mahaichuan
mahaichuan
Thank you very much for your answer! This helps me a lot. By the way, do you have the **PIRM-test HR images**? And is it convenient for you to share?...
It seems to be (-1,1), which is what the original tensorflow implementation recommends.
Use $n_bits=8$.
The pytorch implementation has been released. You can find its link in the ReadMe file.
I use Tensorflow 1.4.
Sorry for the late reply！ There is a link in README under the model folder.