Difference between CycleGAN

[udithhaputhanthri]

Hi,

Thank you for the really great work. Going through the literature, we have found a similar work CycleGAN (https://arxiv.org/pdf/1703.10593.pdf), which uses similar losses as in MAGAN. [CycleGAN is cited in the MAGAN paper as well]

eg: reconstruction loss in MAGAN = cycle consistency in CycleGAN correspondence loss in MAGAN = identity loss in CycleGAN

As I understood, the difference in MAGAN paper compared to CycleGAN is the discriminator loss where there is a term including x_121, x_212 whereas in CycleGAN, there is no such term in discriminator loss. Even though there is this loss difference in the paper, I was not able to find this loss term in this repository.

Please correct me if there is any misunderstanding of the concept.