-
Notifications
You must be signed in to change notification settings - Fork 0
/
datasets.py
232 lines (194 loc) · 8.49 KB
/
datasets.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
import numpy as np
from torch.utils.data import Dataset
from PIL import Image
from torchvision import transforms
from PIL import ImageFilter
import random
from officehome import OfficeHome
from visdac import VISDAC
from domainnet import DomainNet
class GaussianBlur(object):
"""Gaussian blur augmentation in SimCLR https://arxiv.org/abs/2002.05709"""
def __init__(self, sigma=[0.1, 2.0]):
self.sigma = sigma
def __call__(self, x):
sigma = random.uniform(self.sigma[0], self.sigma[1])
x = x.filter(ImageFilter.GaussianBlur(radius=sigma))
return x
normalize = transforms.Normalize(
mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]
)
class dataset(Dataset):
def __init__(self, dataset, root, imb, mode, transform, noisy_path=None):
self.dataset = dataset
self.root = root
self.imb = imb
self.mode = mode
self.transform = transform
self.noisy_path = noisy_path
self.parse_dataset()
def parse_dataset(self):
if self.dataset.split('/')[0] == 'officehome':
return self.get_officehome()
elif self.dataset.split('/')[0] == 'visdac':
return self.get_visdac()
elif self.dataset.split('/')[0] == 'domainnet':
return self.get_domainnet()
def get_officehome(self):
domain = self.dataset.split('/')[-1]
if self.mode == 'all':
train_set = OfficeHome(root=self.root,
imbalanced=self.imb,
domain=domain,
train=True,
transform=self.transform,
from_file=False
)
test_set = OfficeHome(root=self.root,
imbalanced=self.imb,
domain=domain,
train=False,
transform=self.transform,
from_file=False
)
data = np.concatenate((train_set.data, test_set.data))
labels = np.concatenate((train_set.labels, test_set.labels))
else:
train = True if self.mode == 'train' else False
dataset = OfficeHome(root=self.root,
imbalanced=self.imb,
domain=domain,
train=train,
transform=self.transform,
from_file=False
)
data = dataset.data
labels = dataset.labels
self.strong_augmentation = transforms.Compose(
[
transforms.RandomResizedCrop(224, scale=(0.2, 1.0), antialias=None),
transforms.RandomApply(
[transforms.ColorJitter(0.4, 0.4, 0.4, 0.1)],
p=0.8, # not strengthened
),
transforms.RandomGrayscale(p=0.2),
transforms.RandomApply([GaussianBlur([0.1, 2.0])], p=0.5),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
normalize,
]
)
self.data = data
self.labels = labels
def get_visdac(self):
domain = self.dataset.split('/')[-1]
if self.mode == 'all':
train_set = VISDAC(root=self.root,
imbalanced=self.imb,
domain=domain,
train=True,
transform=self.transform,
from_file=False
)
test_set = VISDAC(root=self.root,
imbalanced=self.imb,
domain=domain,
train=False,
transform=self.transform,
from_file=False
)
data = np.concatenate((train_set.data, test_set.data))
labels = np.concatenate((train_set.labels, test_set.labels))
else:
train = True if self.mode == 'train' else False
dataset = VISDAC(root=self.root,
imbalanced=self.imb,
domain=domain,
train=train,
transform=self.transform,
from_file=False
)
data = dataset.data
labels = dataset.labels
self.strong_augmentation = transforms.Compose(
[
transforms.RandomResizedCrop(224, scale=(0.2, 1.0), antialias=None),
transforms.RandomApply(
[transforms.ColorJitter(0.4, 0.4, 0.4, 0.1)],
p=0.8, # not strengthened
),
transforms.RandomGrayscale(p=0.2),
transforms.RandomApply([GaussianBlur([0.1, 2.0])], p=0.5),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
normalize,
]
)
self.data = data
self.labels = labels
def get_domainnet(self):
domain = self.dataset.split('/')[-1]
if self.mode == 'all':
train_set = DomainNet(root=self.root,
imbalanced=self.imb,
domain=domain,
train=True,
transform=self.transform,
from_file=False
)
test_set = DomainNet(root=self.root,
imbalanced=self.imb,
domain=domain,
train=False,
transform=self.transform,
from_file=False
)
data = np.concatenate((train_set.data, test_set.data))
labels = np.concatenate((train_set.labels, test_set.labels))
else:
train = True if self.mode == 'train' else False
dataset = DomainNet(root=self.root,
imbalanced=self.imb,
domain=domain,
train=train,
transform=self.transform,
from_file=False
)
data = dataset.data
labels = dataset.labels
self.strong_augmentation = transforms.Compose(
[
transforms.RandomResizedCrop(224, scale=(0.2, 1.0), antialias=None),
transforms.RandomApply(
[transforms.ColorJitter(0.4, 0.4, 0.4, 0.1)],
p=0.8, # not strengthened
),
transforms.RandomGrayscale(p=0.2),
transforms.RandomApply([GaussianBlur([0.1, 2.0])], p=0.5),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
normalize,
]
)
self.data = data
self.labels = labels
def load_noisy_labels(self):
idx = np.load(self.noisy_path+"_idx.npy")
labels = np.load(self.noisy_path+"_noisylab.npy")
return idx, labels
def __getitem__(self, index):
"""
Args:
index (int): Index
Returns:
tuple: (image, target) where target is index of the target class.
"""
img, target = self.data[index], self.labels[index]
noisy_target = self.noisy_labels[index] if self.noisy_path is not None else self.labels[index]
img = Image.open(img).convert('RGB')
strong_augmented = self.strong_augmentation(img)
strong_augmented2 = self.strong_augmentation(img)
weak_augmented = self.transform(img) if self.transform is not None else img
return weak_augmented, strong_augmented, target, index, noisy_target, strong_augmented2#, img
def __len__(self) -> int:
return len(self.data)