Hi.
In the equation (3) in the paper (L = L_{reconstruction}(E_c, E_p,D)+αL_{similarity}(E_c)+β(L_{adversarial}(E_p)+L_{adversarial}(C))), hyper parameter beta is applied to both L_{adversarial}(E_p) and L_{adversarial}(C). However, in the code, sd_weight seems to be used only within train function and not to be used within train_scene_discriminator function. Why is that?
Thanks.
Hi.
In the equation (3) in the paper (L = L_{reconstruction}(E_c, E_p,D)+αL_{similarity}(E_c)+β(L_{adversarial}(E_p)+L_{adversarial}(C))), hyper parameter beta is applied to both L_{adversarial}(E_p) and L_{adversarial}(C). However, in the code, sd_weight seems to be used only within train function and not to be used within train_scene_discriminator function. Why is that?
Thanks.