search

sigsep / sigsep-mus-io

Tools to convert sigsep mus dataset from STEMS <-> WAV
MIT License
8 stars 4 forks source link

0it [00:00, ?it/s] - nothing more #1

Closed Raprodent closed 5 years ago

Raprodent commented 5 years ago

Hey, I'm new to this, but have done everything as long as the guides go - decoded the STEM databases etc. When I try to run the command python IRM.py --eval_dir ./Evaluation_Dir --audio_dir ./Audio_Dir it simply yields the result: 0it [00:00, ?it/s] Then it returns to "base".

Any help highly appreciated, thanks! :)

faroit commented 5 years ago

Hi, sorry for the late reply.

I guess your questions is related to the code in the oracle repository. Here, it is important to enable the decoded wav flag of the musdb parser. See here. I hope that solves the problem.

Raprodent commented 5 years ago

Hey mate, no worries. The oracle program is just for the stem.mp4's and not for other audio files? Merry Xmas!

On Mon, 24 Dec 2018, 12:25 Fabian-Robert Stöter <notifications@github.com wrote:

Hi, sorry for the late reply.

I guess your questions is related to the code in the oracle repository https://github.com/sigsep/sigsep-mus-oracle. Here, it is important to enable the decoded wav flag of the musdb parser. See here https://github.com/sigsep/sigsep-mus-db#use-a-decoded-version. I hope that solves the problem.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/sigsep/sigsep-mus-io/issues/1#issuecomment-449722011, or mute the thread https://github.com/notifications/unsubscribe-auth/AgJMpAeLpEeeMBbLuYMrit36rg5F--IAks5u8LmTgaJpZM4ZaaxN .

faroit commented 5 years ago

Exactly, so you just need to add the is_wav flag

Raprodent commented 5 years ago

Do you have any insight in which vocal separator that currently work best? :)

On Mon, Dec 24, 2018 at 12:51 PM Fabian-Robert Stöter < notifications@github.com> wrote:

Exactly, so you just need to add the is_wav flag

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/sigsep/sigsep-mus-io/issues/1#issuecomment-449724690, or mute the thread https://github.com/notifications/unsubscribe-auth/AgJMpE8fs54wcpxKMaJqAIxd7YQfXQD9ks5u8L_SgaJpZM4ZaaxN .

faroit commented 5 years ago

Have a look at the SiSEC2018 results the best performing system from Sony is not open source, but you find a reference to reimplement it yourself here. We are currently developing an open source separation tool that will be released very soon in the next weeks. You can find more information about it here