質問編集履歴
4
誤字
test
CHANGED
File without changes
|
test
CHANGED
@@ -4,7 +4,9 @@
|
|
4
4
|
|
5
5
|
何とかコードはできましたが、
|
6
6
|
|
7
|
-
学習してからのテストデータの精度を調べる方法がわか
|
7
|
+
学習してからのテストデータの精度を調べる方法がわからず、
|
8
|
+
|
9
|
+
エラーが発生します。
|
8
10
|
|
9
11
|
初心者ゆえの疑問です。
|
10
12
|
|
3
コード修正
test
CHANGED
File without changes
|
test
CHANGED
@@ -162,6 +162,36 @@
|
|
162
162
|
|
163
163
|
validation_split=0.2)
|
164
164
|
|
165
|
+
# 精度の評価
|
166
|
+
|
167
|
+
scores = model.evaluate(data_set, verbose=1)
|
168
|
+
|
169
|
+
print('Test loss:', scores[0])
|
170
|
+
|
171
|
+
print('Test accuracy:', scores[1])
|
172
|
+
|
173
|
+
|
174
|
+
|
175
|
+
# データの可視化(テストデータの先頭の10枚)
|
176
|
+
|
177
|
+
for i in range(10):
|
178
|
+
|
179
|
+
plt.subplot(2, 5, i+1)
|
180
|
+
|
181
|
+
plt.imshow(data_set[i].reshape((28,28)), 'gray')
|
182
|
+
|
183
|
+
plt.suptitle("テストデータの先頭の10枚",fontsize=20)
|
184
|
+
|
185
|
+
plt.show()
|
186
|
+
|
187
|
+
|
188
|
+
|
189
|
+
# 予測(テストデータの先頭の10枚)
|
190
|
+
|
191
|
+
pred = np.argmax(model.predict(data_set[0:10]), axis=1)
|
192
|
+
|
193
|
+
print(pred)
|
194
|
+
|
165
195
|
|
166
196
|
|
167
197
|
model.summary()
|
@@ -174,60 +204,50 @@
|
|
174
204
|
|
175
205
|
_________________________________________________________________
|
176
206
|
|
177
|
-
|
178
|
-
|
179
|
-
|
180
|
-
|
181
|
-
|
182
|
-
|
183
|
-
|
184
|
-
|
185
|
-
|
186
|
-
|
187
|
-
_
|
188
|
-
|
189
|
-
|
190
|
-
|
191
|
-
|
192
|
-
|
193
|
-
|
194
|
-
|
195
|
-
__
|
196
|
-
|
197
|
-
|
198
|
-
|
199
|
-
|
200
|
-
|
201
|
-
|
202
|
-
|
203
|
-
|
204
|
-
|
205
|
-
|
206
|
-
|
207
|
-
|
208
|
-
|
209
|
-
de
|
210
|
-
|
211
|
-
____
|
212
|
-
|
213
|
-
|
214
|
-
|
215
|
-
_
|
216
|
-
|
217
|
-
|
218
|
-
|
219
|
-
|
220
|
-
|
221
|
-
|
222
|
-
|
223
|
-
|
224
|
-
|
225
|
-
activation_4 (Activation) (None, 2219) 0
|
226
|
-
|
227
|
-
=================================================================
|
228
|
-
|
229
|
-
Total params: 4,040,747
|
230
|
-
|
231
|
-
Trainable params: 4,040,747
|
232
|
-
|
233
|
-
Non-trainable params: 0
|
207
|
+
---------------------------------------------------------------------------
|
208
|
+
|
209
|
+
IndexError Traceback (most recent call last)
|
210
|
+
|
211
|
+
<ipython-input-2-d7df5a824556> in <module>()
|
212
|
+
|
213
|
+
110
|
214
|
+
|
215
|
+
111 # 精度の評価
|
216
|
+
|
217
|
+
--> 112 scores = model.evaluate(data_set, verbose=1)
|
218
|
+
|
219
|
+
113 print('Test loss:', scores[0])
|
220
|
+
|
221
|
+
114 print('Test accuracy:', scores[1])
|
222
|
+
|
223
|
+
|
224
|
+
|
225
|
+
~\Anaconda3\lib\site-packages\keras\engine\training.py in evaluate(self, x, y, batch_size, verbose, sample_weight, steps)
|
226
|
+
|
227
|
+
1111 batch_size=batch_size,
|
228
|
+
|
229
|
+
1112 verbose=verbose,
|
230
|
+
|
231
|
+
-> 1113 steps=steps)
|
232
|
+
|
233
|
+
1114
|
234
|
+
|
235
|
+
1115 def predict(self, x,
|
236
|
+
|
237
|
+
|
238
|
+
|
239
|
+
~\Anaconda3\lib\site-packages\keras\engine\training_arrays.py in test_loop(model, f, ins, batch_size, verbose, steps)
|
240
|
+
|
241
|
+
353 indices_for_conversion_to_dense = []
|
242
|
+
|
243
|
+
354 for i in range(len(feed)):
|
244
|
+
|
245
|
+
--> 355 if issparse(ins[i]) and not K.is_sparse(feed[i]):
|
246
|
+
|
247
|
+
356 indices_for_conversion_to_dense.append(i)
|
248
|
+
|
249
|
+
357
|
250
|
+
|
251
|
+
|
252
|
+
|
253
|
+
IndexError: list index out of range
|
2
コード修正
test
CHANGED
File without changes
|
test
CHANGED
@@ -64,19 +64,7 @@
|
|
64
64
|
|
65
65
|
my_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
|
66
66
|
|
67
|
-
|
67
|
+
|
68
|
-
|
69
|
-
#幅、高さ 80*28
|
70
|
-
|
71
|
-
"""
|
72
|
-
|
73
|
-
#幅、高さ 80*28
|
74
|
-
|
75
|
-
my_img= cv2.resize(gray, (80, 28))
|
76
|
-
|
77
|
-
"""
|
78
|
-
|
79
|
-
# my_img= cv2.resize(gray, (80, 28))
|
80
68
|
|
81
69
|
|
82
70
|
|
@@ -86,66 +74,22 @@
|
|
86
74
|
|
87
75
|
my_img = my_img.flatten().tolist()
|
88
76
|
|
89
|
-
# my_img = my_img.tolist()
|
90
|
-
|
91
77
|
data_set.append(my_img)
|
92
78
|
|
93
79
|
labels.append(item.split('.')[0])
|
94
80
|
|
95
|
-
|
81
|
+
|
96
|
-
|
97
|
-
|
82
|
+
|
98
|
-
|
99
|
-
|
83
|
+
|
100
|
-
|
101
|
-
# labels.append(item.split('.')[0])
|
102
84
|
|
103
85
|
|
104
86
|
|
105
|
-
# フィルタを定義
|
106
|
-
|
107
|
-
# filt = np.array([[0, 1, 0],
|
108
|
-
|
109
|
-
# [1, 0, 1],
|
110
|
-
|
111
|
-
# [0, 1, 0]], np.uint8)
|
112
|
-
|
113
|
-
|
114
|
-
|
115
|
-
|
116
|
-
|
117
|
-
# 膨張
|
118
|
-
|
119
|
-
# my_img =cv2.dilate(img, filt)
|
120
|
-
|
121
|
-
# data_set.append(my_img.flatten().tolist())
|
122
|
-
|
123
|
-
# labels.append(item.split('.')[0])
|
124
|
-
|
125
|
-
# 収縮
|
126
|
-
|
127
|
-
# my_img =cv2.erode(img, filt)
|
128
|
-
|
129
|
-
# data_set.append(my_img.flatten().tolist())
|
130
|
-
|
131
|
-
# labels.append(item.split('.')[0])
|
132
|
-
|
133
|
-
|
134
|
-
|
135
|
-
# cv2.imwrite(str(file) + "_" + str(count) + ".jpg", img)
|
136
|
-
|
137
|
-
|
138
|
-
|
139
|
-
|
140
|
-
|
141
87
|
# Numpyへ戻して
|
142
88
|
|
143
89
|
|
144
90
|
|
145
91
|
data_set = np.array(data_set)/255
|
146
92
|
|
147
|
-
# print(data_set)
|
148
|
-
|
149
93
|
import pandas as pd
|
150
94
|
|
151
95
|
|
@@ -156,24 +100,10 @@
|
|
156
100
|
|
157
101
|
# 正規化
|
158
102
|
|
159
|
-
#data_set = data_set/255
|
160
|
-
|
161
103
|
data_set = data_set.reshape([-1, 28, 80, 1])
|
162
104
|
|
163
105
|
|
164
106
|
|
165
|
-
#if not os.path.exists("data_set"):
|
166
|
-
|
167
|
-
# os.mkdir("data_set")
|
168
|
-
|
169
|
-
|
170
|
-
|
171
|
-
#for num, im in enumerate(data_set):
|
172
|
-
|
173
|
-
# まず保存先のディレクトリ"data_set/"を指定、番号を付けて保存
|
174
|
-
|
175
|
-
# cv2.imwrite("data_set/" + str(num) + ".jpg" ,im)
|
176
|
-
|
177
107
|
|
178
108
|
|
179
109
|
# モデルの定義
|
1
書式改善
test
CHANGED
@@ -1 +1 @@
|
|
1
|
-
python
|
1
|
+
python MNISTを利用した文字認証
|
test
CHANGED
@@ -1,14 +1,226 @@
|
|
1
|
-
pythonを学習し初めて
|
1
|
+
pythonを学習し初めて機械学習を利用した文字認証を行おうとしています。
|
2
2
|
|
3
3
|
文字は4文字の英字で、幅80 高さ28のjpgファイルです。
|
4
4
|
|
5
|
+
何とかコードはできましたが、
|
6
|
+
|
5
|
-
|
7
|
+
学習してからのテストデータの精度を調べる方法がわかりません。
|
8
|
+
|
6
|
-
|
9
|
+
初心者ゆえの疑問です。
|
10
|
+
|
11
|
+
|
12
|
+
|
7
|
-
|
13
|
+
ご教示いただければと思います。
|
14
|
+
|
15
|
+
|
16
|
+
|
8
|
-
|
17
|
+
```ここに言語を入力
|
18
|
+
|
9
|
-
|
19
|
+
from keras.datasets import mnist
|
20
|
+
|
10
|
-
|
21
|
+
from keras.layers import Dense, Dropout, Flatten, Activation
|
22
|
+
|
23
|
+
from keras.layers import Conv2D, MaxPooling2D
|
24
|
+
|
25
|
+
from keras.models import Sequential, load_model
|
26
|
+
|
27
|
+
from keras.utils.np_utils import to_categorical
|
28
|
+
|
29
|
+
from keras.utils.vis_utils import plot_model
|
30
|
+
|
31
|
+
import numpy as np
|
32
|
+
|
33
|
+
import matplotlib.pyplot as plt
|
34
|
+
|
35
|
+
#import cv2
|
36
|
+
|
37
|
+
import os
|
38
|
+
|
39
|
+
%matplotlib inline
|
40
|
+
|
41
|
+
|
42
|
+
|
43
|
+
path = "./data/"
|
44
|
+
|
45
|
+
files = os.listdir(path)
|
46
|
+
|
47
|
+
|
48
|
+
|
49
|
+
data_set = []
|
50
|
+
|
51
|
+
labels=[]
|
52
|
+
|
53
|
+
|
54
|
+
|
55
|
+
for item in files:
|
56
|
+
|
57
|
+
img = cv2.imread(os.path.join(path, item))
|
58
|
+
|
59
|
+
#追加
|
60
|
+
|
61
|
+
img= cv2.resize(img, (80, 28))
|
62
|
+
|
63
|
+
# グレースケール変換
|
64
|
+
|
65
|
+
my_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
|
66
|
+
|
67
|
+
# my_img = cv2.cvtColor(my_img, cv2.COLOR_GRAY2BGR)
|
68
|
+
|
11
|
-
|
69
|
+
#幅、高さ 80*28
|
70
|
+
|
71
|
+
"""
|
72
|
+
|
73
|
+
#幅、高さ 80*28
|
74
|
+
|
75
|
+
my_img= cv2.resize(gray, (80, 28))
|
76
|
+
|
77
|
+
"""
|
78
|
+
|
79
|
+
# my_img= cv2.resize(gray, (80, 28))
|
80
|
+
|
81
|
+
|
82
|
+
|
83
|
+
# リスト型へ変換してappendで追加
|
84
|
+
|
85
|
+
#print(my_img)
|
86
|
+
|
87
|
+
my_img = my_img.flatten().tolist()
|
88
|
+
|
89
|
+
# my_img = my_img.tolist()
|
90
|
+
|
91
|
+
data_set.append(my_img)
|
92
|
+
|
93
|
+
labels.append(item.split('.')[0])
|
94
|
+
|
95
|
+
#ぼかし
|
96
|
+
|
97
|
+
# my_img = cv2.GaussianBlur(img, (3, 3), 0)
|
98
|
+
|
99
|
+
# data_set.append(my_img.flatten().tolist())
|
100
|
+
|
101
|
+
# labels.append(item.split('.')[0])
|
102
|
+
|
103
|
+
|
104
|
+
|
105
|
+
# フィルタを定義
|
106
|
+
|
107
|
+
# filt = np.array([[0, 1, 0],
|
108
|
+
|
109
|
+
# [1, 0, 1],
|
110
|
+
|
111
|
+
# [0, 1, 0]], np.uint8)
|
112
|
+
|
113
|
+
|
114
|
+
|
115
|
+
|
116
|
+
|
117
|
+
# 膨張
|
118
|
+
|
119
|
+
# my_img =cv2.dilate(img, filt)
|
120
|
+
|
121
|
+
# data_set.append(my_img.flatten().tolist())
|
122
|
+
|
123
|
+
# labels.append(item.split('.')[0])
|
124
|
+
|
125
|
+
# 収縮
|
126
|
+
|
127
|
+
# my_img =cv2.erode(img, filt)
|
128
|
+
|
129
|
+
# data_set.append(my_img.flatten().tolist())
|
130
|
+
|
131
|
+
# labels.append(item.split('.')[0])
|
132
|
+
|
133
|
+
|
134
|
+
|
135
|
+
# cv2.imwrite(str(file) + "_" + str(count) + ".jpg", img)
|
136
|
+
|
137
|
+
|
138
|
+
|
139
|
+
|
140
|
+
|
141
|
+
# Numpyへ戻して
|
142
|
+
|
143
|
+
|
144
|
+
|
145
|
+
data_set = np.array(data_set)/255
|
146
|
+
|
147
|
+
# print(data_set)
|
148
|
+
|
149
|
+
import pandas as pd
|
150
|
+
|
151
|
+
|
152
|
+
|
153
|
+
labels_set=pd.get_dummies(labels).values
|
154
|
+
|
155
|
+
|
156
|
+
|
157
|
+
# 正規化
|
158
|
+
|
159
|
+
#data_set = data_set/255
|
160
|
+
|
161
|
+
data_set = data_set.reshape([-1, 28, 80, 1])
|
162
|
+
|
163
|
+
|
164
|
+
|
165
|
+
#if not os.path.exists("data_set"):
|
166
|
+
|
167
|
+
# os.mkdir("data_set")
|
168
|
+
|
169
|
+
|
170
|
+
|
171
|
+
#for num, im in enumerate(data_set):
|
172
|
+
|
173
|
+
# まず保存先のディレクトリ"data_set/"を指定、番号を付けて保存
|
174
|
+
|
175
|
+
# cv2.imwrite("data_set/" + str(num) + ".jpg" ,im)
|
176
|
+
|
177
|
+
|
178
|
+
|
179
|
+
# モデルの定義
|
180
|
+
|
181
|
+
model = Sequential()
|
182
|
+
|
183
|
+
model.add(Conv2D(filters=32, kernel_size=(3, 3),input_shape=(28,80,1)))
|
184
|
+
|
185
|
+
model.add(Activation('relu'))
|
186
|
+
|
187
|
+
model.add(Conv2D(filters=64, kernel_size=(3, 3)))
|
188
|
+
|
189
|
+
model.add(Activation('relu'))
|
190
|
+
|
191
|
+
model.add(MaxPooling2D(pool_size=(2, 2)))
|
192
|
+
|
193
|
+
model.add(Dropout(0.25))
|
194
|
+
|
195
|
+
model.add(Flatten())
|
196
|
+
|
197
|
+
model.add(Dense(128))
|
198
|
+
|
199
|
+
model.add(Activation('relu'))
|
200
|
+
|
201
|
+
model.add(Dropout(0.5))
|
202
|
+
|
203
|
+
|
204
|
+
|
205
|
+
#かっこの中は重複なしのファイル数とする。 model.add(Dense(10))
|
206
|
+
|
207
|
+
model.add(Dense(2219))
|
208
|
+
|
209
|
+
|
210
|
+
|
211
|
+
model.add(Activation('softmax'))
|
212
|
+
|
213
|
+
|
214
|
+
|
215
|
+
|
216
|
+
|
217
|
+
model.compile(loss='categorical_crossentropy',
|
218
|
+
|
219
|
+
optimizer='adadelta',
|
220
|
+
|
221
|
+
metrics=['accuracy'])
|
222
|
+
|
223
|
+
#変更する余地あり epochs=100
|
12
224
|
|
13
225
|
model.fit(data_set,labels_set,
|
14
226
|
|
@@ -20,258 +232,72 @@
|
|
20
232
|
|
21
233
|
validation_split=0.2)
|
22
234
|
|
23
|
-
|
24
|
-
|
235
|
+
|
236
|
+
|
25
|
-
|
237
|
+
model.summary()
|
26
|
-
|
27
|
-
|
238
|
+
|
28
|
-
|
29
|
-
|
239
|
+
|
30
|
-
|
31
|
-
コード
|
32
240
|
|
33
241
|
```
|
34
242
|
|
35
|
-
|
36
|
-
|
37
|
-
|
38
|
-
|
39
|
-
|
40
|
-
|
41
|
-
|
42
|
-
|
43
|
-
|
44
|
-
|
45
|
-
|
46
|
-
|
47
|
-
|
48
|
-
|
49
|
-
|
50
|
-
|
51
|
-
|
52
|
-
|
53
|
-
|
54
|
-
|
55
|
-
|
56
|
-
|
57
|
-
|
58
|
-
|
59
|
-
|
60
|
-
|
61
|
-
|
62
|
-
|
63
|
-
|
64
|
-
|
65
|
-
|
66
|
-
|
67
|
-
f
|
68
|
-
|
69
|
-
|
70
|
-
|
71
|
-
|
72
|
-
|
73
|
-
|
74
|
-
|
75
|
-
i
|
76
|
-
|
77
|
-
|
78
|
-
|
79
|
-
|
80
|
-
|
81
|
-
|
82
|
-
|
83
|
-
|
84
|
-
|
85
|
-
|
86
|
-
|
87
|
-
|
88
|
-
|
89
|
-
|
90
|
-
|
91
|
-
|
92
|
-
|
93
|
-
|
94
|
-
|
95
|
-
|
96
|
-
|
97
|
-
#追加
|
98
|
-
|
99
|
-
img= cv2.resize(img, (80, 28))
|
100
|
-
|
101
|
-
# グレースケール変換
|
102
|
-
|
103
|
-
my_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
|
104
|
-
|
105
|
-
my_img = cv2.cvtColor(my_img, cv2.COLOR_GRAY2BGR)
|
106
|
-
|
107
|
-
#幅、高さ 80*28
|
108
|
-
|
109
|
-
"""
|
110
|
-
|
111
|
-
#幅、高さ 80*28
|
112
|
-
|
113
|
-
my_img= cv2.resize(gray, (80, 28))
|
114
|
-
|
115
|
-
"""
|
116
|
-
|
117
|
-
# my_img= cv2.resize(gray, (80, 28))
|
118
|
-
|
119
|
-
|
120
|
-
|
121
|
-
# リスト型へ変換してappendで追加
|
122
|
-
|
123
|
-
#print(my_img)
|
124
|
-
|
125
|
-
my_img = my_img.tolist()
|
126
|
-
|
127
|
-
data_set.append(my_img)
|
128
|
-
|
129
|
-
labels.append(item.split('.')[0])
|
130
|
-
|
131
|
-
#ぼかし
|
132
|
-
|
133
|
-
my_img = cv2.GaussianBlur(img, (5, 5), 0)
|
134
|
-
|
135
|
-
data_set.append(my_img.tolist())
|
136
|
-
|
137
|
-
labels.append(item.split('.')[0])
|
138
|
-
|
139
|
-
|
140
|
-
|
141
|
-
# フィルタを定義
|
142
|
-
|
143
|
-
filt = np.array([[0, 1, 0],
|
144
|
-
|
145
|
-
[1, 0, 1],
|
146
|
-
|
147
|
-
[0, 1, 0]], np.uint8)
|
148
|
-
|
149
|
-
|
150
|
-
|
151
|
-
|
152
|
-
|
153
|
-
# 膨張
|
154
|
-
|
155
|
-
my_img =cv2.dilate(img, filt)
|
156
|
-
|
157
|
-
data_set.append(my_img.tolist())
|
158
|
-
|
159
|
-
labels.append(item.split('.')[0])
|
160
|
-
|
161
|
-
# 収縮
|
162
|
-
|
163
|
-
my_img =cv2.erode(img, filt)
|
164
|
-
|
165
|
-
data_set.append(my_img.tolist())
|
166
|
-
|
167
|
-
labels.append(item.split('.')[0])
|
168
|
-
|
169
|
-
|
170
|
-
|
171
|
-
# cv2.imwrite(str(file) + "_" + str(count) + ".jpg", img)
|
172
|
-
|
173
|
-
|
174
|
-
|
175
|
-
|
176
|
-
|
177
|
-
# Numpyへ戻して
|
178
|
-
|
179
|
-
|
180
|
-
|
181
|
-
data_set = np.array(data_set)
|
182
|
-
|
183
|
-
# print(data_set)
|
184
|
-
|
185
|
-
import pandas as pd
|
186
|
-
|
187
|
-
|
188
|
-
|
189
|
-
labels_set=pd.get_dummies(labels).values
|
190
|
-
|
191
|
-
|
192
|
-
|
193
|
-
# 正規化
|
194
|
-
|
195
|
-
data_set = data_set/255
|
196
|
-
|
197
|
-
data_set = data_set.reshape([-1, 80, 28, 1])
|
198
|
-
|
199
|
-
|
200
|
-
|
201
|
-
if not os.path.exists("data_set"):
|
202
|
-
|
203
|
-
os.mkdir("data_set")
|
204
|
-
|
205
|
-
|
206
|
-
|
207
|
-
for num, im in enumerate(data_set):
|
208
|
-
|
209
|
-
# まず保存先のディレクトリ"data_set/"を指定、番号を付けて保存
|
210
|
-
|
211
|
-
cv2.imwrite("data_set/" + str(num) + ".jpg" ,im)
|
212
|
-
|
213
|
-
|
214
|
-
|
215
|
-
# モデルの定義
|
216
|
-
|
217
|
-
model = Sequential()
|
218
|
-
|
219
|
-
model.add(Conv2D(filters=32, kernel_size=(3, 3),input_shape=(80,28,1)))
|
220
|
-
|
221
|
-
model.add(Activation('relu'))
|
222
|
-
|
223
|
-
model.add(Conv2D(filters=64, kernel_size=(3, 3)))
|
224
|
-
|
225
|
-
model.add(Activation('relu'))
|
226
|
-
|
227
|
-
model.add(MaxPooling2D(pool_size=(2, 2)))
|
228
|
-
|
229
|
-
model.add(Dropout(0.25))
|
230
|
-
|
231
|
-
model.add(Flatten())
|
232
|
-
|
233
|
-
model.add(Dense(128))
|
234
|
-
|
235
|
-
model.add(Activation('relu'))
|
236
|
-
|
237
|
-
model.add(Dropout(0.5))
|
238
|
-
|
239
|
-
|
240
|
-
|
241
|
-
#かっこの中は重複なしのファイル数とする。 model.add(Dense(10))
|
242
|
-
|
243
|
-
model.add(Dense(2219))
|
244
|
-
|
245
|
-
|
246
|
-
|
247
|
-
model.add(Activation('softmax'))
|
248
|
-
|
249
|
-
|
250
|
-
|
251
|
-
|
252
|
-
|
253
|
-
model.compile(loss='categorical_crossentropy',
|
254
|
-
|
255
|
-
optimizer='adadelta',
|
256
|
-
|
257
|
-
metrics=['accuracy'])
|
258
|
-
|
259
|
-
#変更する余地あり epochs=100
|
260
|
-
|
261
|
-
model.fit(data_set,labels_set,
|
262
|
-
|
263
|
-
batch_size=128,
|
264
|
-
|
265
|
-
epochs=100,
|
266
|
-
|
267
|
-
verbose=1,
|
268
|
-
|
269
|
-
validation_split=0.2)
|
270
|
-
|
271
|
-
|
272
|
-
|
273
|
-
model.summary()
|
274
|
-
|
275
|
-
|
276
|
-
|
277
|
-
```
|
243
|
+
実行結果:
|
244
|
+
|
245
|
+
_________________________________________________________________
|
246
|
+
|
247
|
+
Layer (type) Output Shape Param #
|
248
|
+
|
249
|
+
=================================================================
|
250
|
+
|
251
|
+
conv2d_1 (Conv2D) (None, 26, 78, 32) 320
|
252
|
+
|
253
|
+
_________________________________________________________________
|
254
|
+
|
255
|
+
activation_1 (Activation) (None, 26, 78, 32) 0
|
256
|
+
|
257
|
+
_________________________________________________________________
|
258
|
+
|
259
|
+
conv2d_2 (Conv2D) (None, 24, 76, 64) 18496
|
260
|
+
|
261
|
+
_________________________________________________________________
|
262
|
+
|
263
|
+
activation_2 (Activation) (None, 24, 76, 64) 0
|
264
|
+
|
265
|
+
_________________________________________________________________
|
266
|
+
|
267
|
+
max_pooling2d_1 (MaxPooling2 (None, 12, 38, 64) 0
|
268
|
+
|
269
|
+
_________________________________________________________________
|
270
|
+
|
271
|
+
dropout_1 (Dropout) (None, 12, 38, 64) 0
|
272
|
+
|
273
|
+
_________________________________________________________________
|
274
|
+
|
275
|
+
flatten_1 (Flatten) (None, 29184) 0
|
276
|
+
|
277
|
+
_________________________________________________________________
|
278
|
+
|
279
|
+
dense_1 (Dense) (None, 128) 3735680
|
280
|
+
|
281
|
+
_________________________________________________________________
|
282
|
+
|
283
|
+
activation_3 (Activation) (None, 128) 0
|
284
|
+
|
285
|
+
_________________________________________________________________
|
286
|
+
|
287
|
+
dropout_2 (Dropout) (None, 128) 0
|
288
|
+
|
289
|
+
_________________________________________________________________
|
290
|
+
|
291
|
+
dense_2 (Dense) (None, 2219) 286251
|
292
|
+
|
293
|
+
_________________________________________________________________
|
294
|
+
|
295
|
+
activation_4 (Activation) (None, 2219) 0
|
296
|
+
|
297
|
+
=================================================================
|
298
|
+
|
299
|
+
Total params: 4,040,747
|
300
|
+
|
301
|
+
Trainable params: 4,040,747
|
302
|
+
|
303
|
+
Non-trainable params: 0
|