-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdataset.py
257 lines (218 loc) · 10.5 KB
/
dataset.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
import os
import sys
import inspect
currentdir = os.path.dirname(os.path.abspath(inspect.getfile(inspect.currentframe())))
parentdir = os.path.dirname(currentdir)
sys.path.insert(0, parentdir)
parentdir = os.path.dirname(parentdir)
sys.path.insert(0, parentdir)
parentdir = os.path.dirname(parentdir)
sys.path.insert(0, parentdir)
from torchvision import datasets, transforms
from torch.utils.data import Dataset
import PIL.Image as Image
import torch
import cv2
import numpy as np
from config_cifar10 import *
def get_data_from_pt(pt_root, max_num=None, expand=None):
# normalized data
data_dict = torch.load(pt_root)
adv_input_dict = data_dict['adv_inputs']
label_dict = data_dict['labels']
img_name_dict = data_dict['img_names']
num_cls = len(adv_input_dict)
img_list = []
label_list = []
img_name_list = []
for cls in adv_input_dict:
num = len(adv_input_dict[cls])
if max_num is None:
max_num = num
img_list.append(adv_input_dict[cls][:max_num])
label_list.append(label_dict[cls][:max_num])
img_name_list += img_name_dict[cls][:max_num]
adv_input_list_cat = torch.cat(img_list, 0)
label_list_cat = torch.cat(label_list, 0)
if expand is not None:
for idx in range(len(label_list_cat)):
y = label_list_cat[idx] + expand*num_cls
label_list_cat[idx] = y
return adv_input_list_cat, label_list_cat, img_name_list
def make_train_dict(train_dict, attack_method, _adv_train_root, sub_root=None, eps = 8, max_num=None):
if sub_root is None:
sub_root = attack_method + '_eps' + str(eps) + '.pt'
train_root = os.path.join(_adv_train_root, sub_root)
adv_input_list_cat, label_list_cat, img_name_list = get_data_from_pt(train_root, max_num=max_num)
train_dict['image_list'].append(adv_input_list_cat)
train_dict['label_list'].append(label_list_cat)
train_dict['img_name_list'] += img_name_list
return train_dict
def make_train_dict_expand(train_dict, attack_method, _adv_train_root, sub_root=None, eps = 8, max_num=None, expand=0):
if sub_root is None:
sub_root = attack_method + '_eps' + str(eps) + '.pt'
train_root = os.path.join(_adv_train_root, sub_root)
adv_input_list_cat, label_list_cat, img_name_list = get_data_from_pt(train_root, max_num=max_num, expand=expand)
train_dict['image_list'].append(adv_input_list_cat)
train_dict['label_list'].append(label_list_cat)
train_dict['img_name_list'] += img_name_list
return train_dict
def make_test_dicts(test_dicts, attack_method, _adv_test_root, sub_root=None, eps = 8, max_num=None):
if sub_root is None:
sub_root = attack_method + '_eps' + str(eps) + '.pt'
test_root = os.path.join(_adv_test_root, sub_root)
adv_input_list_cat, label_list_cat, img_name_list = get_data_from_pt(test_root, max_num=max_num,)
test_dicts[attack_method] = {}
test_dicts[attack_method]['image_list'] = adv_input_list_cat
test_dicts[attack_method]['label_list'] = label_list_cat
test_dicts[attack_method]['img_name_list'] = img_name_list
return test_dicts
class torchdict_Dataset(Dataset):
def __init__(self, adv_input_list_cat, label_list_cat, img_name_list, max_num=None):
if max_num is not None:
self.image_list = adv_input_list_cat[:max_num]
self.label_list = label_list_cat[:max_num]
self.img_name_list = img_name_list[:max_num]
else:
self.image_list = adv_input_list_cat
self.label_list = label_list_cat
self.img_name_list = img_name_list[:max_num]
# random.shuffle(self.image_list)
def __getitem__(self, item):
img = self.image_list[item]
label = self.label_list[item]
img_name = self.img_name_list[item]
return img, label, img_name
def __len__(self):
return len(self.image_list)
class torchdict2imagelist_Dataset(Dataset):
def __init__(self, adv_input_list_cat, label_list_cat, img_name_list, transform=None, max_num=None):
if max_num is not None:
self.image_list = adv_input_list_cat[:max_num]
self.label_list = label_list_cat[:max_num]
self.img_name_list = img_name_list[:max_num]
else:
self.image_list = adv_input_list_cat
self.label_list = label_list_cat
self.img_name_list = img_name_list[:max_num]
# random.shuffle(self.image_list)
self.transform = transform
self.image_list = self.image_list.permute(0, 2, 3, 1).cpu().numpy() * 255 # (b,H,W,C) numpy.array
def __getitem__(self, item):
img = self.image_list[item]
img = Image.fromarray(img.astype(np.uint8)[:,:,[2,1,0]], mode='RGB')
if self.transform is not None:
img = self.transform(img)
label = int(self.label_list[item])
img_name = self.img_name_list[item]
return img, label, img_name
def __len__(self):
return len(self.image_list)
def get_dataset_all_pt(args, attacks, w_clean = True):
if args.dataset == 'cifar10':
_adv_train_root = adv_train_pt_root
_adv_test_root = adv_test_pt_root
transform_train = transforms.Compose([
transforms.RandomCrop(32, padding = 4),
transforms.RandomHorizontalFlip(),
transforms.ToTensor()
])
transform_test = transforms.Compose([
transforms.ToTensor()
])
else:
raise Exception
train_dict = {}
train_dict['image_list'] = []
train_dict['label_list'] = []
train_dict['img_name_list'] = []
test_dicts = {}
for attack in attacks:
train_dict = make_train_dict(train_dict, attack, _adv_train_root, eps = args.eps)
test_dicts = make_test_dicts(test_dicts, attack, _adv_test_root, eps = args.eps)
if w_clean:
attack_method = 'Clean'
sub_root = sub_train_root
train_dict = make_train_dict(train_dict, attack_method, cifar10_root, sub_root, eps = args.eps)
sub_root = sub_test_root
test_dicts = make_test_dicts(test_dicts, attack_method, cifar10_root, sub_root, eps = args.eps)
if not w_clean and attacks is None:
train_dataset = None
else:
train_dict['image_list'] = torch.cat(train_dict['image_list'], 0)
train_dict['label_list'] = torch.cat(train_dict['label_list'], 0)
train_dataset = torchdict2imagelist_Dataset(train_dict['image_list'], train_dict['label_list'], train_dict['img_name_list'], transform=transform_train)
if not w_clean and attacks is None:
test_datasets = None
else:
test_datasets = {}
for attack_method in attacks:
test_dict = test_dicts[attack_method]
test_datasets[attack_method] = torchdict2imagelist_Dataset(test_dict['image_list'], test_dict['label_list'], test_dict['img_name_list'], transform=transform_test)
if w_clean:
attack_method = 'Clean'
test_dict = test_dicts[attack_method]
test_datasets[attack_method] = torchdict_Dataset(test_dict['image_list'], test_dict['label_list'], test_dict['img_name_list'])
return train_dataset, test_datasets
def get_dataset_session0_pt(args, attack, proto=False):
train_dict = {}
train_dict['image_list'] = []
train_dict['label_list'] = []
train_dict['img_name_list'] = []
test_dicts = {}
if args.dataset == 'cifar10':
_adv_train_root = adv_train_pt_root
_adv_test_root = adv_test_pt_root
if proto:
transform_train = transforms.Compose([
transforms.ToTensor()
])
else:
transform_train = transforms.Compose([
transforms.RandomCrop(32, padding = 4),
transforms.RandomHorizontalFlip(),
transforms.ToTensor()
])
transform_test = transforms.Compose([
transforms.ToTensor()
])
else:
raise Exception
train_dict = make_train_dict(train_dict, attack, _adv_train_root, eps = args.eps)
test_dicts = make_test_dicts(test_dicts, attack, _adv_test_root, eps = args.eps)
train_dict['image_list'] = torch.cat(train_dict['image_list'], 0)
train_dict['label_list'] = torch.cat(train_dict['label_list'], 0)
train_dataset_base = torchdict2imagelist_Dataset(train_dict['image_list'], train_dict['label_list'], train_dict['img_name_list'], transform=transform_train)
test_dataset_base = torchdict2imagelist_Dataset(test_dicts[attack]['image_list'], test_dicts[attack]['label_list'], test_dicts[attack]['img_name_list'], transform=transform_test)
return train_dataset_base, test_dataset_base
def get_dataset_fewshot_pt(args, attack, expand=0, num_shot=1, max_num=1000, proto=False):
train_dict = {}
train_dict['image_list'] = []
train_dict['label_list'] = []
train_dict['img_name_list'] = []
test_dicts = {}
if args.dataset == 'cifar10':
_adv_train_root = adv_train_pt_root
_adv_test_root = adv_test_pt_root
if proto:
transform_train = transforms.Compose([
transforms.ToTensor()
])
else:
transform_train = transforms.Compose([
transforms.RandomCrop(32, padding = 4),
transforms.RandomHorizontalFlip(),
transforms.ToTensor()
])
transform_test = transforms.Compose([
transforms.ToTensor()
])
else:
raise Exception
train_dict = make_train_dict_expand(train_dict, attack, _adv_train_root, eps = args.eps, max_num=num_shot, expand=expand)
test_dicts = make_test_dicts(test_dicts, attack, _adv_test_root, eps = args.eps, max_num=max_num)
train_dict['image_list'] = torch.cat(train_dict['image_list'], 0)
train_dict['label_list'] = torch.cat(train_dict['label_list'], 0)
train_dataset_fewshot = torchdict2imagelist_Dataset(train_dict['image_list'], train_dict['label_list'], train_dict['img_name_list'], transform=transform_train)
test_dataset_fewshot = torchdict2imagelist_Dataset(test_dicts[attack]['image_list'], test_dicts[attack]['label_list'], test_dicts[attack]['img_name_list'], transform=transform_test)
return train_dataset_fewshot, test_dataset_fewshot