jis478 / anigan

Simple implementation of AniGAN

Hi there 👋

About Me

I’m an AI Research Engineer from South Korea. I truly enjoy challenging and improving myself by learning new skills on a daily basis in order to be a full stack AI engineer, covering from formulating the problem to conducting research to deploying models with a focus on data-centric mindset.

anigan's Issues

About the Encoder

After tracing the coding, I've noticed that the Encoder used in main.py is different from the original AniGAN paper.
Does it have a better performance than the original architecture?

I'm looking forward to your response.

Regards,
Lawrence

About recon_loss

Hi, thanks for your awesome work, it inspired me a lot!

I wonder that why you add recon_loss(recon_B, images_B) to the original recon loss, since the author merely add recon_loss(recon_A, images_A) to the recon loss?

I also noticed that you didn't follow the hyper parameters reported in the papar, does it affect the results?

Looking forward to your reply, thanks again!