Doubts with inference

[txusser]

Hi enochkan,

Thank you very much for sharing the code, it is a great resource for us coming from plain image analysis into DL. This is not properly an issue but:

I have trained your net using Nifti-1 images. I modified your code so it accepts Nifti as an input instead of hf5. My input data are 3D DATSPECT images of 128x128x128.

After the training, I am trying to perform inference using the saved generator_n.pth. I have written the following code:

generator = GeneratorUNet()
generator.load_state_dict(torch.load(opt.model_pth))
generator.eval().cuda(device=0)

img = nib.load(opt.in_path)
img_data = img.get_fdata()
tensor_arr = torch.tensor(img_data)

output = generator(tensor_arr).data.cpu().numpy()

Nevertheless, this code seems to fail with the following error:

RuntimeError: Expected 5-dimensional input for 5-dimensional weight [64, 1, 4, 4, 4], but got 3-dimensional input of size [128, 128, 128] instead

Is there something I am missing?