Кластеризация

def mnist_cnn_model():
   image_size = 28
   num_channels = 1  # 1 for grayscale images
   num_classes = 10  # Number of outputs
   model = Sequential()
   model.add(Conv2D(filters=32, kernel_size=(3,3), activation='relu', padding='same', input_shape=(image_size, image_size, num_channels)))
   model.add(MaxPooling2D(pool_size=(2, 2)))
   model.add(Conv2D(filters=64, kernel_size=(3, 3), activation='relu', padding='same'))
   model.add(MaxPooling2D(pool_size=(2, 2)))
   model.add(Conv2D(filters=64, kernel_size=(3, 3), activation='relu', padding='same'))
   model.add(MaxPooling2D(pool_size=(2, 2)))
   model.add(Flatten())
   # Densely connected layers
   model.add(Dense(128, activation='relu'))
   # Output layer
   model.add(Dense(num_classes, activation='softmax'))
   model.compile(optimizer=Adam(), loss='categorical_crossentropy', metrics=['accuracy'])
   return model
def mnist_cnn_train(model):
   (train_digits, train_labels), (test_digits, test_labels) = keras.datasets.mnist.load_data()
 
   # Get image size
   image_size = 28
   num_channels = 1  # 1 for grayscale images
 
   # re-shape and re-scale the images data
   train_data = np.reshape(train_digits, (train_digits.shape[0], image_size, image_size, num_channels))
   train_data = train_data.astype('float32') / 255.0
   # encode the labels - we have 10 output classes
   # 3 -> [0 0 0 1 0 0 0 0 0 0], 5 -> [0 0 0 0 0 1 0 0 0 0]
   num_classes = 10
   train_labels_cat = keras.utils.to_categorical(train_labels, num_classes)
 
   # re-shape and re-scale the images validation data
   val_data = np.reshape(test_digits, (test_digits.shape[0], image_size, image_size, num_channels))
   val_data = val_data.astype('float32') / 255.0
   # encode the labels - we have 10 output classes
   val_labels_cat = keras.utils.to_categorical(test_labels, num_classes)
 
   print("Training the network...")
   t_start = time.time()
 
   # Start training the network
   model.fit(train_data, train_labels_cat, epochs=8, batch_size=64, validation_data=(val_data, val_labels_cat))
 
   print("Done, dT:", time.time() - t_start)
 
   return model
model = mnist_cnn_model()
mnist_cnn_train(model)
model.save('cnn_digits_28x28.h5')
def cnn_digits_predict(model, image_file):
    image_size = 28
    img = keras.preprocessing.image.load_img(image_file, target_size=(image_size, image_size), color_mode='grayscale')
    img_arr = np.expand_dims(img, axis=0)
    img_arr = 1 - img_arr/255.0
    img_arr = img_arr.reshape((1, 28, 28, 1))
 
    result = model.predict_classes([img_arr])
    return result[0]
 
model = tf.keras.models.load_model('cnn_digits_28x28.h5')
print(cnn_digits_predict(model, 'digit_0.png'))