About training model

[princess200816]

Hello! when I try to train LCNN+LFCC ,I meet some problems. The first problem : In FakeAVCeleb_v1.2 dataset, after I convert the MP4 format to MP3 format,it may go wrong as picture_1.But when I convert the MP4 format to WAV format,it run normal,just as picture_2. The second problem : when I run train_models.py, it show the warning:UserWarning: At least one mel filterbank has all zero values. The value for n_mels (128) may be set too high. Or, the value for n_freqs (257) may be set too low.I show it in picture_3. And I want to know how to preprocess the AAD .Thank you.

[piotrkawa]

Hello! Regarding torchaudio - unfortunately I did not encounter the problem you describe, but I guess it might be a problem with a backend for torchaudio i.e. ffmpeg and sox. I researched a bit and stumbled upon this issue. I’d start with changing their versions.

Regarding mel filterbanks - yes, some of them will be empty - these are the values which we conducted the experiments on, so ignore this warning.

[princess200816]

Thank you for your answer!

---Original--- From: "Piotr @.> Date: Thu, Apr 13, 2023 19:33 PM To: @.>; Cc: @.@.>; Subject: Re: [piotrkawa/attack-agnostic-dataset] About training model (Issue#2)

Hello! Regarding torchaudio - unfortunately I did not encounter the problem you describe, but I guess it might be a problem with a backend for torchaudio i.e. ffmpeg and sox. I researched a bit and stumbled upon this issue. I’d start with changing their versions.

Regarding mel filterbanks - yes, some of them will be empty - these are the values which we conducted the experiments on, so ignore this warning.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

[smuhib]

hello hope you are doing Great ! need your assistance @princess200816 if u have time i am student of UET taxila [email protected]