search

jnzhng / keras-unet-vocal-separation

17 stars 7 forks source link

pre-trained model #4

Closed Vichoko closed 5 years ago

Vichoko commented 5 years ago

i've just trained the model with the same dataset and hyper-parameters. You can download it from here: model And use it in the 3rd notebook.

28s 3ms/step - loss: 0.0032 - val_loss: 0.0032

Vichoko commented 5 years ago

trained on ccmixter + dsd100 model 75s 2ms/step - loss: 0.0032 - val_loss: 0.0067