Coverting a Pytorch tensor of shape [1,3,19,19] to [1,3,9,9]

[Mathedelic]

I am using a single image GAN model(ConSINGan) for image generation. The dimension of the input image is [1,3,9,9]. However, the size of the image tensor used for interpolation is [1,3,19,19], which I want to change to [1,3,9,9]. How do I reshape torch tensors so that only the last 2 indices of the tensor are reshaped or downsampled.

I tried changing it into numpy first and then doing it, but is there any other way to do it directly without changing it to numpy?

Thank you in advance. !!!

[Nikronic (Nikan Doosti)]

Hi,

You can use torch.nn.functional.interpolate to the down/up scaling for any tensors:

Resize tensor without converting to PIL image?

Hi, You can do it using interpolate function and it supports different methods. Here is an example: import PIL.Image as Image from torchvision.transforms import ToTensor, ToPILImage import torch.nn.functional as F img = Image.open('data/Places365_val_00000001.jpg') img = ToTensor()(img) out = F.interpolate(img, size=128) #The resize operation on tensor. ToPILImage()(out).save('test.png', mode='png') Bests Nik

Bests

[Mathedelic]

Thank you so much BOSS ! This worked !