ozora-ogino / lcnn Goto Github PK

The feature.py doesn't have a save_feature function.So may I have your ' save_feature function' code ?

I appreciate that your code can be used as a reference, but I don't know how to run them, could you please add more comments in the README file again? Thanks.

The protocol files include the PA files as csv. Does this script work for LA attack as well?

I am using your code to train data on LCNN

Now i am dealing with the problem of choosing access type PA or LA

When I choose the access type as LA, the error keep occurs.

I suppose the error is because there is no file about LA on the protocol file.

train_protocol.csv is filled only with PA files.

LA file does not need any protocol path or something>?

I beg your help.

Thank you

In the "model.fit " function I tried setting shuffle=True.
But still loss: 0.2522 - accuracy: 0.8989 -
val_loss: 0.0000e+00 - val_accuracy: 0.0000e+00 <-- (PS: Classification problem)
model.compile(
optimizer=Adam(learning_rate=lr),
loss="sparse_categorical_crossentropy",
metrics=["accuracy"]
)
model.fit(
x_train,
y_train,
epochs=epochs,
batch_size=batch_size,
validation_data=[x_val, y_val],
shuffle=True,
callbacks=[es, cp_cb])

The train and dev features get extracted successfully; While building the LCNN model, I get the following error:
ValueError: Input 0 of layer conv2d is incompatible with the layer: : expected min_ndim=4, found ndim=3.
I did not make any changes in the script, simply running the original script with same dataset (ASV spoof 2019)