Hello author,

I wanted to clarify if the results from the Cascade model in the infer.py (for 3D Fourier-Net+) at line 120:
vout_1, vout_2, vout_3 = model(x.float().to(device), y.float().to(device)) are in the spatial domain or still in the frequency domain? I am asking because when running the code, I get an error that leads back to Dv_xy having a shape of torch.Size([1, 3, 280, 336, 392]) instead of torch.Size([1, 1, 160, 192, 224]) like x_seg. And this comes from the padding being done at line 150: p3d = (84, 84, 72, 72, 60, 60).

If the model returns the result in the spatial domain, why is it necessary to redo the FFT, inverse FFT, and padding again during main() in the infer.py file? (I also noticed this section is commented out in train.py)