PIL:使用Python自带图像处理库读取出来的图片格式
numpy：使用Python-opencv库读取出来的图片格式
tensor：pytorch中训练时所采取的向量格式
import torch import torchvision.transforms as transforms
PIL to Tensor
1 def PIL2tensor(img): 2 loader = transforms.Compose([ 3 transforms.ToTensor() 4 ]) 5 image = loader(img).unsqueeze(0) 6 return image.to(torch.device,torch.float)
Tensor to PIL
1 def tensor2PIL(tensor): # 将tensor-> PIL 2 unloader = transforms.ToPILImage() 3 image = tensor.cpu().clone() 4 image = image.squeeze(0) 5 image = unloader(image) 6 return image
Tensor to numpy
1 def tensor2numpy(tensor): 2 img = tensor.mul(255).byte() 3 img = img.cpu().numpy().squeeze(0).tranpose((1,2,0)) 4 return img
numpy to tensor
if isinstance(pic,np.ndarray):
img = torch.from_numpy(pic.transpose((2,0,1)))
return img.float().div(255)
pytorch_13_pytorch 中tensor，numpy，PIL的转换
原文：https://www.cnblogs.com/shuangcao/p/11930743.html