-
Notifications
You must be signed in to change notification settings - Fork 5
/
dataset.py
138 lines (113 loc) · 5.19 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
import torch
import torch.utils.data as torchdata
import torchvision
import torchvision.transforms as transforms
import numpy as np
import os
class CIFAR10(torchdata.Dataset):
def __init__(self, dataroot, target, imagesize, train):
self.target = target
self.train = train
if imagesize is None:
transform = transforms.Compose([
transforms.ToTensor(), # first, convert image to PyTorch tensor [0.,1.]
])
else:
transform = transforms.Compose([
transforms.Resize(imagesize),
transforms.ToTensor(), # first, convert image to PyTorch tensor [0.,1.]
])
self.raw_dataset = torchvision.datasets.CIFAR10(root=dataroot, train = train, transform=transform, download = True )
targets = np.asarray(self.raw_dataset.targets)
if self.train:
self.idxs = np.where(targets == target)[0]
else: # test
self.idxs = np.arange(targets.size)
def __len__(self):
return self.idxs.size
def __getitem__(self, i):
idx = self.idxs[i]
data = self.raw_dataset[idx][0]
min, max = data.min(), data.max()
data = (data-min)/(max-min)
if self.train: # when training, label is not considered
label = torch.LongTensor([1])
else: # if label is one, it means it is anomalous
label = torch.LongTensor([self.raw_dataset.targets[idx]!=self.target])
return data,label
class MNIST(torchdata.Dataset):
def __init__(self, dataroot, target, imagesize, train):
self.target = target
self.train = train
if imagesize is None:
transform = transforms.Compose([
transforms.ToTensor(), # first, convert image to PyTorch tensor [0.,1.]
])
else:
transform = transforms.Compose([
transforms.Resize(imagesize),
transforms.ToTensor(), # first, convert image to PyTorch tensor [0.,1.]
])
self.raw_dataset = torchvision.datasets.MNIST(root=dataroot, train = train, transform=transform, download = True)
targets = self.raw_dataset.targets.numpy()
if self.train:
self.idxs = np.where(targets == target)[0]
else: # test
self.idxs = np.arange(targets.size)
def __len__(self):
return self.idxs.size
def __getitem__(self, i):
idx = self.idxs[i]
data = self.raw_dataset[idx][0]
if self.train: # when training, label is not considered
label = torch.LongTensor([1])
else: # if label is one, it means it is anomalous
label = torch.LongTensor([self.raw_dataset.targets[idx]!=self.target])
return data,label
class FMNIST(torchdata.Dataset):
def __init__(self, dataroot, target, imagesize, train):
self.target = target
self.train = train
if imagesize is None:
transform = transforms.Compose([
transforms.ToTensor(), # first, convert image to PyTorch tensor [0.,1.]
])
else:
transform = transforms.Compose([
transforms.Resize(imagesize),
transforms.ToTensor(), # first, convert image to PyTorch tensor [0.,1.]
])
self.raw_dataset = torchvision.datasets.FashionMNIST(root=dataroot, train = train, transform=transform, download = True)
targets = self.raw_dataset.targets.numpy()
if self.train:
self.idxs = np.where(targets == target)[0]
else: # test
self.idxs = np.arange(targets.size)
def __len__(self):
return self.idxs.size
def __getitem__(self, i):
idx = self.idxs[i]
data = self.raw_dataset[idx][0]
if self.train: # when training, label is not considered
label = torch.LongTensor([1])
else: # if label is one, it means it is anomalous
label = torch.LongTensor([self.raw_dataset.targets[idx]!=self.target])
return data,label
def get_trainloader(datatype, dataroot, target, batchsize, nworkers, imagesize = None):
if datatype.lower() in ['mnist']:
traindataset = MNIST(dataroot, target, imagesize = imagesize, train=True)
elif datatype.lower() in ['cifar10']:
traindataset = CIFAR10(dataroot, target, imagesize = imagesize, train=True)
elif datatype.lower() in ['fmnist']:
traindataset = FMNIST(dataroot, target, imagesize = imagesize, train=True)
trainloader = torchdata.DataLoader(traindataset, batch_size = batchsize, shuffle = True, num_workers = nworkers)
return trainloader
def get_testloader(datatype, dataroot, target, batchsize, nworkers, imagesize = None):
if datatype.lower() in ['mnist']:
testdataset = MNIST(dataroot, target, imagesize = imagesize, train=False)
elif datatype.lower() in ['cifar10']:
testdataset = CIFAR10(dataroot, target, imagesize = imagesize, train=False)
elif datatype.lower() in ['fmnist']:
testdataset = FMNIST(dataroot, target, imagesize = imagesize, train=False)
testloader = torchdata.DataLoader(testdataset, batch_size = batchsize, shuffle = False, num_workers = nworkers)
return testloader