def build_audio_D(self):
# load pretrain weights
audio_input = Input(batch_shape=(self.batch_size, self.audio_len),
name='audio_input')
spectro = logMelSpectrogram(sample_rate=self.audio_sr)(audio_input)
# Block 1
x = Conv2D(64, (3, 3), strides=(1, 1), padding='same',
name='conv1')(spectro)
x = LeakyReLU(alpha=0.2)(x)
x = MaxPooling2D((2, 2), strides=(2, 2), padding='same',
name='pool1')(x)
# Block 2
x = Conv2D(128, (3, 3), strides=(1, 1), padding='same',
name='conv2')(x)
x = LeakyReLU(alpha=0.2)(x)
x = MaxPooling2D((2, 2), strides=(2, 2), padding='same',
name='pool2')(x)
# Block 3
x = Conv2D(256, (3, 3),
strides=(1, 1),
padding='same',
name='conv3/conv3_1')(x)
x = LeakyReLU(alpha=0.2)(x)
x = Conv2D(256, (3, 3),
strides=(1, 1),
padding='same',
name='conv3/conv3_2')(x)
x = LeakyReLU(alpha=0.2)(x)
x = MaxPooling2D((2, 2), strides=(2, 2), padding='same',
name='pool3')(x)
# Block 4
x = Conv2D(512, (3, 3),
strides=(1, 1),
padding='same',
name='conv4/conv4_1')(x)
x = LeakyReLU(alpha=0.2)(x)
x = Conv2D(512, (3, 3),
strides=(1, 1),
padding='same',
name='conv4/conv4_2')(x)
x = LeakyReLU(alpha=0.2)(x)
x = MaxPooling2D((2, 2), strides=(2, 2), padding='same',
name='pool4')(x)
fea = Flatten(name='flatten_')(x)
#x = Dense(4096, name='fc1')(x)
#x = LeakyReLU(alpha=0.2)(x)
#x = Dense(self.audio_emb_dim, name='embeddings')(x)
#fea = LeakyReLU(alpha=0.2)(x)
valid = Dense(1, activation='sigmoid', name='predictions')(fea)
labels = Dense(self.classes + 1, activation='softmax')(fea)
audio_D = Model(inputs=audio_input, outputs=[valid, labels])
return audio_D
def build_audio_C(self):
img_input = Input(batch_shape=(self.batch_size, self.img_rows, self.img_cols, self.channels),
name='img_input')
x = Encoding_layer(name='vOICe')(img_input)
spectro = logMelSpectrogram(name='logSpectrogram')(x)
# Block 1
x = Conv2D(64, (3, 3), strides=(1, 1), activation='relu', padding='same', name='conv1')(spectro)
#x = LeakyReLU(alpha=0.2)(x)
x = MaxPooling2D((2, 2), strides=(2, 2), padding='same', name='pool1')(x)
# Block 2
x = Conv2D(128, (3, 3), strides=(1, 1), activation='relu', padding='same', name='conv2')(x)
#x = LeakyReLU(alpha=0.2)(x)
x = MaxPooling2D((2, 2), strides=(2, 2), padding='same', name='pool2')(x)
# Block 3
x = Conv2D(256, (3, 3), strides=(1, 1), activation='relu', padding='same', name='conv3/conv3_1')(x)
#x = LeakyReLU(alpha=0.2)(x)
x = Conv2D(256, (3, 3), strides=(1, 1), activation='relu', padding='same', name='conv3/conv3_2')(x)
#x = LeakyReLU(alpha=0.2)(x)
x = MaxPooling2D((2, 2), strides=(2, 2), padding='same', name='pool3')(x)
# Block 4
x = Conv2D(512, (3, 3), strides=(1, 1), activation='relu', padding='same', name='conv4/conv4_1')(x)
#x = LeakyReLU(alpha=0.2)(x)
x = Conv2D(512, (3, 3), strides=(1, 1), activation='relu', padding='same', name='conv4/conv4_2')(x)
#x = LeakyReLU(alpha=0.2)(x)
x = MaxPooling2D((2, 2), strides=(2, 2), padding='same', name='pool4')(x)
x = Flatten(name='flatten_')(x)
x = Dense(4096, activation='relu', name='fc1')(x)
embeddings = Dense(self.audio_emb_dim, activation='relu', name='embeddings')(x)
predicts = Dense(self.classes, activation='softmax', name='prediction')(embeddings)
audio_model = Model(inputs=img_input, outputs=predicts)
return audio_model