Request code for quantitative comparison

[budui]

Hello, thank you very much for your excellent work! Could you please provide the code and classifier model for calculating mean-AD, max-AD in your paper? These are necessary for a fair comparison with StyleSpace in the next papers.

[betterze]

Dear Wrong Wang,

Thank you for your interest in our work. We will upload the codes about AD in the next few days.

Best Wishes,

Zongze

[xrenaa]

Dear Wrong Wang,

Thank you for your interest in our work. We will upload the codes about AD in the next few days.

Best Wishes,

Zongze

Hi, thanks for your excellent work. I am wondering if you will opensource the way to calculate DCI on latent spaces of StyleGAN2 in terms of FFHQ dataset.

[betterze]

Dear @xrenaa ，

Thank you for your interest in our work. We used the implementation of DCI in google disentanglement_lib.

We will realse the full codes about DCI and AD during weekend.

Best Wishes,

Zongze

[betterze]

@budui I am sorry, I have a deadline soon, it takes all of my energy. I may not be able to upload it in time.

The classifers are from [StyleGAN metric}(https://drive.google.com/drive/folders/1MvYdWCBuMfnoYGptRH-AgKLbPTsIQLhl)

[di-mi-ta]

Hi @betterze,

I wonder that how long you run the evaluation code for computing DCI? I have tried to re-implement DCI metric but I have a time bottleneck at using face attribute classifier (31) to prepare data for DCI (500k images). It took a very long time even though I used GPU.

Thanks!

[enisimsar]

Hi all,

Does anyone have the code for the DCI metric for this paper?

Thanks

[betterze]

@enisimsar @di-mi-ta @xrenaa The DCI metric is online now. Sorry for the delay.

[enisimsar]

Thanks a lot :)