Problem reproducing CRST-MRKLD result

[Nadavc220]

Hi, I ran the mrkld.sh script with the right data paths. I did not change anything else. The training process seems to have something wrong, The mIoU is going down drastically in every round all the way down to around 4% success.

Is there anything else needed to ne done on order to reproduce the 47% result?

[yzou2]

This is a similar issue as #22 .

The code is tested in Pytorch 0.4.0 and Python 2.7. We found running the code with other Pytorch versions will give different results. I suggest to run the code with the exact Pytorch version 0.4.0. Different performances on even 0.4.1 were reported by other users of this code.

[Nadavc220]

forgot to thank you for your comment. Thanks!

[quantacode]

Hi,

I found for crst_seg.py, commenting out the random seed gave me correct test results. Otherwise the accuracies are way too low. Nevertheless, running the training as it is still leads to the same problem. Can you verify if the random seed is indeed the issue?