teratail
回答率
85.35%
teratail header banner
teratail header banner
質問するログイン新規登録
トップに関する質問google colaboratoryでのエラー

編集履歴

質問編集履歴

6

全体的な修正

2021/11/24 08:07

投稿

退会済みユーザー
test CHANGED
File without changes
test CHANGED
@@ -32,11 +32,11 @@
32
32
 
33
33
  %matplotlib inline
34
34
 
35
-
36
-
35
+
36
+
37
- ----------------------------
37
+ #----------------------------
38
-
39
-
38
+
39
+
40
40
 
41
41
  !wget https://github.com/GANs-in-Action/gans-in-action/raw/Chapter-10/chapter-10/DSC_0897.jpg
42
42
 
@@ -44,9 +44,9 @@
44
44
 
45
45
  !wget https://raw.githubusercontent.com/GANs-in-Action/gans-in-action/master/chapter-10/initialization_vals_for_noise.csv
46
46
 
47
- ------------------------------------------------------
47
+ #------------------------------------------------------
48
-
49
-
48
+
49
+
50
50
 
51
51
  def load_image(img_path: str):
52
52
 
@@ -62,11 +62,11 @@
62
62
 
63
63
  image = load_image('DSC_0897.jpg')
64
64
 
65
-
66
-
65
+
66
+
67
- -------------------------------------------------------
67
+ #-------------------------------------------------------
68
-
69
-
68
+
69
+
70
70
 
71
71
  keras.backend.set_learning_phase(0)
72
72
 
@@ -122,15 +122,15 @@
122
122
 
123
123
  pp.pprint(decode_predictions(new_preds, top=20)[0]))
124
124
 
125
-
126
-
125
+
126
+
127
- ------------------------------------------------------------
127
+ # ------------------------------------------------------------
128
-
129
-
128
+
129
+
130
130
 
131
131
  plt.figure(figsize=(12,12))
132
132
 
133
-
133
+
134
134
 
135
135
  plt.subplot(1, 3, 1)
136
136
 
@@ -166,17 +166,17 @@
166
166
 
167
167
  plt.show()
168
168
 
169
-
170
-
169
+
170
+
171
- --------------------------------------------
171
+ # --------------------------------------------
172
-
173
-
172
+
173
+
174
174
 
175
175
  max_vals = pd.read_csv('initialization_vals_for_noise.csv')
176
176
 
177
-
178
-
177
+
178
+
179
- -----------------------------------------
179
+ #-----------------------------------------
180
180
 
181
181
 
182
182
 
@@ -260,11 +260,11 @@
260
260
 
261
261
  plt.show()
262
262
 
263
-
264
-
263
+
264
+
265
- ----------------------------------------------------
265
+ ##----------------------------------------------------
266
-
267
-
266
+
267
+
268
268
 
269
269
  fig = plt.figure(figsize=(20,20))
270
270
 
@@ -320,7 +320,7 @@
320
320
 
321
321
 
322
322
 
323
-
323
+
324
324
 
325
325
  for i in range(1,101):
326
326
 
@@ -340,9 +340,9 @@
340
340
 
341
341
  plt.show()
342
342
 
343
-
344
-
343
+
344
+
345
- --------------------------------------------------
345
+ #--------------------------------------------------
346
346
 
347
347
 
348
348
 
@@ -362,17 +362,17 @@
362
362
 
363
363
  AssertionError Traceback (most recent call last)
364
364
 
365
- <ipython-input-12-729512c0bf43> in <module>()
365
+ <ipython-input-3-9446ca790563> in <module>()
366
-
366
+
367
- 31 mu = mu_list.pop(0)
367
+ 164 mu = mu_list.pop(0)
368
-
368
+
369
- 32 sigma = sigma_list.pop(0)
369
+ 165 sigma = sigma_list.pop(0)
370
-
370
+
371
- ---> 33 make_subplot(10,10, i, new_row=True)
371
+ --> 166 make_subplot(10,10, i, new_row=True)
372
-
372
+
373
- 34 else:
373
+ 167 else:
374
-
374
+
375
- 35 make_subplot(10,10, i)
375
+ 168 make_subplot(10,10, i)
376
376
 
377
377
 
378
378
 

5

修正

2021/11/24 08:07

投稿

退会済みユーザー
test CHANGED
File without changes
test CHANGED
@@ -46,19 +46,21 @@
46
46
 
47
47
  ------------------------------------------------------
48
48
 
49
+
50
+
49
- def load_image(img_path: str):
51
+ def load_image(img_path: str):
50
-
52
+
51
- image = img.load_img(img_path, target_size=(224, 224))
53
+ image = img.load_img(img_path, target_size=(224, 224))
52
-
54
+
53
- plt.imshow(image)
55
+ plt.imshow(image)
54
-
56
+
55
- x = img.img_to_array(image)
57
+ x = img.img_to_array(image)
56
-
58
+
57
- return x
59
+ return x
58
-
59
-
60
-
60
+
61
+
62
+
61
- image = load_image('DSC_0897.jpg')
63
+ image = load_image('DSC_0897.jpg')
62
64
 
63
65
 
64
66
 

4

修正

2021/11/24 07:44

投稿

退会済みユーザー
test CHANGED
File without changes
test CHANGED
@@ -284,9 +284,13 @@
284
284
 
285
285
  label = np.argmax(first_pred)
286
286
 
287
- attack = foolbox.attacks.ProjectedGradientDescentAttack(fmodel, threshold=.999, distance=foolbox.distances.Linfinity, criterion=ConfidentMisclassification(.999))
288
-
289
- adversarial = attack(rand_noise[:, :, ::-1], label) ##ここです##
287
+ attack = foolbox.attacks.ProjectedGradientDescentAttack(fmodel, threshold=.999,
288
+
289
+ distance=foolbox.distances.Linfinity,
290
+
291
+ criterion=ConfidentMisclassification(.999))
292
+
293
+ adversarial = attack(rand_noise[:, :, ::-1], label) ##ここです##
290
294
 
291
295
  noise_preds = kmodel.predict(np.expand_dims(adversarial, axis=0))
292
296
 

3

インテンドの修正

2021/11/24 07:26

投稿

退会済みユーザー
test CHANGED
File without changes
test CHANGED
@@ -1,46 +1,48 @@
1
1
  ```
2
2
 
3
- !pip install tensorflow==2.0.0
3
+ !pip install tensorflow==2.0.0
4
-
4
+
5
- !pip install keras==2.3.1
5
+ !pip install keras==2.3.1
6
-
6
+
7
- !pip install foolbox==2.4.0
7
+ !pip install foolbox==2.4.0
8
-
8
+
9
- !pip3 install h5py==2.10.0
9
+ !pip3 install h5py==2.10.0
10
-
10
+
11
- import h5py
11
+ import h5py
12
-
12
+
13
- import numpy as np
13
+ import numpy as np
14
-
14
+
15
- from keras.applications.resnet50 import ResNet50
15
+ from keras.applications.resnet50 import ResNet50
16
-
16
+
17
- from foolbox.criteria import Misclassification, ConfidentMisclassification
17
+ from foolbox.criteria import Misclassification, ConfidentMisclassification
18
-
18
+
19
- from keras.preprocessing import image as img
19
+ from keras.preprocessing import image as img
20
-
20
+
21
- from keras.applications.resnet50 import preprocess_input, decode_predictions
21
+ from keras.applications.resnet50 import preprocess_input, decode_predictions
22
-
22
+
23
- import matplotlib.pyplot as plt
23
+ import matplotlib.pyplot as plt
24
-
24
+
25
- import foolbox
25
+ import foolbox
26
-
26
+
27
- import pprint as pp> import keras
27
+ import pprint as pp
28
+
28
-
29
+ import keras
30
+
29
- import pandas as pd
31
+ import pandas as pd
30
-
32
+
31
- %matplotlib inline
33
+ %matplotlib inline
32
34
 
33
35
 
34
36
 
35
37
  ----------------------------
36
38
 
39
+
40
+
37
- !wget https://github.com/GANs-in-Action/gans-in-action/raw/Chapter-10/chapter-10/DSC_0897.jpg
41
+ !wget https://github.com/GANs-in-Action/gans-in-action/raw/Chapter-10/chapter-10/DSC_0897.jpg
38
42
 
39
43
  !wget https://github.com/GANs-in-Action/gans-in-action/raw/Chapter-10/chapter-10/DSC_0896.jpg
40
44
 
41
- !wget https://raw.githubusercontent.com/GANs-in-Action/gans-in-action/master/chapter-10/initialization_vals_for_noise.csv
45
+ !wget https://raw.githubusercontent.com/GANs-in-Action/gans-in-action/master/chapter-10/initialization_vals_for_noise.csv
42
-
43
-
44
46
 
45
47
  ------------------------------------------------------
46
48
 
@@ -64,45 +66,59 @@
64
66
 
65
67
 
66
68
 
67
- keras.backend.set_learning_phase(0)
69
+ keras.backend.set_learning_phase(0)
68
-
70
+
69
- kmodel = ResNet50(weights='imagenet')
71
+ kmodel = ResNet50(weights='imagenet')
70
-
72
+
71
- preprocessing = (np.array([104, 116, 123]), 1)
73
+ preprocessing = (np.array([104, 116, 123]), 1)
74
+
72
-
75
+ #Kerasからfoolboxモデルのオブジェクトを生成する
73
-
74
-
76
+
75
- fmodel = foolbox.models.KerasModel(kmodel, bounds=(0, 255),
77
+ fmodel = foolbox.models.KerasModel(kmodel, bounds=(0, 255),
76
-
78
+
77
- preprocessing=preprocessing)
79
+ preprocessing=preprocessing)
80
+
78
-
81
+ #画像を(1,224,224,3)にしてResNet-50に合わせ、
82
+
79
-
83
+ #推定に使う画像がバッチに収まるようにする
80
-
84
+
81
- to_classify = np.expand_dims(image, axis=0)
85
+ to_classify = np.expand_dims(image, axis=0)
86
+
82
-
87
+ #predictを呼び出し、結果を表示する
88
+
83
- preds = kmodel.predict(to_classify)
89
+ preds = kmodel.predict(to_classify)
90
+
84
-
91
+ print('Predicted:',
92
+
85
- print('Predicted:', > pp.pprint(decode_predictions(preds, top=20)[0]))
93
+ pp.pprint(decode_predictions(preds, top=20)[0]))
94
+
86
-
95
+ #あとでラベルとして使うため、一番大きな数のインデックスを得る
96
+
87
- label = np.argmax(preds)
97
+ label = np.argmax(preds)
98
+
99
+
100
+
88
-
101
+ #Keras ReNet-50はRGBではなくBGRのため、::-1で色チャンネルを逆転させる
89
-
90
-
102
+
91
- image = image[:, :, ::-1]
103
+ image = image[:, :, ::-1]
104
+
92
-
105
+ #高い分類間違い基準を持った、攻撃のためのオブジェクトを作る
106
+
93
- attack = foolbox.v1.attacks.FGSM(fmodel, threshold=.9,
107
+ attack = foolbox.v1.attacks.FGSM(fmodel, threshold=.9,
94
-
108
+
95
- criterion=ConfidentMisclassification(.9))
109
+ criterion=ConfidentMisclassification(.9))
110
+
96
-
111
+ #ソース画像に攻撃を加える
112
+
97
- adversarial = attack(image, label)
113
+ adversarial = attack(image, label)
114
+
98
-
115
+ #敵対的サンプルのラベルを新たに推定する
99
-
100
-
116
+
101
- new_preds = kmodel.predict(np.expand_dims(adversarial, axis=0))
117
+ new_preds = kmodel.predict(np.expand_dims(adversarial, axis=0))
102
-
118
+
103
- print('Predicted:',
119
+ print('Predicted:',
104
-
120
+
105
- pp.pprint(decode_predictions(new_preds, top=20)[0]))
121
+ pp.pprint(decode_predictions(new_preds, top=20)[0]))
106
122
 
107
123
 
108
124
 
@@ -110,43 +126,43 @@
110
126
 
111
127
 
112
128
 
113
- plt.figure(figsize=(12,12))
129
+ plt.figure(figsize=(12,12))
114
-
115
-
116
-
130
+
131
+
132
+
117
- plt.subplot(1, 3, 1)
133
+ plt.subplot(1, 3, 1)
118
-
134
+
119
- plt.title('Original')
135
+ plt.title('Original')
120
-
136
+
121
- plt.imshow(image[:,:,::-1] / 255) # division by 255 to convert [0, 255] to [0, 1]
137
+ plt.imshow(image[:,:,::-1] / 255) # division by 255 to convert [0, 255] to [0, 1]
122
-
138
+
123
- plt.axis('off')
139
+ plt.axis('off')
124
-
125
-
126
-
140
+
141
+
142
+
127
- plt.subplot(1, 3, 2)
143
+ plt.subplot(1, 3, 2)
128
-
144
+
129
- plt.title('Adversarial')
145
+ plt.title('Adversarial')
130
-
146
+
131
- plt.imshow(adversarial[:, :, ::-1] / 255) # ::-1 to convert BGR to RGB
147
+ plt.imshow(adversarial[:, :, ::-1] / 255) # ::-1 to convert BGR to RGB
132
-
148
+
133
- plt.axis('off')
149
+ plt.axis('off')
134
-
135
-
136
-
150
+
151
+
152
+
137
- plt.subplot(1, 3, 3)
153
+ plt.subplot(1, 3, 3)
138
-
154
+
139
- plt.title('Difference')
155
+ plt.title('Difference')
140
-
156
+
141
- difference = adversarial[:, :, ::-1] - image
157
+ difference = adversarial[:, :, ::-1] - image
142
-
158
+
143
- plt.imshow(difference / abs(difference).max() * 0.2 + 0.5)
159
+ plt.imshow(difference / abs(difference).max() * 0.2 + 0.5)
144
-
160
+
145
- plt.axis('off')
161
+ plt.axis('off')
146
-
147
-
148
-
162
+
163
+
164
+
149
- plt.show()
165
+ plt.show()
150
166
 
151
167
 
152
168
 
@@ -154,91 +170,93 @@
154
170
 
155
171
 
156
172
 
157
- max_vals = pd.read_csv('initialization_vals_for_noise.csv')
173
+ max_vals = pd.read_csv('initialization_vals_for_noise.csv')
158
174
 
159
175
 
160
176
 
161
177
  -----------------------------------------
162
178
 
163
- fig = plt.figure(figsize=(20,20))
164
-
165
- #these are just lists of means and variances as floats
166
-
167
- sigma_list = list(max_vals.sigma)
168
-
169
- mu_list = list(max_vals.mu)
170
-
171
- conf_list = []>
172
-
173
-
174
-
175
- #this is the core function that
176
-
177
- def make_subplot(x, y, z, new_row=False):
178
-
179
- # sample noise as per for each mean and variance
180
-
181
- rand_noise = np.random.normal(loc=mu, scale=sigma, size=(224,224, 3))
182
-
183
- # only 0-255 pixel values permitted
184
-
185
- rand_noise = np.clip(rand_noise, 0, 255.)
186
-
187
- # get our first prediction
188
-
189
- noise_preds = kmodel.predict(np.expand_dims(rand_noise, axis=0))
190
-
191
- # get the predicted class and confidence respectively
192
-
193
- prediction, num = decode_predictions(noise_preds, top=20)[0][0][1:3]
194
-
195
- num = round(num * 100, 2)
196
-
197
- conf_list.append(num)
198
-
199
- # set up annotating code for the figure and then add the annotations and text
200
-
201
- ax = fig.add_subplot(x,y,z)
202
-
203
- ax.annotate(prediction, xy=(0.1, 0.6), xycoords=ax.transAxes, fontsize=16, olor='yellow')
204
-
205
- ax.annotate(f'{num}%' , xy=(0.1, 0.4), xycoords=ax.transAxes, fontsize=20, olor='orange')
206
-
207
- if new_row:
208
-
209
- ax.annotate(f'$\mu$:{mu}, $\sigma$:{sigma}' , xy=(-.2, 0.8), ycoords=ax.transAxes,
210
-
211
- rotation=90, fontsize=16, color='black')
212
-
213
- # division by 255 to convert [0, 255] to [0, 1]
214
-
215
- ax.imshow(rand_noise / 255)
216
-
217
- ax.axis('off')
218
-
219
-
220
-
221
-
222
-
223
- #this is the main for-loop that allows us to insert subplots into the figure
224
-
225
- for i in range(1,101):
226
-
227
- if (i-1) % 10==0:
228
-
229
- mu = mu_list.pop(0)
230
-
231
- sigma = sigma_list.pop(0)
232
-
233
- make_subplot(10,10, i, new_row=True)
234
-
235
- else:
236
-
237
- make_subplot(10,10, i)
238
-
239
-
240
-
241
- plt.show()
179
+
180
+
181
+ fig = plt.figure(figsize=(20,20))
182
+
183
+ # 平均と分散をfloatのリストにする
184
+
185
+ sigma_list = list(max_vals.sigma)
186
+
187
+ mu_list = list(max_vals.mu)
188
+
189
+ conf_list = []
190
+
191
+
192
+
193
+ # 図10.8を表示するためのコア関数
194
+
195
+ def make_subplot(x, y, z, new_row=False):
196
+
197
+ # それぞれの平均と分散を持ったのノイズをサンプリング
198
+
199
+ rand_noise = np.random.normal(loc=mu, scale=sigma, size=(224,224, 3))
200
+
201
+ # 0-255の画素値のみを許す
202
+
203
+ rand_noise = np.clip(rand_noise, 0, 255.)
204
+
205
+ # 最初の推定を得る
206
+
207
+ noise_preds = kmodel.predict(np.expand_dims(rand_noise, axis=0))
208
+
209
+ # 推定したクラスと確信度を得る
210
+
211
+ prediction, num = decode_predictions(noise_preds, top=20)[0][0][1:3]
212
+
213
+ num = round(num * 100, 2)
214
+
215
+ conf_list.append(num)
216
+
217
+ # 図10.8のための注釈用コードをセットアップしたのちに、注釈文を足す
218
+
219
+ ax = fig.add_subplot(x,y,z)
220
+
221
+ ax.annotate(prediction, xy=(0.1, 0.6), xycoords=ax.transAxes, fontsize=16, color='yellow')
222
+
223
+ ax.annotate(f'{num}%' , xy=(0.1, 0.4), xycoords=ax.transAxes, fontsize=20, color='orange')
224
+
225
+ if new_row:
226
+
227
+ ax.annotate(f'$\mu$:{mu}, $\sigma$:{sigma}' , xy=(-.2, 0.8), xycoords=ax.transAxes,
228
+
229
+ rotation=90, fontsize=16, color='black')
230
+
231
+ # [0, 255]を[0, 1]に変換するため、255で割る
232
+
233
+ ax.imshow(rand_noise / 255)
234
+
235
+ ax.axis('off')
236
+
237
+
238
+
239
+
240
+
241
+ # 画像内に文字を挿入するためのループ本体
242
+
243
+ for i in range(1,101):
244
+
245
+ if (i-1) % 10==0:
246
+
247
+ mu = mu_list.pop(0)
248
+
249
+ sigma = sigma_list.pop(0)
250
+
251
+ make_subplot(10,10, i, new_row=True)
252
+
253
+ else:
254
+
255
+ make_subplot(10,10, i)
256
+
257
+
258
+
259
+ plt.show()
242
260
 
243
261
 
244
262
 
@@ -246,101 +264,87 @@
246
264
 
247
265
 
248
266
 
249
- from keras.applications.inception_v3 import InceptionV3, decode_predictions
250
-
251
-
252
-
253
- #instantiate model
254
-
255
- keras.backend.set_learning_phase(0)
256
-
257
- kmodel_v3 = InceptionV3(weights='imagenet')
267
+ fig = plt.figure(figsize=(20,20))
268
+
269
+ sigma_list = list(max_vals.sigma)
270
+
271
+ mu_list = list(max_vals.mu)
272
+
273
+ sum_pred = []
274
+
275
+
276
+
277
+ def make_subplot(x, y, z, new_row=False):
278
+
279
+ rand_noise = np.random.normal(loc=mu, scale=sigma, size=(224,224, 3))
280
+
281
+ rand_noise = np.clip(rand_noise, 0, 255.)
282
+
283
+ first_pred = kmodel.predict(np.expand_dims(rand_noise, axis=0))
284
+
285
+ label = np.argmax(first_pred)
286
+
287
+ attack = foolbox.attacks.ProjectedGradientDescentAttack(fmodel, threshold=.999, distance=foolbox.distances.Linfinity, criterion=ConfidentMisclassification(.999))
288
+
289
+ adversarial = attack(rand_noise[:, :, ::-1], label) ##ここです##
290
+
291
+ noise_preds = kmodel.predict(np.expand_dims(adversarial, axis=0))
292
+
293
+ prediction, num = decode_predictions(noise_preds, top=20)[0][0][1:3]
294
+
295
+ num = round(num * 100, 2)
296
+
297
+ sum_pred.append(num)
298
+
299
+ ax = fig.add_subplot(x,y,z)
300
+
301
+ ax.annotate(prediction, xy=(0.1, 0.6), xycoords=ax.transAxes, fontsize=16, color='yellow')
302
+
303
+ ax.annotate(f'{num}%', xy=(0.1, 0.4), xycoords=ax.transAxes, fontsize=20, color='orange')
304
+
305
+ if new_row:
306
+
307
+ ax.annotate(f'$\mu$:{mu}, $\sigma$:{sigma}' , xy=(-.2, 0.8), xycoords=ax.transAxes,
308
+
309
+ rotation=90, fontsize=16, color='black')
310
+
311
+ ax.imshow(adversarial / 255) # division by 255 to convert [0, 255] to [0, 1]
312
+
313
+ ax.axis('off')
314
+
315
+
316
+
317
+
318
+
319
+ for i in range(1,101):
320
+
321
+ if (i-1) % 10==0:
322
+
323
+ mu = mu_list.pop(0)
324
+
325
+ sigma = sigma_list.pop(0)
326
+
327
+ make_subplot(10,10, i, new_row=True)
328
+
329
+ else:
330
+
331
+ make_subplot(10,10, i)
332
+
333
+
334
+
335
+ plt.show()
258
336
 
259
337
 
260
338
 
261
339
  --------------------------------------------------
262
340
 
263
- fig = plt.figure(figsize=(20,20))
264
-
265
- sigma_list = list(max_vals.sigma)
266
-
267
- mu_list = list(max_vals.mu)
268
-
269
- sum_pred = []
270
-
271
-
272
-
273
- def make_subplot(x, y, z, new_row=False):
274
-
275
- rand_noise = np.random.normal(loc=mu, scale=sigma, size=(224,224, 3))
276
-
277
- rand_noise = np.clip(rand_noise, 0, 255.)
278
-
279
- first_pred = kmodel.predict(np.expand_dims(rand_noise, axis=0))
280
-
281
- label = np.argmax(first_pred)
282
-
283
- attack = foolbox.attacks.ProjectedGradientDescentAttack(fmodel, threshold=.999,
284
-
285
- distance=foolbox.distances.Linfinity,
286
-
287
- criterion=ConfidentMisclassification(.999))
288
-
289
- → → adversarial = attack(rand_noise[:, :, ::-1], label)
290
-
291
- noise_preds = kmodel.predict(np.expand_dims(adversarial, axis=0))
292
-
293
- prediction, num = decode_predictions(noise_preds, top=20)[0][0][1:3]
294
-
295
- num = round(num * 100, 2)
296
-
297
- sum_pred.append(num)
298
-
299
- ax = fig.add_subplot(x,y,z)
300
-
301
- ax.annotate(prediction, xy=(0.1, 0.6), xycoords=ax.transAxes, fontsize=16, color='yellow')
302
-
303
- ax.annotate(f'{num}%', xy=(0.1, 0.4), xycoords=ax.transAxes, fontsize=20, color='orange')
304
-
305
- if new_row:
306
-
307
- ax.annotate(f'$\mu$:{mu}, $\sigma$:{sigma}' , xy=(-.2, 0.8), xycoords=ax.transAxes,
308
-
309
- rotation=90, fontsize=16, color='black')
310
-
311
- ax.imshow(adversarial / 255) # division by 255 to convert [0, 255] to [0, 1]
312
-
313
- ax.axis('off')
314
-
315
-
316
-
317
-
318
-
319
- for i in range(1,101):
320
-
321
- if (i-1) % 10==0:
322
-
323
- mu = mu_list.pop(0)
324
-
325
- sigma = sigma_list.pop(0)
326
-
327
- make_subplot(10,10, i, new_row=True)
328
-
329
- else:
330
-
331
- make_subplot(10,10, i)
332
-
333
-
334
-
335
- plt.show()
336
-
337
341
 
338
342
 
339
343
  ```
340
344
 
341
345
 
342
346
 
343
- を実行しようとすると、最後の段落矢印のところで
347
+ を実行しようとすると、最後の段落の##ここです##のところで
344
348
 
345
349
  -----------------------------------------
346
350
 

2

2021/11/24 07:21

投稿

退会済みユーザー
test CHANGED
File without changes
test CHANGED
@@ -340,7 +340,7 @@
340
340
 
341
341
 
342
342
 
343
- を実行しようとすると、最後の段落で
343
+ を実行しようとすると、最後の段落矢印のところで
344
344
 
345
345
  -----------------------------------------
346
346
 
@@ -391,3 +391,9 @@
391
391
 
392
392
 
393
393
  というエラーが表示されてしまします。いろいろ調べてもわかりません。教えてほしいです。お願いします。
394
+
395
+
396
+
397
+ 引用は、主にhttps://colab.research.google.com/drive/1bsmaB23bSEuu1zWyPSajDJ_DO3QqlovQ
398
+
399
+ こちらです。

1

2021/11/23 15:51

投稿

退会済みユーザー
test CHANGED
File without changes
test CHANGED
@@ -286,7 +286,7 @@
286
286
 
287
287
  criterion=ConfidentMisclassification(.999))
288
288
 
289
- adversarial = attack(rand_noise[:, :, ::-1], label)
289
+ → → adversarial = attack(rand_noise[:, :, ::-1], label)
290
290
 
291
291
  noise_preds = kmodel.predict(np.expand_dims(adversarial, axis=0))
292
292