3DPN

[cong-code]

Hello, author, I read the article xraysyn in the first stage of training output corresponds to Volume CT and Bone Mask, but your code seems to output 3 channels, corresponding to：
bone_mask = vol_pred_temp[:,[0]]
bone_ct = vol_pred_temp[:,[1]]bone_mask
tissue_ct = vol_pred_temp[:,[2]](1-bone_mask)
Is there any difference between vol_pred_temp[:,[1]] and vol_pred_temp[:,[2]]?

[cpeng93]

implementation wise we ultimately went with predicting volume CT (bone and tissue separately) and Bone mask. the tissue component is refined by the inverse of bone mask. This seemed to produce better results.