Hey, I'm running your code on a subset of the FakeAVCeleb dataset and I've tried Iam values of 0,1,5,10 and the output AUC I obtain by selecting the best threshold is around 79-80%. Quite far from the 94 in the paper. I wanted to ask, to fairly reproduce your experiments, which parameter value should I use while running the tests?
Hey, I'm running your code on a subset of the FakeAVCeleb dataset and I've tried Iam values of 0,1,5,10 and the output AUC I obtain by selecting the best threshold is around 79-80%. Quite far from the 94 in the paper. I wanted to ask, to fairly reproduce your experiments, which parameter value should I use while running the tests?
Thank you!