Question about other network performance.

[YuZheng9]

Thank you for your work! Your work has inspired me a lot.

I have a little question. In table.3 of your paper, you quantitative compare different methods.

But as far as I know, some methods have no official training results in some datasets(e.g. FFA-NET in NTIRE-2019).

I want to know how you get these results through training. Do you use the same training settings as your proposed Two branch network.

Thank you and look forward to your reply.

[liuh127]

Thank you for your work! Your work has inspired me a lot.

I have a little question. In table.3 of your paper, you quantitative compare different methods.

But as far as I know, some methods have no official training results in some datasets(e.g. FFA-NET in NTIRE-2019).

I want to know how you get these results through training. Do you use the same training settings as your proposed Two branch network.

Thank you and look forward to your reply.

The numbers we reported in our paper are either copied from the original paper or obtained by our reimplementation. For those methods that do not provide official testing results on several datasets, we carefully trained and tuned them to their best performance on each dataset. In other words, the hyper-parameters we used to train the compared methods might be different from that in our released training code. The released code should be only used to reproduce the performance of our proposed method.