Bad ssim result for the removal setting.

[Ha0Tang]

I used the provided script train_cityscapes.sh to retrain a modal and then used the script test_cityscapes_addition.sh to test the model on the addition setting, resulting in almost the same SSIM result reported in your paper.

However, I used the script test_cityscapes_removal.sh test on the mentioned-above modal to obtain the SSIM result of the removal setting, leading to only 0.597, which is quite worse than the one (i.e., 0.797) reported in your paper.

Am I missing something? Thanks.

[entavelis]

Hello Hao,

This is indeed a typo in our part, I am going to create a note pointing this out in the README file.

Thank you!