APCER/BPCER definitions in paper

[AIshutin]

Hi!

I've read appendix of the paper and experiments part of OULU-NPU, which is referenced in paper, however, in my understanding definitions of APCER in CelebA-Spoof and OULU-NPU are different (and that seems strange). Can you please clarify what exactly do you mean by APCER and BPCER?

In OULU-NPU:

i. e. this maximum error across different attacks.

In CelebA-Spoof:

which seems to be simply FAR and FRR.

Thanks.

[ZhangYuanhan-AI]

Hi: Yes, it is different from the definition of Oulu-NPU, because we follow the calculation fashion of APCER and BPCER from this repo, this paper is accepted by CVPR.

[AIshutin]

Thanks for help.

[AIshutin]

You wrote in the paper:

(ACER is calculated at EER)

On the other hand, in CDCN: result ACER is minimum of ACER(thr_i).

Just to finally confirm, should I follow the code in CDCN`s implementation, not the CelebA-Spoof paper?

[AIshutin]

Sorry, I still don't understand( . Can you please explain, what's the difference between FAR and APCER, when calculated at EER threshold?

[AIshutin]

Then, BPCER = FRR_i (which FRR_i == FAR_i) -> BPCER = APCER, no?

(From the paper I suppose not, but I can't find the mistake.)

[ZhangYuanhan-AI]

Thanks for your remind, we realize we do totally follow the calculation fashion from this repo.

[AIshutin]

Again, big thanks for help.