Is it typo that the metric of IJB-C 1:N is TNIR(%)@FPIR=10^-2 in the paper?
pansanity666 opened this issue · comments
Hi, it's me again LOL, sorry for so many questions.
In the paper you said that the metric for IJB-C 1:N is TNIR(%)@ FPIR. But to the best of my knowledge, TNIR is equal to 1-FPIR.
Maybe you mean TPIR(%)@ FPIR=10^-2, which is also known as DET(Decision Error Trade-off) curve, as in this paper ?
By the way, for the open-set test protocol, should I calculate the metric for probe-G1 and probe-G2, separately, and then average them?
Hi, it's me again LOL, sorry for so many questions.
In the paper you said that the metric for IJB-C 1:N is TNIR(%)@ FPIR. But to the best of my knowledge, TNIR is equal to 1-FPIR.
Maybe you mean TPIR(%)@ FPIR=10^-2, which is also known as DET(Decision Error Trade-off) curve, as in this paper ?By the way, for the open-set test protocol, should I calculate the metric for probe-G1 and probe-G2, separately, and then average them?
I am deeply sorry for this typo! I will revise it right now (only the table caption is wrong, the experiment metric setting introduction part (Sec 3.1) is right).
It should be TPIR@FPIR as the standard testing protocol. Thanks for spotting the typo.
The openset TPIR@FPIR values are calcuated on each FPIR level for the two splits and averaged to the value presented in the paper.
An example to understand this
| FPIR | TPIR (G1) | TPIR (G2) | TPIR |
|---|---|---|---|
| 10-02 | value_a | value_b | (value_a + value_b) / 2 |
| 10-03 | value_c | value_d | (value_c + value_d) / 2 |
Thank you very much~