The performance in the paper can't to be reproduced.

[twowwj]

Hi authors,

I have retrained the model with the config "mic/gtaHR2csHR_mic_hrda_512.py".
However, the results are ~5miou lower than the performance you listed in the paper.
Do you have any explanations for this?

[leo-lab-511]

I want to ask something about how to solve the problem of this picture