search

TomTomTommi / HiNet

Official PyTorch implementation of "HiNet: Deep Image Hiding by Invertible Network" (ICCV 2021)
170 stars 36 forks source link

Confusion about low frequency loss #8

Closed minzhang-1 closed 2 years ago

minzhang-1 commented 2 years ago

Hi, the low frequency loss in your code seems to be different from what you describe in the paper. You only select the first image in the batch, each with 12 channels to calculate loss. (If so, it will has no difference with the guide loss.) And the lambda is also not 10. Could you please explain a little bit?

TomTomTommi commented 2 years ago

Thanks for your careful observation, maybe it happened during polishing. I have fixed this problem with a "narrow" operation. But I haven't re-run the code, so if there exists any more bugs, please let me know. As for the hyper-parameter lambda, the setting depends on the situation. You could set as 1 in the pre-stage and just fine-tune this hyper-parameter to make the results of cover and secret images balance.