mfcc_cnn_winnner.py

import keras
from keras.layers import Conv2D, Dense, Dropout, Flatten, MaxPooling2D, GlobalAveragePooling2D, BatchNormalization
from keras.models import Sequential
from keras.optimizers import Adam
from keras import regularizers

def model_cnn_reg(input_shape):
  model=Sequential()
  model.add(Conv2D(filters=64,kernel_size=(3,3), input_shape=input_shape, activation='relu'))
  model.add(MaxPooling2D(pool_size=(2,2)))
  model.add(Conv2D(filters=32,kernel_size=(2,2),activation='relu'))
  model.add(BatchNormalization())
  model.add(Flatten())
  model.add(Dense(256,activation='relu',kernel_regularizer=regularizers.l1_l2(l1=3e-4, l2=3e-3),bias_regularizer=regularizers.l2(3e-3),activity_regularizer=regularizers.l2(3e-4)))
  model.add(Dropout(0.5))
  model.add(Dense(128,activation='relu',kernel_regularizer=regularizers.l1_l2(l1=1e-3, l2=1e-2),bias_regularizer=regularizers.l2(1e-2),activity_regularizer=regularizers.l2(1e-3)))
  model.add(Dropout(0.3))
  model.add(Dense(1,activation='sigmoid'))
  optimiser = keras.optimizers.Adam(learning_rate=0.0001)
  model.compile(optimizer=optimiser,
                    loss='binary_crossentropy',
                    metrics=['accuracy'])

  return model