512/512 shape from MEL-spectogram by librosa.

andabi / music-source-separation

Deep neural networks for separating singing voice from music written in TensorFlow

795 stars 150 forks source link

512/512 shape from MEL-spectogram by librosa. #41

Open Tortoise17 opened 5 years ago

Tortoise17 commented 5 years ago

HI.

I have to ask a question. How do you define here for MIR-1K dataset that SECONDS = 8.192 / # To get 512,512 in melspectogram will give us this shape with this length of seconds.

Can you guide?

Greetings.