tomosu/mind-pytorch

torch implementation of MIND

mind-pytorch's Issues

What are the meaning of non local region and neigbor size?

Hi,
Thanks for sharing your cool code. I would like to ask what are the exact meaning of nonlocal region size and neighbor size in your code? I didn't find the corresponding concept in the paper. Could you help me out? Thanks！

Best,
JD

Variance calculation question

Hi,
I see your implementation of variance calculation is different from the original paper.

neighbor_images =self.neighbor_summation_patcher(self.neighbors(orig))
Vx =neighbor_images.var(dim =1).unsqueeze(dim =1)

The neighbor_images corresponds to the summation in the patch.
Thus, the variance is computed based on the patch summation. In contrast, the original MIND paper computes the var based on patch distance.
So if I want to implement the var as in the paper, I should write:
Vx = Dx_alpha.var(dim=1).unsqueeze(dim=1)

Could you please clarify if my understanding is correct?