labels of spoofing scores in the implemention of "One-class Learning Towards Synthetic Voice Spoofing Detection" #46
-
|
In the function of t-dcf that ASV2019 published, the labels for bonafide or spoofing must be 1 and 0 accordingly no? |
Beta Was this translation helpful? Give feedback.
Answered by
yzyouzhang
Dec 30, 2023
Replies: 1 comment
-
|
Referring to https://github.com/yzyouzhang/AIR-ASVspoof/blob/master/eval_metrics.py#L157, the target is bonafide speech, and the non-target is spoofing attacks. I use bonafide 0 and spoofing 1 in my code, but reversing is also fine, as long as you are consistent. |
Beta Was this translation helpful? Give feedback.
0 replies
Answer selected by
yzyouzhang
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Referring to https://github.com/yzyouzhang/AIR-ASVspoof/blob/master/eval_metrics.py#L157, the target is bonafide speech, and the non-target is spoofing attacks. I use bonafide 0 and spoofing 1 in my code, but reversing is also fine, as long as you are consistent.