.from_numpy(np.asarray(image))

 但是报错:
max_pool2d” not implemented for ‘Byte’
解决办法

image = torch.from_numpy(np.asarray(image)).float()

2 Tensor格式转Image

torch.Tensor需要格式是(3, 512,512)

from torchvision import transforms
unloader = transforms.ToPILImage()
image = unloader(image)
image.save('example.jpg')