Generate cubic input and data augmentation

[PengfeiRen96]

I have two questions as follows: (1) Firstly, I know the corpped img size is 96(pixel) in DeepPrior++ and REN,but i was confused about why you set oringinalSz =96 in generate_cubic_input()?And why you use "originalSz" and "croppedSz" variables. (2) Secondly, in generate_cubic_input(), why you use the following code to resize the data instead of multiplying a coefficient directly？ if newSz < 100 then coord = coord / originalSz math.floor(originalSznewSz/100) + math.floor(originalSz/2 - originalSz/2newSz/100) elseif newSz > 100 then coord = coord / originalSz math.floor(originalSznewSz/100) - math.floor(originalSz/2newSz/100 - originalSz/2) end

[mks0601]

(1) I don`t know the cropped img size in DeepPrior++ and REN is 96, but that is not related with my setting. The reason why I defined originalSz and croppedSz is for translation data augmentation.

(2) what do you mean by multiplying a coefficient? which coefficient do you mean?

[PengfeiRen96]

Thank you for your reply. (1)Now i understand the originalSz and croppedSz is defined for translation data augmentation. And i wonder to know does the translation is [-4, 4] voxels in your code: local trans = torch.Tensor(worldDim) trans[1] = torch.random(1,originalSz-croppedSz+1) trans[2] = torch.random(1,originalSz-croppedSz+1) trans[3] = torch.random(1,originalSz-croppedSz+1) --discretize coord = discretize(coord) coord = coord + (originalSz/2 - croppedSz/2) --translation coord[1] = coord[1] - (trans[1] - 1) coord[2] = coord[2] - (trans[2] - 1) coord[3] = coord[3] - (trans[3] - 1) (2)I mistaked for the meaning of “newSz”.I used to think is stand for voxel.

[mks0601]

Sorry for late reply. (1) No. The translation is [-8. 8] as described in the paper.