pytorch优雅的反归一化

非常优雅的两句话
编程的魅力

        for i in range(hyperparameters['batch_size']):

            img_gen = x_ba[i].cpu()
            img_GT = x_a[i].cpu()

            #img_GT = img_GT.numpy()
            #print(img_GT.shape)
            #img_GT.save("./output_GT.jpg")
#########unnormalize
            mean = (0.5, 0.5, 0.5)
            std = (0.5, 0.5, 0.5)
            #t_mean = torch.FloatTensor(mean).view(3,1,1).expand(3,H,W).contiguous().view(1,3,H,W)
            #t_std = torch.FloatTensor(std).view(3,1,1).expand(3,H,W).contiguous().view(1,3,H,W)
            #x = y * t_std.expand(B,3,H,W) +t_mean.expand(B,3,H,W)
            t_mean = torch.FloatTensor(mean).view(3,1,1).expand(3, 120, 120)
            t_std = torch.FloatTensor(std).view(3,1,1).expand(3, 120, 120)

            print(t_mean.shape)
            print(t_std.shape)
            print(img_GT.shape)

            img_GT = img_GT * t_std +t_mean

            img = torchvision.transforms.ToPILImage()(img_GT).convert('RGB')


            img.save("./output_GT.jpg")