lixucuhk / asv-anti-spoofing-with-res2net Goto Github PK

Hello dear Dr. Xu Li

My name is Shahryar Ghayour an I am working on your paper
''ASV-anti-spoofing-with-Res2Net''.

I have a question if you can help me please.

I have installed kaldi and made changes to your ''extract_feats.sh '' file and added my directory to the dataset. the question is when I put the file in terminal on linux there is an error in step 1.

when it reaches to this line : [ -d data/spec/${name} ] cp -r data/${name} data/spec/${name}

an error comes which is [: missing `]'

how can I make it work?

can you please tell me?

Thank you so much

Did you try on the out of domain data for replay attack detection? Because most of these algorithms are not working good on the real-world data which i saw. Do you have any opinion about it?

Thanks, Best Regards
Mesut

Hi!
After evaluation, I just got some .txt and .eer files. And I want to calculate the EER, can u give me some suggestions?
Thanks for your reply!

您好，我看到path.sh里面有个KALDI_ROOT,我不太清楚这个地方需要改的是什么路径，我是否需要先安装kaldi链接

你好 @lixucuhk
我对这个工程很感兴趣而且浏览一些源码，如果我只用CQT 特征，是不是就不需要安装kaldi，是不是只有LFCC特征需要用到kaldi？谢谢！

Hi Dr. Xu Li

I could fortunately run the training command and successfully run "start.sh" command.
now that I got the output as txt file I am running your "err" command to get the output and here are the results:

scoring/evaluate_tDCF_asvspoof19.py:36: DeprecationWarning: np.float is a deprecated alias for the builtin float. To silence this warning, use float by itself. Doing this will not modify any behavior and is safe. If you specifically wanted the numpy scalar type, use np.float64 here.
Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations
asv_scores = asv_data[:, 2].astype(np.float)
scoring/evaluate_tDCF_asvspoof19.py:43: DeprecationWarning: np.float is a deprecated alias for the builtin float. To silence this warning, use float by itself. Doing this will not modify any behavior and is safe. If you specifically wanted the numpy scalar type, use np.float64 here.
Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations
cm_scores = cm_data[:, 3].astype(np.float)
0.024604674888228765
0.024581005586592177
0.23934754703985472
t-DCF evaluation from [Nbona=7355, Nspoof=63882] trials

t-DCF MODEL
Ptar = 0.94050 (Prior probability of target user)
Pnon = 0.00950 (Prior probability of nontarget user)
Pspoof = 0.05000 (Prior probability of spoofing attack)
Cfa_asv = 10.00000 (Cost of ASV falsely accepting a nontarget)
Cmiss_asv = 1.00000 (Cost of ASV falsely rejecting target speaker)
Cfa_cm = 10.00000 (Cost of CM falsely passing a spoof to ASV system)
Cmiss_cm = 1.00000 (Cost of CM falsely blocking target utterance which never reaches ASV)

Implied normalized t-DCF function (depends on t-DCF parameters and ASV errors), s=CM threshold)
tDCF_norm(s) = 2.40595 x Pmiss_cm(s) + Pfa_cm(s)

ASV SYSTEM
EER = 2.45778 % (Equal error rate (target vs. nontarget discrimination)
Pfa = 2.46047 % (False acceptance rate of nontargets)
Pmiss = 2.45810 % (False rejection rate of targets)
1-Pmiss,spoof = 76.06525 % (Spoof false acceptance rate)

CM SYSTEM
EER = 13.10529 % (Equal error rate for countermeasure)

TANDEM

I am a bit confused at this stage. I appreciate if you can help me what should I do now.
This is the command to compute both system EER and t-DCF:
python scoring/evaluate_tDCF_asvspoof19.py scoring/la_asv_scores/ASVspoof2019.LA.asv.eval.gi.trl.scores.txt cm_scores/seres2net50_26w_8s-epoch11-eval_scores.txt

Am I getting the right results? what can I do to get visual output? and use it in a paper?
Thanks

I am not able to run extract_feats.sh file , it give me error. , i already installed all dependencies in my conda environment and kaldi to run shell script..Error is given below:

$ ./extract_feats.sh --stage 2
Stage 2: extract CQT feats.
Traceback (most recent call last):
File "D:\asv_res2net\feats_extraction\compute_CQT.py", line 4, in
import numpy as np
ModuleNotFoundError: No module named 'numpy'

Why it is giving me numpy error? Please help me out