Reviews: Image-to-image translation for cross-domain disentanglement

In overall, I think this paper proposes a nice framework to learn the disentangled features. The authors addressed my questions in the rebuttal. And I suggest the author add the original BicycleGAN (with skip connections) results to the final version if the paper is accepted. Besides, I also suggest the author soften the claim about GRL loss or explain when the GRL loss can improve the performance and when not. My final rating is accept.