search

JuanFMontesinos / Acappella-YNet

Official implementation of A cappella: Audio-visual Singing VoiceSeparation, from BMVC21
https://ipcv.github.io/Acappella/
14 stars 4 forks source link

Metrics #5

Closed EmreOzkose closed 2 years ago

EmreOzkose commented 2 years ago

Hi,

I run run.py and it gives me some metrics like sdr/ds, etc.. Are these metrics equal to ones without ds? I search ds term in code, but couldn't find what it is exactly. For example I got sdr/ds=13.4 for my training, can I say sdr=13.4 ?

JuanFMontesinos commented 2 years ago

Hi, You can find the implementation of sdr at https://github.com/JuanFMontesinos/Acappella-YNet/blob/main/VnBSS/trainer.py

The -ds too

It's just for tensorboard to group the parameters in a certain way. So yes, that's sdr for the validation set.

In theory if you are using traces for the validation set, that metric should run on exactly the same pair samples.

EmreOzkose commented 2 years ago

Thank you :)