mishimario
And I want to know about preprocessing when evaluating. In the paper, there is face detection and alignment. What is this alignment? It's not only removing the background?
Thank you. I got it. How about the number of dim for FID? The dim is 2048?