-
Notifications
You must be signed in to change notification settings - Fork 0
/
load_data.py
171 lines (140 loc) · 5.98 KB
/
load_data.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
import os
from torchvision import transforms
from loaders.data_list import Imagelists_VISDA, return_classlist, Imagelists_VISDA_twice
import torch
class ResizeImage():
def __init__(self, size):
if isinstance(size, int):
self.size = (int(size), int(size))
else:
self.size = size
def __call__(self, img):
th, tw = self.size
return img.resize((th, tw))
class TransformTwice:
def __init__(self, transform):
self.transform = transform
def __call__(self, inp):
out1 = self.transform(inp)
out2 = self.transform(inp)
return out1, out2
def return_dataset(args):
base_path = args.exp_dir+'/data_list'
if args.dataset == 'office':
root = ''
else:
root = '../dataset/%s' % args.dataset
src_img_pth_file = os.path.join(base_path, args.source + '_img.txt')
src_label_pth_file = os.path.join(base_path, args.source + '_label.txt')
trg_img_pth_file_unl = os.path.join(base_path, args.target + '_img.txt')
trg_label_pth_file_unl = os.path.join(base_path, args.target + '_label.txt')
crop_size = 224
data_transforms = {
'train': transforms.Compose([
ResizeImage(256),
transforms.RandomHorizontalFlip(),
transforms.RandomCrop(crop_size),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
'val': transforms.Compose([
ResizeImage(256),
transforms.RandomHorizontalFlip(),
transforms.RandomCrop(crop_size),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
'test': transforms.Compose([
ResizeImage(256),
transforms.CenterCrop(crop_size),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
}
source_dataset = Imagelists_VISDA(src_img_pth_file, src_label_pth_file, root=root,
transform=data_transforms['train'])
target_dataset_unl = Imagelists_VISDA_twice(trg_img_pth_file_unl, trg_label_pth_file_unl, root=root,
transform=TransformTwice(data_transforms['train']))
return source_dataset, target_dataset_unl
def return_psu_dataset(args):
base_path = args.exp_dir+'/data_list'
if args.dataset == 'office':
root = ''
else:
root = '../dataset/%s'%args.dataset
trg_img_pth_file_psu = os.path.join(base_path,
'pesudo_target_images_' + args.target + '_img.txt')
trg_label_pth_file_psu = os.path.join(base_path,
'pesudo_target_images_' + args.target + '_label.txt')
crop_size = 224
data_transforms = {
'train': transforms.Compose([
ResizeImage(256),
transforms.RandomHorizontalFlip(),
transforms.RandomCrop(crop_size),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
'val': transforms.Compose([
ResizeImage(256),
transforms.RandomHorizontalFlip(),
transforms.RandomCrop(crop_size),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
'test': transforms.Compose([
ResizeImage(256),
transforms.CenterCrop(crop_size),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
}
target_dataset_psu = Imagelists_VISDA(trg_img_pth_file_psu, trg_label_pth_file_psu, root=root,
transform=data_transforms['train'])
return target_dataset_psu
def retrun_test_dataset(args):
base_path = args.exp_dir+'/data_list'
if args.dataset == 'office':
root = ''
else:
root = '../dataset/%s' % args.dataset
src_img_pth_file = os.path.join(base_path, args.source + '_img.txt')
src_label_pth_file = os.path.join(base_path, args.source + '_label.txt')
trg_img_pth_file_unl = os.path.join(base_path, args.target + '_img.txt')
trg_label_pth_file_unl = os.path.join(base_path, args.target + '_label.txt')
crop_size = 224
data_transforms = {
'train': transforms.Compose([
ResizeImage(256),
transforms.RandomHorizontalFlip(),
transforms.RandomCrop(crop_size),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
'val': transforms.Compose([
ResizeImage(256),
transforms.RandomHorizontalFlip(),
transforms.RandomCrop(crop_size),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
'test': transforms.Compose([
ResizeImage(256),
transforms.CenterCrop(crop_size),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
}
source_dataset = Imagelists_VISDA(src_img_pth_file, src_label_pth_file, root=root,
transform=data_transforms['test'])
target_dataset_unl = Imagelists_VISDA(trg_img_pth_file_unl, trg_label_pth_file_unl, root=root,
transform=data_transforms['test'])
return source_dataset, target_dataset_unl
def per_image_standardization(x):
y = x.view(-1, x.shape[1]*x.shape[2]*x.shape[3])
mean = y.mean(dim=1, keepdim = True).expand_as(y)
std = y.std(dim=1, keepdim = True).expand_as(y)
adjusted_std = torch.max(std, 1.0/torch.sqrt(torch.cuda.FloatTensor([x.shape[1]*x.shape[2]*x.shape[3]])))
y = (y- mean)/ adjusted_std
standarized_input = y.view(x.shape[0],x.shape[1],x.shape[2],x.shape[3])
return standarized_input