scale generator network output to [-1, 1]

[pyguan88]

This generator network does not scale the output to [-1, 1] with tanh activation function. Does anyone know why?

[yjia757-2]

I have that question too. it's output is directly used to compare with the hr_image, which is [-1, 1]. My guess is the the biases setting of the generator network make the output in rage [-1, 1]? Did you figure it out?