AI 事始め10：Get Started with TensorFlow :Fashion MNIST その２

Get Started with TensorFlowの冒頭のコードが気になります。1. Basic classificationのコードとは、model、model.compile、model.evaluateなどの記述が異なります。

先に作成した、Fashion MNISTのJyupiter Notebookを下記のように変更してみました。

<tensorflow.python.keras.callbacks.History at 0x252ec4e5390>のようなメッセージは出力されません。Trainの時間が長くなっています。精度も少し良くなっているように見えます。動作の仕組みに関して知識がゼロの状態では、なんですが、実験はできそうな感じですね。

 In [1]:

# TensorFlow and tf.keras
import tensorflow as tf
from tensorflow import keras  # 最新のtensorflowにはkerasが同梱されているのかな？

# Helper libraries
import numpy as np
import matplotlib.pyplot as plt

print(tf.__version__)

 

1.11.0

In [2]:

fashion_mnist = keras.datasets.fashion_mnist

(train_images, train_labels), (test_images, test_labels) = fashion_mnist.load_data()

In [3]:

class_names = ['T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat', 
               'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle boot']

In [4]:

train_images.shape

Out[4]:

(60000, 28, 28)

In [5]:

len(train_labels)

Out[5]:

60000

In [6]:

train_labels

Out[6]:

array([9, 0, 0, ..., 3, 0, 5], dtype=uint8)

In [7]:

test_images.shape

Out[7]:

(10000, 28, 28)

In [8]:

len(test_labels)

Out[8]:

10000

In [9]:

plt.figure()
plt.imshow(train_images[0])
plt.colorbar()
plt.grid(False)

 

In [10]:

train_images = train_images / 255.0

test_images = test_images / 255.0

In [11]:

plt.figure(figsize=(10,10))
for i in range(25):
    plt.subplot(5,5,i+1)
    plt.xticks([])
    plt.yticks([])
    plt.grid(False)
    plt.imshow(train_images[i], cmap=plt.cm.binary)
    plt.xlabel(class_names[train_labels[i]])

 

In [31]:

# MODELを変えてみました。
model = tf.keras.models.Sequential([
  tf.keras.layers.Flatten(),
  tf.keras.layers.Dense(512, activation=tf.nn.relu),
  tf.keras.layers.Dropout(0.2),
  tf.keras.layers.Dense(10, activation=tf.nn.softmax)
])
# ここの記述を変えてみました。
model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])
model.fit(train_images, train_labels, epochs=5)
model.evaluate(test_images, test_labels)

 

Epoch 1/5
60000/60000 [==============================] - 14s 232us/step - loss: 0.4727 - acc: 0.8321
Epoch 2/5
60000/60000 [==============================] - 14s 225us/step - loss: 0.3605 - acc: 0.8686
Epoch 3/5
60000/60000 [==============================] - 14s 225us/step - loss: 0.3234 - acc: 0.8799
Epoch 4/5
60000/60000 [==============================] - 14s 227us/step - loss: 0.3010 - acc: 0.8897
Epoch 5/5
60000/60000 [==============================] - 14s 227us/step - loss: 0.2808 - acc: 0.8959
10000/10000 [==============================] - 0s 47us/step

Out[31]:

[0.3366610364437103, 0.8792]

In [16]:

predictions = model.predict(test_images)

In [17]:

predictions[0]

Out[17]:

array([2.2103798e-07, 1.3197257e-07, 1.4327797e-07, 1.7504580e-09,
       3.8520842e-08, 2.0662786e-02, 7.2375616e-07, 4.4677880e-02,
       2.1749284e-07, 9.3465793e-01], dtype=float32)

In [18]:

np.argmax(predictions[0])

Out[18]:

9

In [19]:

test_labels[0]

Out[19]:

9

In [20]:

def plot_image(i, predictions_array, true_label, img):
  predictions_array, true_label, img = predictions_array[i], true_label[i], img[i]
  plt.grid(False)
  plt.xticks([])
  plt.yticks([])
  
  plt.imshow(img, cmap=plt.cm.binary)

  predicted_label = np.argmax(predictions_array)
  if predicted_label == true_label:
    color = 'blue'
  else:
    color = 'red'
  
  plt.xlabel("{} {:2.0f}% ({})".format(class_names[predicted_label],
                                100*np.max(predictions_array),
                                class_names[true_label]),
                                color=color)

def plot_value_array(i, predictions_array, true_label):
  predictions_array, true_label = predictions_array[i], true_label[i]
  plt.grid(False)
  plt.xticks([])
  plt.yticks([])
  thisplot = plt.bar(range(10), predictions_array, color="#777777")
  plt.ylim([0, 1]) 
  predicted_label = np.argmax(predictions_array)
 
  thisplot[predicted_label].set_color('red')
  thisplot[true_label].set_color('blue')

In [21]:

i = 0
plt.figure(figsize=(6,3))
plt.subplot(1,2,1)
plot_image(i, predictions, test_labels, test_images)
plt.subplot(1,2,2)
plot_value_array(i, predictions,  test_labels)

 

In [22]:

i = 12
plt.figure(figsize=(6,3))
plt.subplot(1,2,1)
plot_image(i, predictions, test_labels, test_images)
plt.subplot(1,2,2)
plot_value_array(i, predictions,  test_labels)

 

In [23]:

# Plot the first X test images, their predicted label, and the true label
# Color correct predictions in blue, incorrect predictions in red
num_rows = 5
num_cols = 3
num_images = num_rows*num_cols
plt.figure(figsize=(2*2*num_cols, 2*num_rows))
for i in range(num_images):
  plt.subplot(num_rows, 2*num_cols, 2*i+1)
  plot_image(i, predictions, test_labels, test_images)
  plt.subplot(num_rows, 2*num_cols, 2*i+2)
  plot_value_array(i, predictions, test_labels)

 

In [24]:

# Grab an image from the test dataset
img = test_images[0]

print(img.shape)

 

(28, 28)

In [25]:

# Add the image to a batch where it's the only member.
img = (np.expand_dims(img,0))

print(img.shape)

 

(1, 28, 28)

In [26]:

predictions_single = model.predict(img)

print(predictions_single)

 

2.2103858e-07 1.3197256e-07 1.4327823e-07 1.7504711e-09 3.8520980e-08
  2.0662796e-02 7.2375815e-07 4.4677928e-02 2.1749364e-07 9.3465781e-01

In [27]:

plot_value_array(0, predictions_single, test_labels)
_ = plt.xticks(range(10), class_names, rotation=45)

 

In [28]:

np.argmax(predictions_single[0])

Out[28]:

9

In [29]:

#@title MIT License
#
# Copyright (c) 2017 François Chollet
#
# Permission is hereby granted, free of charge, to any person obtaining a
# copy of this software and associated documentation files (the "Software"),
# to deal in the Software without restriction, including without limitation
# the rights to use, copy, modify, merge, publish, distribute, sublicense,
# and/or sell copies of the Software, and to permit persons to whom the
# Software is furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE.