npz data files creation

[oriohayon]

Hi,
Thanks for the nice repository, I have a question regarding train test phase in some of the notebooks:
You are loading a npc data files (e.g., '/media/hdd1/khaled/npz_files/final_version/test_data.npz') that constructs the train and test dataset, however in the repository I don't find them and I didn't understand how you created them.

Any help finding out how you packed the data files will be great, thanks!

[hechmik]

Hi! These files are produced by asvtorch and saved in .npz format for better convenience: in this way I don't have to use every time the loaders inside asvtorch and I have my dataframe ready to be used.
Therefore what you will have to do is to copy the copy inside that repo, split all observations in train and test according to the train and test files I uploaded and you should be good to go :)

[oriohayon]

Hey,
Thanks for the answer!

So if I'm only interested in generating the dataset (including the audio files) used in the paper without the asvtorch environment what will be the easiest way to do so?

[hechmik]

If you only want the raw recordings then you should get in touch with the original VoxCeleb <https://www.robots.ox.ac.uk/~vgg/data/voxceleb/> team in order to get all the recordings, as some YT files are no longer available due to copyright issues, account deletion etc. However, if you want also the processed recordings (MFCC etc) then it is best to use asvtorch as it wraps all the Kaldi operations Il giorno gio 23 giu 2022 alle ore 14:12 oriohayon ***@***.***> ha scritto:

…

Hey, Thanks for the answer! So if I'm only interested in generating the dataset (including the audio files) used in the paper without the asvtorch environment what will be the easiest way to do so? — Reply to this email directly, view it on GitHub <#3 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AH3GRTKO7PCFGUEMRNAQJ6DVQRIBPANCNFSM5ZELQUAQ> . You are receiving this because you commented.Message ID: ***@***.***>

[oriohayon]

Thanks again for your answer.
My problem with the asvtorch environment is that it requires to many dependencies and other installations (such as kaldi tools) and I don't have a stand alone environment to do so but I am working with Colab.

Is it possible to run the asvtorch with Colab notebooks? I have troubles in Colab with many things such as Conda and I don't even get to the point where I can run anything.

[hechmik]

Oh ok, now I understand the context! It is quite tricky, as asvtorch uses Kaldi under the hood for the spectrograms, mfcc etc etc. In theory you can compute them with librosa, however the resulting mfcc from librosa will quite likely be different than the kaldi based on what I read in the past, even though I do not have any direct experience on it. My suggestion is to run asvtorch on a local machine, then upload the output on your Drive, so that you have your data ready to be used from Colab. Especially because the asvtorch pipeline take several hours to run, therefore you might risk sudden interruption on Colab. Il giorno ven 24 giu 2022 alle ore 12:57 oriohayon ***@***.***> ha scritto:

…

Thanks again for your answer. My problem with the asvtorch environment is that it requires to many dependencies and other installations (such as kaldi tools) and I don't have a stand alone environment to do so but I am working with Colab. Is it possible to run the asvtorch with Colab notebooks? I have troubles in Colab with many things such as Conda and I don't even get to the point where I can run anything. — Reply to this email directly, view it on GitHub <#3 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AH3GRTKNLFQSZTZFY25JO43VQWIBTANCNFSM5ZELQUAQ> . You are receiving this because you commented.Message ID: ***@***.***>