-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdata_loader.py
238 lines (219 loc) · 12.2 KB
/
data_loader.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
# Helper functions to load data for in-distribution and out-of-distribution datasets
import numpy as np
import torch
import torchvision.transforms as transforms
import torchvision.datasets as dset
import heart_loader as heart
def get_iid_loader(database, subset_size_train=-1, subset_size_valid=-1, outlier_exposure=True, seed=2020):
if database == 'cifar10':
# mean and standard deviation of channels of CIFAR-10 images
mean = [x / 255 for x in [125.3, 123.0, 113.9]]
std = [x / 255 for x in [63.0, 62.1, 66.7]]
train_set, train_loader, _ = _cifar10_loader(mean, std, subset_size_train)
valid_set, valid_loader, classes = _cifar10_loader(mean, std, subset_size_valid, mode='valid')
class_weights = np.ones(10)
if database == 'heart':
mean = [0.122, 0.122, 0.122]
std = [0.184, 0.184, 0.184]
train_set, train_loader, classes, class_weights, qual_c, qual_w = _heart_loader(mean, std, subset_size_train, 'train', outlier_exposure)
valid_set, valid_loader, _, _, _, _ = _heart_loader(mean, std, subset_size_valid, 'valid', outlier_exposure)
# if database == 'lung':
# mean = [0.185, 0.185, 0.185]
# std = [0.196, 0.196, 0.196]
# out_set, out_loader = _image_folder_loader("../data/lung_image_dataset", mean, std, subset_size_train)
if database == 'ob':
mean = [0.136, 0.136, 0.136]
std = [0.193, 0.193, 0.193]
train_set, train_loader, classes, class_weights = _ob_loader(mean, std, subset_size_train, 'train')
valid_set, valid_loader, _, _ = _ob_loader(mean, std, subset_size_valid, 'valid')
#class_weights = np.ones(5)
return train_set, valid_set, train_loader, valid_loader, classes, class_weights#, qual_c, qual_w
def get_ood_loader(database, subset_size, target_mean, target_std, seed=2020):
# extracts images to be trained alongside the training set with the IID data
# if we'd like to use this for training/test of outlier detection, need to postprocess the batch of labels
# for training refinement, requires mean/std of IID dataset for whitening
if database == 'cifar':
out_set, out_loader, _ = _cifar10_loader(target_mean, target_std, subset_size)
if database == 'svhn':
out_set, out_loader = _svhn_ood_loader(target_mean, target_std, subset_size)
if database == 'dtd':
out_set, out_loader = _image_folder_loader("../data/dtd/images", target_mean, target_std, subset_size)
#_dtd_loader(target_mean, target_std, subset_size)
if database == 'imagenet200':
out_set, out_loader = _image_folder_loader("../data/tiny-imagenet-200/train", target_mean, target_std, subset_size)
#_imagenet200_loader(target_mean, target_std, subset_size)
if database == 'heart':
out_set, out_loader, _, _, _, _ = _heart_loader(target_mean, target_std, subset_size, 'train')
if database == 'heart_valid':
out_set, out_loader, _, _, _, _ = _heart_loader(target_mean, target_std, subset_size, 'valid')
if database == 'heart_test':
out_set, out_loader, _, _, _, _ = _heart_loader(target_mean, target_std, subset_size, 'test')
if database == 'lung':
out_set, out_loader = _image_folder_loader("../data/lung_image_dataset", target_mean, target_std, subset_size)
#_lung_loader(target_mean, target_std, subset_size)
if database == 'ob_test':
out_set, out_loader, _, _ = _ob_loader(target_mean, target_std, subset_size, 'test')
return out_set, out_loader
def _cifar10_loader(mean, std, subset_size, mode='train'):
# if mode == 'train':
# transform = transforms.Compose(
# [transforms.Resize(256),
# transforms.RandomCrop(224),
# transforms.Grayscale(num_output_channels=3),
# transforms.ToTensor(),
# transforms.Normalize(mean, std)])
# dataset = dset.CIFAR10(root='../data/cifarpy', train=True,
# download=False, transform=transform)
# else:
# transform = transforms.Compose(
# [transforms.Resize(256),
# transforms.CenterCrop(224),
# transforms.Grayscale(num_output_channels=3),
# transforms.ToTensor(),
# transforms.Normalize(mean, std)])
# dataset = dset.CIFAR10(root='../data/cifarpy', train=False,
# download=False, transform=transform)
# if mode == 'train':
# transform = _get_transform('train')
# dataset = dset.CIFAR10(root='../data/cifarpy', train=True,
# download=False, transform=transform)
# else:
# transform = _get_transform('valid')
# dataset = dset.CIFAR10(root='../data/cifarpy', train=False,
# download=False, transform=transform)
transform = _get_transform(mean, std, mode)
dataset = dset.CIFAR10(root='../data/cifarpy', train=(mode=='train'),
download=False, transform=transform)
dataset_1 = _take_subset(dataset, subset_size)
data_loader = torch.utils.data.DataLoader(dataset_1, batch_size=4,
shuffle=True, num_workers=2)
classes = ('plane', 'car', 'bird', 'cat',
'deer', 'dog', 'frog', 'horse', 'ship', 'truck')
num_classes = 10
return dataset_1, data_loader, classes
def _ob_loader(mean, std, subset_size, mode='train', outlier_exposure=True):
transform = _get_transform(mean, std, mode, to_pil_image=True)
dataset = ob.OB("../data/ob_image_dataset", transform, mode)
classes = dataset.get_classes()
class_weights = dataset.get_class_weights()
dataset_1 = _take_subset(dataset, subset_size)
# data_loader = torch.utils.data.DataLoader(dataset_1, batch_size=4,
# shuffle=True, num_workers=2)
data_loader = torch.utils.data.DataLoader(dataset_1, batch_size=4,
shuffle=True, num_workers=2)
return dataset, data_loader, classes, class_weights
def _heart_loader(mean, std, subset_size, mode='train', outlier_exposure=False):
transform = _get_transform(mean, std, mode, to_pil_image=True)
if mode == 'train':
dataset = heart.Heart('../database_path/train_labels_2.txt', transform, outlier_exposure)
elif mode == 'valid':
dataset = heart.Heart('../database_path/valid_labels_2.txt', transform, outlier_exposure)
elif mode == 'test':
dataset = heart.Heart('../database_path/test_labels_2.txt', transform, outlier_exposure)
view_c, qual_c = dataset.get_classes()
view_w, qual_w = dataset.get_label_sampling_weights()
view_w_per_item = dataset.get_label_sampling_weights_per_item()
#dataset_1 = _take_subset(dataset, subset_size)
# data_loader = torch.utils.data.DataLoader(dataset_1, batch_size=4,
# shuffle=True, num_workers=2)
bs = 4
if (subset_size == -1):
if mode=='train':
samp = torch.utils.data.WeightedRandomSampler(view_w_per_item, num_samples=len(dataset) )
else:
samp = torch.utils.data.SequentialSampler(dataset)
else:
if mode=='train':
samp = torch.utils.data.WeightedRandomSampler(view_w_per_item, num_samples=subset_size)
else:
samp = torch.utils.data.RandomSampler(dataset, replacement=True, num_samples=subset_size)
data_loader = torch.utils.data.DataLoader(dataset, batch_size=bs, sampler=samp, num_workers=2)
return dataset, data_loader, view_c, view_w, qual_c, qual_w
def _svhn_ood_loader(mean, std, subset_size):
transform = _get_transform(mean, std, 'valid')
dataset = svhn.SVHN(root='../data/svhn/', split="test", transform=transform, download=False)
dataset_1 = _take_subset(dataset, subset_size)
data_loader = torch.utils.data.DataLoader(dataset_1, batch_size=4,
shuffle=True, num_workers=2)
return dataset_1, data_loader
def _image_folder_loader(folder_name, mean, std, subset_size):
transform = _get_transform(mean, std, 'valid')
dataset = dset.ImageFolder(root=folder_name,
transform=transform)
dataset_1 = _take_subset(dataset, subset_size)
data_loader = torch.utils.data.DataLoader(dataset_1, batch_size=4, shuffle=True,
num_workers=2)
return dataset_1, data_loader
# def _dtd_loader(mean, std, subset_size):
# transform = transforms.Compose(
# [transforms.Resize((224, 224)),
# transforms.Grayscale(num_output_channels=3),
# transforms.ToTensor(),
# transforms.Normalize(mean, std)])
# dataset = dset.ImageFolder(root="../data/dtd/images",
# transform=transform)
# dataset_1 = _take_subset(dataset, subset_size)
# data_loader = torch.utils.data.DataLoader(dataset_1, batch_size=4, shuffle=True,
# num_workers=2)
# return dataset_1, data_loader
# def _imagenet200_loader(mean, std, subset_size):
# transform = transforms.Compose(
# [transforms.Resize(224),
# transforms.Grayscale(num_output_channels=3),
# transforms.ToTensor(),
# transforms.Normalize(mean, std)])
# dataset = dset.ImageFolder(root="../data/tiny-imagenet-200/train",transform=transform)
# dataset_1 = _take_subset(dataset, subset_size)
# data_loader = torch.utils.data.DataLoader(dataset_1, batch_size=4, shuffle=True,
# num_workers=2)
# return dataset_1, data_loader
# def _lung_loader(mean, std, subset_size):
# transform = transforms.Compose(
# [transforms.Resize((224,224)),
# transforms.Grayscale(num_output_channels=3),
# transforms.ToTensor(),
# transforms.Normalize(mean, std)])
# dataset = dset.ImageFolder(root="../data/lung_image_dataset",transform=transform)
# dataset_1 = _take_subset(dataset, subset_size)
# data_loader = torch.utils.data.DataLoader(dataset_1, batch_size=4, shuffle=True,
# num_workers=2)
# return dataset_1, data_loader
def _get_transform(mean, std, mode='valid', to_pil_image=False):
# if mode=='train':
# t_transform = transforms.Compose(
# [transforms.Resize((256,256)),
# transforms.RandomRotation(5),
# transforms.RandomCrop((224,224)),
# transforms.Grayscale(num_output_channels=3),
# transforms.ToTensor()])#,
# #transforms.Normalize(mean, std)])
# else: # mode=='valid'
# t_transform = transforms.Compose(
# [transforms.Resize((224,224)),
# transforms.Grayscale(num_output_channels=3),
# transforms.ToTensor()])#,
# #transforms.Normalize(mean, std)])
if mode=='train':
aug_transform = transforms.Compose(
[transforms.Resize((256,256)),
transforms.RandomRotation(5),
transforms.RandomCrop((224,224)) ])
else: # mode=='valid'
aug_transform = transforms.Resize((224,224))
t_transform = transforms.Compose(
[aug_transform,
transforms.Grayscale(num_output_channels=3),
transforms.ToTensor(),
transforms.Normalize(mean, std)
])
if to_pil_image:
t_transform = transforms.Compose([transforms.ToPILImage(),
t_transform])
return t_transform
def _take_subset(dataset, subset_size):
np.random.seed(2020)
if (subset_size == -1):
dataset_1 = dataset # use full dataset
else: # use subset, if subset_size > dataset_size, this will implement oversampling
dataset_1 = torch.utils.data.Subset(dataset, np.random.choice(len(dataset), subset_size) )
return dataset_1