search

jerryli27 / TwinGAN

Twin-GAN -- Unpaired Cross-Domain Image Translation with Weight-Sharing GANs
Apache License 2.0
719 stars 98 forks source link

How many encoders do you use? #31

Closed wyixiang closed 5 years ago

wyixiang commented 5 years ago

Your paper said that "your TwinGAN architecture uses two encoders, E1 and E2". But in your blog, I saw that "As mentioned in the Facebook paper, letting the human and anime portraits share the same network will help the network realize that......" So, how many encoders are there in the TwinGAN architecture?

jerryli27 commented 5 years ago

There are two encoders, one for each domain. They have the same structure and share some weights with each other, namely the conv kernel weights. They do not share the batch normalization parameters (it's similar to AdaIN).

wyixiang commented 5 years ago

OK, I get it. Thank you!