Refinement Module in E2E-VarNet Implementation #229
-
|
Hello, I am reading through the 2020 paper from Sriram A et al on their work with E2E-VarNet and have a question on this line in the fastMRI repository. Specifically, line 10 of their paper suggests that the refinement module output should be added to the previous predicted kspace, whereas the implementation in the repository seems like the refinement module output is being subtracted. Just want to check to see which approach is intended and if I'm thinking about their implementation correctly. Thank you in advance! |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
|
Hello @michael-s-yao, it looks like this formulation is in the original open-sourcing by the authors here. The private internal code used for this paper also follows this convention, so I think equation (10) in the paper has a typo. |
Beta Was this translation helpful? Give feedback.
-
|
Got it, thanks @mmuckley for the explanation! |
Beta Was this translation helpful? Give feedback.
Hello @michael-s-yao, it looks like this formulation is in the original open-sourcing by the authors here. The private internal code used for this paper also follows this convention, so I think equation (10) in the paper has a typo.
CC @anuroopsriram