search

Purdue-M2 / Fairness-Generalization

MIT License
21 stars 1 forks source link

Fairness Metrics and Testing Data Distribution #4

Closed qiqitao77 closed 3 months ago

qiqitao77 commented 3 months ago

Hi, thanks a lot for the code!

I'm trying to reproduce the experiment results in the paper.

I want to confirm whether the highlighted outputs below correspond to the F_FPR, F_OAE, F_DP, F_MEO metrics in the paper?

image

Also, I'm interested in whether you control the real/fake distribution and demographic distribution of testing data balanced while evaluating the detection performance and fairness, and how you select testing samples from raw datasets of Celeb-DF etc. (except for FF++).

Thank you and look forward to your kind reply.

Purdue-M2 commented 3 months ago

Thank you for your interest in our work. The metrics you highlighted are indeed the ones we used in our paper. We did not balance the testing set in terms of real/fake content and demographic groups.

qiqitao77 commented 3 months ago

Thank you for your interest in our work. The metrics you highlighted are indeed the ones we used in our paper. We did not balance the testing set in terms of real/fake content and demographic groups.

Thank you for your prompt reply! Looking forward to your release of data splits and checkpoint.😀