def build_img_G(self):
audio_embeddings = Input(batch_shape=(self.batch_size, self.audio_emb_dim))
audio_noise = Input(batch_shape=(self.batch_size, self.noise_dim))
#audio_label_embedding = ()(Embedding(self.classes, self.audio_emb_dim)(audio_labels))
audio_input = Concatenate()([audio_embeddings, audio_noise])
x = Dense(256)(audio_input)
x = LeakyReLU(alpha=0.2)(x)
x = BatchNormalization(momentum=0.8)(x)
x = Dense(512)(x)
x = LeakyReLU(alpha=0.2)(x)
x = BatchNormalization(momentum=0.8)(x)
x = Dense(1024)(x)
x = LeakyReLU(alpha=0.2)(x)
x = BatchNormalization(momentum=0.8)(x)
x = Dense(np.prod(self.img_shape), activation='sigmoid')(x)
img = Reshape(self.img_shape, name='generated_img')(x)
encoded_audio = Encoding_layer()(img)
img_G = Model(inputs=[audio_embeddings, audio_noise], outputs=encoded_audio)
return img_G
def build_img_G_conv_full(self):
audio_embeddings = Input(batch_shape=(self.batch_size, self.audio_emb_dim))
audio_noise = Input(batch_shape=(self.batch_size, self.noise_dim))
audio_input = Concatenate()([audio_embeddings, audio_noise])
x = Dense(128*7*7, activation='relu')(audio_input)
x = Reshape((7,7,128))(x)
x = UpSampling2D()(x)
x = Conv2D(128, kernel_size=5, padding="same")(x)
x = BatchNormalization(momentum=0.8)(x)
x = Activation("relu")(x)
x = UpSampling2D()(x)
x = Conv2D(64, kernel_size=5, padding="same")(x)
x = BatchNormalization(momentum=0.8)(x)
x = Activation("relu")(x)
#x = UpSampling2D()(x)
#x = Conv2D(32, kernel_size=5, padding="same")(x)
#x = BatchNormalization(momentum=0.8)(x)
#x = Activation("relu")(x)
x = Conv2D(self.channels, kernel_size=3, padding="same")(x)
img = Activation(activation='sigmoid', name='generated_img')(x)
encoded_audio = Encoding_layer()(img)
img_G = Model(inputs=[audio_embeddings, audio_noise], outputs=encoded_audio)
return img_G
def build_audio_encoder(self):
img_input = Input(batch_shape=(self.batch_size, self.img_rows, self.img_cols, self.channels),
name='img_input')
output = Encoding_layer(name='vOICe')(img_input)
audio_encoder = Model(inputs=img_input, outputs=output)
return audio_encoder
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
