Questions about DIRLAB dataset

[Valeyards]

Hi Vincent, thank you so much for sharing your work and such concise code. I’m trying to reimplement your experiments, but I come across some problems:
While training both losses and TRE keeps going up

I thought it was about my data processing since the 'mhd' file is acquired through numpy, so I checked #2 and followed that instruction to obtain mhd file.
The result is quite surprising for me since their values are actually the same and abnormal I think?(1.mhd is through MATLAB Elastix and case1_T00.mhd is calculated by myself)

I'll be glad if you could reply asap :)