keras imagedatagenerator

kerasで画像判別の学習を行っています。
ローカルに保存してある画像をtrain/test/vali=6:3:1に分け、
trainとtestの画像に対して、imagedatageneratorを適用したいです

python
1# coding:utf-8
2
3import keras
4from keras.utils import np_utils
5from keras.layers.convolutional import Conv2D, MaxPooling2D
6from keras.models import Sequential
7from keras.layers.core import Dense, Dropout, Activation, Flatten
8from keras.preprocessing.image import array_to_img, img_to_array, list_pictures, load_img
9import numpy as np
10import pandas as pd
11from sklearn.model_selection import train_test_split
12import matplotlib.pyplot as plt
13from sklearn.neighbors import KNeighborsClassifier
14
15#【学習データの読み込み】
16X = []
17Y = []#category_name
18
19V = []
20W = []
21
22
23# aの画像
24for picture in list_pictures('Der/a'):
25    img = img_to_array(load_img(picture, target_size=(64,64)))
26    X.append(img)
27
28    Y.append(0)
29
30for picture in list_pictures('Der/a'):
31    img = img_to_array(load_img(picture, target_size=(64,64)))
32    V.append(img)
33
34    W.append(0)
35
36# b
37for picture in list_pictures('Der/b'):
38    img = img_to_array(load_img(picture, target_size=(64,64)))
39    X.append(img)
40
41    Y.append(1)
42    
43for picture in list_pictures('Der/b'):
44    img = img_to_array(load_img(picture, target_size=(64,64)))
45    V.append(img)
46
47    W.append(1)
48    
49# c
50for picture in list_pictures('Der/c'):
51    img = img_to_array(load_img(picture, target_size=(64,64)))
52    X.append(img)
53
54    Y.append(2)
55    
56for picture in list_pictures('Der/c'):
57    img = img_to_array(load_img(picture, target_size=(64,64)))
58    V.append(img)
59
60    W.append(2)
61    
62# dの画像
63for picture in list_pictures('Der/d'):
64    img = img_to_array(load_img(picture, target_size=(64,64)))
65    X.append(img)
66
67    Y.append(3)
68    
69for picture in list_pictures('Der/d'):
70    img = img_to_array(load_img(picture, target_size=(64,64)))
71    V.append(img)
72
73    W.append(3)
74    
75# e
76for picture in list_pictures('Der/e'):
77    img = img_to_array(load_img(picture, target_size=(64,64)))
78    X.append(img)
79
80    Y.append(4)
81    
82for picture in list_pictures('Der/e'):
83    img = img_to_array(load_img(picture, target_size=(64,64)))
84    V.append(img)
85
86    W.append(4)
87    
88# fの画像
89for picture in list_pictures('Der/f'):
90    img = img_to_array(load_img(picture, target_size=(64,64)))
91    X.append(img)
92
93    Y.append(5)
94    
95for picture in list_pictures('Der/f'):
96    img = img_to_array(load_img(picture, target_size=(64,64)))
97    V.append(img)
98
99    W.append(5)
100    
101# gの画像
102for picture in list_pictures('Der/g'):
103    img = img_to_array(load_img(picture, target_size=(64,64)))
104    X.append(img)
105
106    Y.append(6)
107    
108for picture in list_pictures('Der/g'):
109    img = img_to_array(load_img(picture, target_size=(64,64)))
110    V.append(img)
111
112    W.append(6)
113    
114# hの画像
115for picture in list_pictures('Der/h'):
116    img = img_to_array(load_img(picture, target_size=(64,64)))
117    X.append(img)
118
119    Y.append(7)
120    
121for picture in list_pictures('Der/h'):
122    img = img_to_array(load_img(picture, target_size=(64,64)))
123    V.append(img)
124
125    W.append(7)
126    
127# load_imgをarrayに変換
128X = np.asarray(X)
129Y = np.asarray(Y)
130V = np.asarray(V)
131W = np.asarray(W)
132
133# 画素値を0から1の範囲に変換
134X = X.astype('float32')
135X = X / 255.0
136V = V.astype('float32')
137V = V / 255.0
138
139# クラスの形式を変換
140Y = np_utils.to_categorical(Y, 8)#(ベクトル変換したいラベル、配列数)
141W = np_utils.to_categorical(W, 8)
142
143
144
145X_train, X_vali, y_train, y_vali= train_test_split(X, Y, test_size=0.2)
146V_test, W_test = (V, W)
147
148# 【CNNを構築】
149model = Sequential()
150
151
152model.add(Conv2D(32, (3, 3), padding='same',
153                 input_shape=X_train.shape[1:]))
154
155model.add(Activation('relu'))
156model.add(Conv2D(32, (3, 3)))
157model.add(Activation('relu'))
158model.add(MaxPooling2D(pool_size=(2, 2)))
159
160model.add(Dropout(0.4))
161
162model.add(Conv2D(64, (3, 3), padding='same'))
163model.add(Activation('relu'))
164model.add(Conv2D(64, (3, 3)))
165model.add(Activation('relu'))
166model.add(MaxPooling2D(pool_size=(2, 2)))
167model.add(Dropout(0.4))
168
169model.add(Flatten())
170model.add(Dense(512))
171model.add(Activation('relu'))
172model.add(Dropout(0.4))
173model.add(Dense(8))#カテゴリー数
174model.add(Activation('softmax'))
175
176# 【コンパイル】
177model.compile(loss='categorical_crossentropy',
178              optimizer='adam',
179              metrics=['accuracy'])
180
181#【学習実行】
182history = model.fit(X_train, y_train, batch_size=64, epochs=100,
183                   validation_data = (X_vali, y_vali), verbose = 1)
184
185
186
187score = model.evaluate(V_test, W_test, verbose = 1)
188print("正解", score[1], "loss", score[0])