base_model = tf.keras.applications.MobileNetV2(input_shape=(224,224,3),
                                               alpha=1.0,
                                               include_top=False,
                                               weights="imagenet")
for layer in base_model.layers:
    layer.trainable = False
    
model_transfered_1=Sequential()

model_transfered_1.add(base_model)

# Flattening
model_transfered_1.add(Flatten())

# Fully connected layer 1st layer
model_transfered_1.add(Dense(32))
model_transfered_1.add(BatchNormalization())
model_transfered_1.add(Activation('relu'))
model_transfered_1.add(Dropout(0.4))

# Fully connected layer 2nd layer
model_transfered_1.add(Dense(32))
model_transfered_1.add(BatchNormalization())
model_transfered_1.add(Activation('relu'))
model_transfered_1.add(Dropout(0.4))

model_transfered_1.add(Dense(7, activation='softmax'))

model_transfered_1.compile(optimizer=Adam(lr=0.0005), 
                         loss='categorical_crossentropy', metrics=['categorical_accuracy'])
epochs = 10
steps_per_epoch = train_generator.n//train_generator.batch_size
validation_steps = validation_generator.n//validation_generator.batch_size

callbacks = [PlotLossesKerasTF(), checkpoint, reduce_lr]

history = model_transfered_1.fit(
    x=train_generator,
    steps_per_epoch=steps_per_epoch,
    epochs=epochs,
    validation_data = validation_generator,
    validation_steps = validation_steps,
    shuffle=True
)