model.predict_generator(test_generator)

data = img.astype(float)/255
model.predict(data)

def Get_generator_data(dir,img_width, img_height, batch_size):
    datagen = ImageDataGenerator(rescale=1. / 255)
    train_generator = datagen.flow_from_directory(
    dir,
    target_size=(img_width, img_height),
    batch_size=batch_size,
    class_mode='categorical',
    shuffle=False)
    return train_generator

def create_model(outNeron, size):


    # Загружаем предварительно обученную нейронную сеть VGG16
    vgg16_net = VGG16(weights='imagenet', include_top=False,
                      input_shape=(size, size, 3))

    # "Замораживаем" веса предварительно обученной нейронной сети VGG16
    vgg16_net.trainable = False

    # Создаем составную нейронную сеть на основе VGG16
    # Создаем последовательную модель Keras
    model = Sequential()
    # Добавляем в модель сеть VGG16 вместо слоя
    model.add(vgg16_net)
    # Добавляем в модель новый классификатор
    model.add(Flatten())
    model.add(Dense(256))
    model.add(Activation('relu'))
    model.add(Dropout(0.5))
    model.add(Dense(outNeron))
    model.add(Activation('softmax'))
    return model

def Train_Model (model, train_generator, nb_train_samples, batch_size, val_generator, nb_validation_samples):
        сheckpoint = ModelCheckpoint('save/mnist-dense.hdf5',
            monitor='val_acc',
            save_best_only=True)
    # Компилируем составную нейронную сеть
        model.compile(loss=losses.categorical_crossentropy,
            optimizer=Adam(lr=1e-5),
            metrics=['accuracy'])
    # Обучаем модель с использованием генераторов
        model.fit_generator(
            train_generator,
            steps_per_epoch=nb_train_samples // batch_size,
            epochs=35,
            validation_data=val_generator,
            validation_steps=nb_validation_samples // batch_size,
            callbacks=[сheckpoint])

Layer (type)                 Output Shape              Param #
=================================================================
vgg16 (Model)                (None, 1, 1, 512)         14714688
_________________________________________________________________
flatten (Flatten)            (None, 512)               0
_________________________________________________________________
dense (Dense)                (None, 256)               131328
_________________________________________________________________
activation (Activation)      (None, 256)               0
_________________________________________________________________
dropout (Dropout)            (None, 256)               0
_________________________________________________________________
dense_1 (Dense)              (None, 4)                 1028
_________________________________________________________________
activation_1 (Activation)    (None, 4)                 0
=================================================================
Total params: 14,847,044
Trainable params: 132,356
Non-trainable params: 14,714,688
_________________________________________________________________
None