Binary Classifier for FFHQ

[backpropagator]

Hi, thanks for the brilliant work. I was wondering if the binary classifier used to evaluate FFHQ results is same as CLEVR dataset? I mean is same model architecture (for classifiers) employed for both the datasets?

I understand that open-sourcing the classifier might lead to potential privacy breach (#13), however if a different classifier architecture is used, would it be possible to atleast mention the classifier architecture or point to relevant resources to train the same?