rec_B works well, fake_B works poorly, identity 0.5, lambda_A, and lambda_B 10

[dupc2018]

rec_B works well, fake_B works poorly, identity 0.5, lambda_A, and lambda_B 10

[junyanz]

Could you share more details?

[dupc2018]

Using a paired dataset, training the recurrent network was found to be very close rec_B to the real_B, and the fake_B was much different from both rec_B and real_B