【機械学習④】オートエンコーダによって出力される画像の質を上げてみた

# ライブラリのインポート
!pip install japanize-matplotlib

import matplotlib.pyplot as plt
import japanize_matplotlib 
import numpy as np
import pandas as pd
import keras
from keras import layers, losses
from keras.datasets import cifar10
from keras.models import Model

# ============================================================
# 日本語でラベル付け
(x_train, ｙ_train), (x_test, y_test) = cifar10.load_data() #

x_train = x_train.astype('float32') / 255.
x_test = x_test.astype('float32') / 255.

cifar10_labels = np.array(['飛行機','自動車','鳥','猫','鹿','犬','蛙','馬','船','トラック'])
# ============================================================
# モデルの設計と、設計したモデルのインスタンス作成
#
# ここで、２層、３層のモデル設計を行います。
########################################
# 以下の章で記載されるコードを実行して下さい。#
########################################
#
# ============================================================
#モデルの学習
autoencoder.compile(optimizer='adam', loss=losses.MeanSquaredError())

autoencoder.fit(x_train, x_train,
                epochs=10,
                shuffle=True,
                validation_data=(x_test, x_test))

# ============================================================
# 元画像とオートエンコーダの出力画像を表示
encoded_imgs = autoencoder.encoder(x_test).numpy()
decoded_imgs = autoencoder.decoder(encoded_imgs).numpy()

# 画像表示数
n = 10
# 画像番号[0〜(9999-n)]
m = 0
plt.figure(figsize=(20, 4))
for i in range(n):
  # 元画像の表示
  ax = plt.subplot(2, n, i + 1)
  plt.imshow(x_test[m+i])
  plt.title(cifar10_labels[y_test[m+i][0]] + "　元画像")
  ax.get_xaxis().set_visible(False)
  ax.get_yaxis().set_visible(False)

  # 再構築画像の表示
  ax = plt.subplot(2, n, i + 1 + n)
  plt.imshow(decoded_imgs[m+i])
  plt.title("再構築")
  ax.get_xaxis().set_visible(False)
  ax.get_yaxis().set_visible(False)
plt.show()

latent_dim = 512

class Autoencoder(Model):
  def __init__(self, latent_dim):
    super(Autoencoder, self).__init__()
    self.latent_dim = latent_dim   
    self.encoder = keras.Sequential([
      layers.Flatten(),
      layers.Dense(1024, activation='relu'), #この層を追加
      layers.Dense(latent_dim, activation='relu'),
    ])
    self.decoder = keras.Sequential([
      layers.Dense(512, activation='relu'), #この層を追加                               
      layers.Dense(32*32*3, activation='sigmoid'),
      layers.Reshape((32, 32, 3))
    ])

  def call(self, x):
    encoded = self.encoder(x)
    decoded = self.decoder(encoded)
    return decoded
  
autoencoder = Autoencoder(latent_dim)

latent_dim = 512

class Autoencoder(Model):
  def __init__(self, latent_dim):
    super(Autoencoder, self).__init__()
    self.latent_dim = latent_dim
    self.encoder = keras.Sequential([
      layers.Flatten(),
      layers.Dense(3072, activation='relu'), #この層を追加
      layers.Dense(1024, activation='relu'), #この層を追加
      layers.Dense(latent_dim, activation='relu'),
    ])
    self.decoder = keras.Sequential([
      layers.Dense(512, activation='relu'), #この層を追加 
      layers.Dense(1024, activation='relu'), #この層を追加                             
      layers.Dense(32*32*3, activation='sigmoid'),
      layers.Reshape((32, 32, 3))
    ])

  def call(self, x):
    encoded = self.encoder(x)
    decoded = self.decoder(encoded)
    return decoded
  
autoencoder = Autoencoder(latent_dim)

class Autoencoder_Conv(Model):
  def __init__(self):
    super(Autoencoder_Conv, self).__init__()
    self.encoder = keras.Sequential([
      layers.Input(shape=(32, 32, 3)), 
      layers.Conv2D(16, (3,3), activation='relu', padding='same', strides=2),
      layers.Conv2D(8, (3,3), activation='relu', padding='same', strides=2)])
    
    self.decoder = keras.Sequential([
      layers.Conv2DTranspose(8, kernel_size=3, strides=2, activation='relu', padding='same'),
      layers.Conv2DTranspose(16, kernel_size=3, strides=2, activation='relu', padding='same'),
      layers.Conv2D(3, kernel_size=(3,3), activation='sigmoid', padding='same')])
    
  def call(self, x):
    encoded = self.encoder(x)
    decoded = self.decoder(encoded)
    return decoded

autoencoder = Autoencoder_Conv()