-
Notifications
You must be signed in to change notification settings - Fork 0
/
ALD.py
389 lines (334 loc) · 16.1 KB
/
ALD.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
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
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
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
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
"""
在干净样本上训练生成对抗网络，
当对抗补丁引入后，对抗样本所在区域会变的难以生成，
导致生成图像和输入图像在对抗补丁所在区域产生较大的异常损失，
尝试以此方法定位对抗补丁位置
"""
import math
import cv2
import numpy as np
import torch.nn as nn
import torch.nn.functional as F
import torch
from matplotlib import pyplot as plt
from tqdm import tqdm
from target_models import VGGNet_19
def weights_init_normal(m):
classname = m.__class__.__name__
if classname.find("Conv") != -1:
torch.nn.init.normal_(m.weight.data, 0.0, 0.02)
elif classname.find("BatchNorm2d") != -1:
torch.nn.init.normal_(m.weight.data, 1.0, 0.02)
torch.nn.init.constant_(m.bias.data, 0.0)
##############################################
# U-NET----generator of pix2pix
##############################################
class UNetDown(nn.Module):
def __init__(self, in_size, out_size, normalize=True, dropout=0.0):
super(UNetDown, self).__init__()
layers = [nn.Conv2d(in_size, out_size, 4, 2, 1, bias=False)]
if normalize:
layers.append(nn.InstanceNorm2d(out_size))
layers.append(nn.LeakyReLU(0.2))
if dropout:
layers.append(nn.Dropout(dropout))
self.model = nn.Sequential(*layers)
def forward(self, x):
return self.model(x)
class UNetUp(nn.Module):
def __init__(self, in_size, out_size, dropout=0.0):
super(UNetUp, self).__init__()
layers = [
nn.ConvTranspose2d(in_size, out_size, 4, 2, 1, bias=False),
nn.InstanceNorm2d(out_size),
nn.ReLU(inplace=True),
]
if dropout:
layers.append(nn.Dropout(dropout))
self.model = nn.Sequential(*layers)
def forward(self, x, skip_input):
x = self.model(x)
x = torch.cat((x, skip_input), 1)
return x
class GeneratorUNet(nn.Module):
def __init__(self, in_channels=3, out_channels=3):
super(GeneratorUNet, self).__init__()
self.down1 = UNetDown(in_channels, 64, normalize=False)
self.down2 = UNetDown(64, 128)
self.down3 = UNetDown(128, 256)
self.down4 = UNetDown(256, 512, dropout=0.5)
self.down5 = UNetDown(512, 512, dropout=0.5)
self.down6 = UNetDown(512, 512, dropout=0.5)
self.down7 = UNetDown(512, 512, dropout=0.5)
self.down8 = UNetDown(512, 512, normalize=False, dropout=0.5)
self.up1 = UNetUp(512, 512, dropout=0.5)
self.up2 = UNetUp(1024, 512, dropout=0.5)
self.up3 = UNetUp(1024, 512, dropout=0.5)
self.up4 = UNetUp(1024, 512, dropout=0.5)
self.up5 = UNetUp(1024, 256)
self.up6 = UNetUp(512, 128)
self.up7 = UNetUp(256, 64)
self.final = nn.Sequential(
nn.Upsample(scale_factor=2),
nn.ZeroPad2d((1, 0, 1, 0)),
nn.Conv2d(128, out_channels, 4, padding=1),
nn.Tanh(),
)
def forward(self, x):
# U-Net generator with skip connections from encoder to decoder
d1 = self.down1(x) # 64
d2 = self.down2(d1) # 128
d3 = self.down3(d2) # 256
d4 = self.down4(d3) # 512
d5 = self.down5(d4)
d6 = self.down6(d5)
d7 = self.down7(d6)
d8 = self.down8(d7)
u1 = self.up1(d8, d7) # 1024
u2 = self.up2(u1, d6) # 1024
u3 = self.up3(u2, d5) # 1024
u4 = self.up4(u3, d4) # 1024
u5 = self.up5(u4, d3) # 512
u6 = self.up6(u5, d2)
u7 = self.up7(u6, d1)
return self.final(u7)
##############################
# Discriminator
##############################
class Discriminator(nn.Module):
def __init__(self, in_channels=3):
super(Discriminator, self).__init__()
def discriminator_block(in_filters, out_filters, normalization=True):
"""Returns downsampling layers of each discriminator block"""
layers = [nn.Conv2d(in_filters, out_filters, 4, stride=2, padding=1)]
if normalization:
layers.append(nn.InstanceNorm2d(out_filters))
layers.append(nn.LeakyReLU(0.2, inplace=True))
return layers
self.model = nn.Sequential(
*discriminator_block(in_channels * 2, 64, normalization=False),
# *discriminator_block(7, 64, normalization=False),
*discriminator_block(64, 128),
*discriminator_block(128, 256),
*discriminator_block(256, 512),
nn.ZeroPad2d((1, 0, 1, 0)),
nn.Conv2d(512, 1, 4, padding=1, bias=False)
)
def forward(self, img_A, img_B):
# Concatenate image and condition image by channels to produce input
img_input = torch.cat((img_A, img_B), 1)
return self.model(img_input)
def comput_IoU(mask_predict, mask_x, threshold):
mask_predict = torch.where(mask_predict < threshold, 0, 1)
mask_predict = torch.sum(mask_predict, dim=1)
mask_predict = torch.where(mask_predict != 0, 1, 0)
mask_x = torch.sum(mask_x, dim=1)
mask_x = torch.where(mask_x != 0, 1, 0)
iou = 0
for i in range(5):
result = mask_x[i] + mask_predict[i]
Intersection = (result == 2).sum().item()
Union = (result != 0).sum().item()
iou += Intersection / Union
return iou / 5
def comput_Coverage(mask_predict, mask_x, threshold):
mask_predict = torch.where(mask_predict < threshold, 0, 1)
mask_predict = torch.sum(mask_predict, dim=1)
mask_predict = torch.where(mask_predict != 0, 1, 0)
mask_x = torch.sum(mask_x, dim=1)
mask_x = torch.where(mask_x != 0, 1, 0)
Coverage = 0
for i in range(5):
original = (mask_x[i] == 1).sum().item()
result = mask_x[i] + mask_predict[i]
cover = (result == 2).sum().item()
Coverage += cover / original
return Coverage / 5
if __name__ == '__main__':
generator = GeneratorUNet()
cuda = True if torch.cuda.is_available() else False
if cuda:
generator = generator.cuda()
# # ---------------------------------训练好的pix2pix网络------------------------------
generator.load_state_dict(torch.load(" "))
generator.eval()
# # ---------------------------------加载对抗补丁------------------------------
ori_y_path = ["D:\\Study\\wlh\\adversarial_patch\\PatchData\\APA32\\ori_y_2000.npy",
"D:\\Study\\wlh\\adversarial_patch\\PatchData\\APA\\ori_y_2000.npy",
"D:\\Study\\wlh\\adversarial_patch\\PatchData\\APA64\\ori_y_2000.npy",
"D:\\Study\\wlh\\adversarial_patch\\PatchData\\GDPA\\ori_y_2000.npy",
"G:\\wlh\\chapters2\\AEDD_methods\\adv_data\\GDPA48\\ori_y_2000.npy",
"G:\\wlh\\chapters2\\AEDD_methods\\adv_data\\GDPA64\\ori_y_2000.npy"]
adx_x_path = ["D:\\Study\\wlh\\adversarial_patch\\PatchData\\APA32\\APA_Vgg19_32patch_2000.npy",
"D:\\Study\\wlh\\adversarial_patch\\PatchData\\APA\\APA_Vgg19_43x43patch_2000.npy",
"D:\\Study\\wlh\\adversarial_patch\\PatchData\\APA64\\APA_Vgg19_64patch_2000.npy",
"D:\\Study\\wlh\\adversarial_patch\\PatchData\\GDPA\\GDPA_Vgg19_32x32patch_2000.npy",
"G:\\wlh\\chapters2\\AEDD_methods\\adv_data\\GDPA48\\GDPA_Vgg19_48patch_2000.npy",
"G:\\wlh\\chapters2\\AEDD_methods\\adv_data\\GDPA64\\GDPA_Vgg19_64patch_2000.npy"]
mask_x_path = ["D:\\Study\\wlh\\adversarial_patch\\PatchData\\APA32\\APA_Vgg19_32mask_2000.npy",
"D:\\Study\\wlh\\adversarial_patch\\PatchData\\APA\\APA_Vgg19_43x43mask_2000.npy",
"D:\\Study\\wlh\\adversarial_patch\\PatchData\\APA64\\APA_Vgg19_64mask_2000.npy",
"D:\\Study\\wlh\\adversarial_patch\\PatchData\\GDPA\\GDPA_Vgg19_32x32mask_2000.npy",
"G:\\wlh\\chapters2\\AEDD_methods\\adv_data\\GDPA48\\GDPA_Vgg19_48mask_2000.npy",
"G:\\wlh\\chapters2\\AEDD_methods\\adv_data\\GDPA64\\GDPA_Vgg19_64mask_2000.npy"]
thresholds = [0.5, 0.4, 0.3, 0.2, 0.1]
# # ---------------------------------外层循环加载数据，内层循环变更 threshold------------------------------
for index in range(6):
ori_y = np.load(ori_y_path[index])
adv_x = np.load(adx_x_path[index])
mask_x = np.load(mask_x_path[index])
for threshold in thresholds:
print("---------------------", ori_y_path[index].split('\\')[-2], "---", threshold, "---------------------------")
# # # ----------------------------------计算掩码交并比---------------------------
# ori_x = torch.tensor(ori_x)
# ori_y = torch.tensor(ori_y)
adv_x = torch.tensor(adv_x)
mask_x = torch.tensor(mask_x)
IoU = 0
Coverage = 0
for i in tqdm(range(400)):
# ori_x_input = ori_x[i * 5:(i + 1) * 5].cuda()
adv_x_input = adv_x[i * 5:(i + 1) * 5].cuda()
# target = ori_y[i * 5:(i + 1) * 5]
target_mask = mask_x[i * 5:(i + 1) * 5].cuda()
pix2pix_adv = generator(adv_x_input)
mask_predict = torch.abs(pix2pix_adv - adv_x_input)
# adv_x_input = np.array(adv_x_input.cpu())
# adv_x_input = np.transpose(adv_x_input, (0, 2, 3, 1))
# mask_predict = np.array(mask_predict.detach().cpu())
# mask_predict = np.transpose(mask_predict, (0, 2, 3, 1))
# plt.rcParams['font.sans-serif'] = ['SimHei'] # 中文乱码
# for i in range(2):
# for j in range(5):
# num = i * 5 + j + 1
# if i == 0:
# plt.subplot(2, 5, num)
# plt.imshow(adv_x_input[j])
# plt.axis('off')
# elif i == 1:
# plt.subplot(2, 5, num)
# plt.imshow(mask_predict[j])
# plt.axis('off')
# plt.show()
#
IoU += comput_IoU(mask_predict, target_mask, threshold)
Coverage += comput_Coverage(mask_predict, target_mask, threshold)
#
print("交并比：", IoU / 400)
print("覆盖率：", Coverage / 400)
# -------------------------------加载目标分类模型，利用得到的掩码覆盖对抗补丁，计算分类准确率---------------------------
model = VGGNet_19()
model.load_state_dict(torch.load(" "))
model = model.cuda()
model.eval()
ori_y = torch.tensor(ori_y)
adv_x = torch.tensor(adv_x)
total = [0,0,0,0]
correct = [0,0,0,0]
acc_test = [0,0,0,0]
for i in tqdm(range(400)):
# ori_x_input = ori_x[i * 5:(i + 1) * 5].cuda()
adv_x_input = adv_x[i * 5:(i + 1) * 5].cuda()
target = ori_y[i * 5:(i + 1) * 5].cuda()
pix2pix_adv = generator(adv_x_input)
mask_predict = torch.abs(pix2pix_adv - adv_x_input)
mask_predict = torch.where(mask_predict < threshold, 0, 1)
hide = torch.ones((5,3,256,256)).cuda()
mask_predict = torch.sum(mask_predict, dim=1)
mask_predict = torch.where(mask_predict != 0, 1, 0)
mask_predict = torch.unsqueeze(mask_predict, dim=1)
hide_image = torch.where(mask_predict == 1, hide, adv_x_input)
# 图像修复
mask = np.array(torch.where(torch.sum(mask_predict, dim=1)!=0, 1, 0).cpu().detach())
mask = mask.astype(np.uint8)
image = np.array(hide_image.cpu()) * 255
image = np.transpose(image, (0, 2, 3, 1))
image = image.astype(np.uint8)
adv_x_repair = []
for i in range(5):
adv_x_repair.append(cv2.inpaint(image[i], mask[i], 3, cv2.INPAINT_TELEA))
# plt.rcParams['font.sans-serif'] = ['SimHei'] # 中文乱码
# plt.imshow(adv_x_repair[i])
# plt.show()
adv_x_repair = torch.tensor(np.transpose(adv_x_repair, (0, 3, 1, 2)), dtype=torch.float32).cuda()
adv_x_repair = adv_x_repair/255
# output = model(ori_x_input)
# _, predicted = torch.max(output.data, dim=1)
# total[0] += target.size(0)
# correct[0] += (predicted == target).sum().item()
# output = model(adv_x_input)
# _, predicted = torch.max(output.data, dim=1)
# total[1] += target.size(0)
# correct[1] += (predicted == target).sum().item()
output = model(hide_image)
_, predicted = torch.max(output.data, dim=1)
total[2] += target.size(0)
correct[2] += (predicted == target).sum().item()
output = model(adv_x_repair)
_, predicted = torch.max(output.data, dim=1)
total[3] += target.size(0)
correct[3] += (predicted == target).sum().item()
# acc_test[0] = correct[0] / total[0]
# acc_test[1] = correct[1] / total[1]
acc_test[2] = correct[2] / total[2]
acc_test[3] = correct[3] / total[3]
# print("干净样本准确率：%.5f" % acc_test[0])
# print("对抗补丁准确率：%.5f" % acc_test[1])
print("遮挡样本准确率：%.5f" % acc_test[2])
print("修复样本准确率：%.5f" % acc_test[3])
# # -----------------------------------可视化-------------------------------------
# ori_x = torch.tensor(ori_x)
# ori_y = torch.tensor(ori_y)
# adv_x = torch.tensor(adv_x)
# mask_x = torch.tensor(mask_x)
# IoU = 0
# for i in range(400):
# ori_x_input = ori_x[i * 5:(i + 1) * 5].cuda()
# adv_x_input = adv_x[i * 5:(i + 1) * 5].cuda()
# target = ori_y[i * 5:(i + 1) * 5]
# mask_x = mask_x[i * 5:(i + 1) * 5].cuda()
#
# pix2pix_ori = generator(ori_x_input)
# pix2pix_adv = generator(adv_x_input)
#
# real_A = np.array(ori_x_input.cpu())
# real_A = np.transpose(real_A, (0, 2, 3, 1))
#
# fake_A = np.array(pix2pix_ori.detach().cpu())
# fake_A = np.transpose(fake_A, (0, 2, 3, 1))
#
# real_B = np.array(adv_x_input.cpu())
# real_B = np.transpose(real_B, (0, 2, 3, 1))
#
# fake_B = np.array(pix2pix_adv.detach().cpu())
# fake_B = np.transpose(fake_B, (0, 2, 3, 1))
#
# plt.rcParams['font.sans-serif'] = ['SimHei'] # 中文乱码
# for i in range(6):
# for j in range(5):
# num = i * 5 + j + 1
# if i == 0:
# plt.subplot(6, 5, num)
# plt.imshow(real_A[j])
# plt.axis('off')
# elif i == 1:
# plt.subplot(6, 5, num)
# plt.imshow(fake_A[j])
# plt.axis('off')
# elif i == 2:
# plt.subplot(6, 5, num)
# plt.imshow(np.fabs(real_A[j] - fake_A[j]))
# plt.axis('off')
# elif i == 3:
# plt.subplot(6, 5, num)
# plt.imshow(real_B[j])
# plt.axis('off')
# elif i == 4:
# plt.subplot(6, 5, num)
# plt.imshow(fake_B[j])
# plt.axis('off')
# elif i == 5:
# plt.subplot(6, 5, num)
# plt.imshow(np.fabs(real_B[j] - fake_B[j]))
# plt.axis('off')
# plt.show()