inp = Input(get_inception_conv_layers(conv_layer_name,input_size)[-1].output_shape[1:])
x = BatchNormalization(axis=1)(inp)

x = Convolution2D(nr_filters,3,3, activation='relu', border_mode='same')(x)
x = BatchNormalization(axis=1)(x)
x = Dropout(0.5)(x)
x = MaxPooling2D()(x)

x = Convolution2D(nr_filters,3,3, activation='relu', border_mode='same')(x)
x = BatchNormalization(axis=1)(x)
x = Dropout(0.5)(x)
x = MaxPooling2D()(x)

x = Convolution2D(nr_filters,3,3, activation='relu', border_mode='same')(x)
x = BatchNormalization(axis=1)(x)
x = Dropout(0.5)(x)
x = MaxPooling2D()(x)

x = Convolution2D(NR_CLASSES,3,3, border_mode='same')(x)
x = Dropout(0.5)(x)
x = GlobalAveragePooling2D()(x)
output = Activation('softmax')(x)

model = Model(inp, output)
model.compile(Adam(lr=0.001), loss='categorical_crossentropy', metrics=['accuracy'])