Questions about DRIT

[helenmao]

Questions about DRIT

[Yitian-Li]

For question2, here are some image of my test results.
input:

output 0-4:

[HelenMao]

Hi sorry for the late reply.

1. For your question about DRIT, I suggest reading their paper DRIT first to get more details.
   The fake_B_encoded is the generated image using encoded latent vector z from A domain.
   The fake_B_random is the generated image using random noise vector z from A domain.
   The fake_AA_encoded is the self-reconstructed image using encoded latent vector z from domain A.
2. We use 1999 epochs when training cat2dog in DRIT and MSGAN.
3. Please see the links in Quantitative Evaluation Metrics at the bottom of our Github project. And also see the appendix carefully in our paper.