search

zsyzzsoft / co-mod-gan

[ICLR 2021, Spotlight] Large Scale Image Completion via Co-Modulated Generative Adversarial Networks
Other
444 stars 67 forks source link

Use of Normalization Layers in Encoder #48

Closed hamzapehlivan closed 2 years ago

hamzapehlivan commented 2 years ago

Hi,

First of all, thanks for sharing this great work!

My question is about using normalization blocks (batch norm, instance norm, etc.) in the Encoder. I did not find any normalization layer when I investigated the source code. Is this really the case, or am I missing something?

If this is the case, is there any particular reason to omit it?

zsyzzsoft commented 2 years ago

The use of normalization layers in GAN is somewhat tricky. We just find it good enough without using normalization layers in the encoder, while I also believe that some proper use of normalization layers may further enhance the performance.

hamzapehlivan commented 2 years ago

Thanks for the kind response!