-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathdata.py
284 lines (228 loc) · 10.7 KB
/
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
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
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
import glob
from torch.utils.data import Dataset
import os
from PIL import Image
import numpy as np
# import h5py
import random
def dataset(cfgs, flag, trans):
if cfgs['dataset'] == 'BSDS':
dataset = BSDS_500(root=cfgs['dataset_path'], flag=flag, VOC=False, transform=trans)
elif cfgs['dataset'] == 'BSDS-VOC':
dataset = BSDS_500(root=cfgs['dataset_path'], flag=flag, VOC=True, transform=trans)
elif cfgs['dataset'] == 'NYUD-image':
dataset = NYUD(root=cfgs['dataset_path'], flag=flag, rgb=True, transform=trans)
elif cfgs['dataset'] == 'NYUD-hha':
dataset = NYUD(root=cfgs['dataset_path'], flag=flag, rgb=False, transform=trans)
elif cfgs['dataset'] == 'MultiCue-Edge':
dataset = MultiCue(root=cfgs['dataset_path'], flag=flag, edge=True, transform=trans, seq=cfgs['multicue_seq'])
elif cfgs['dataset'] == 'MultiCue-Contour':
dataset = MultiCue(root=cfgs['dataset_path'], flag=flag, edge=False, transform=trans, seq=cfgs['multicue_seq'])
elif cfgs['dataset'] == 'PASCAL-VOC-12':
dataset = PASCAL_VOC12(root=cfgs['dataset_path'], flag=flag, transform=trans)
elif cfgs['dataset'] == 'PASCAL-Context':
dataset = PASCAL_Context(root=cfgs['dataset_path'], flag=flag, transform=trans)
else:
raise NameError
return dataset
'''
#######################################################################################
BSDS_500
#######################################################################################
'''
def read_from_pair_txt(path, filename):
path = path.replace('/', os.sep)
pathfile = open(os.path.join(path, filename)) # 打开这个这个文件
filenames = pathfile.readlines()
pathfile.close()
filenames = [f.strip() for f in filenames] # 去除每行首尾空格，构成一个表格
filenames = [c.split(' ') for c in filenames] # 按中间空格把一个元素分成两个元素数组
filenames = [(os.path.join(path, c[0].replace('/', os.sep)), # 更改为esp兼容linux文件系统
os.path.join(path, c[1].replace('/', os.sep))) for c in filenames] # 给每个相对路径变为绝对路径，
return filenames
class BSDS_500(Dataset):
def __init__(self, root, flag='train', VOC=False, transform=None):
if flag == 'train':
if VOC:
filenames = read_from_pair_txt(root['BSDS-VOC'], 'bsds_pascal_train_pair_s5.lst')
else:
filenames = read_from_pair_txt(root['BSDS'], 'train_pair.lst')
self.im_list = [im_name[0] for im_name in filenames]
self.gt_list = [im_name[1] for im_name in filenames]
elif flag == 'test':
self.im_list = glob.glob(os.path.join(root['BSDS'], r'test/*.jpg'))
self.gt_list = [path.split(os.sep)[-1][:-4] for path in self.im_list]
self.length = self.im_list.__len__()
self.transform = transform
self.flag = flag
def __len__(self):
return self.length
def __getitem__(self, item):
image = Image.open(self.im_list[item])
if self.flag == 'train':
label = np.array(Image.open(self.gt_list[item]).convert('L'))
label = Image.fromarray(label.astype(np.float32) / 255.0)
elif self.flag == 'test':
label = Image.open(self.im_list[item])
sample = {'images': image, 'labels': label}
if self.transform:
sample = self.transform(sample)
return sample
'''
#######################################################################################
NYUD-v2
#######################################################################################
'''
class NYUD(Dataset):
def __init__(self, root, flag='train', rgb=True, transform=None):
if flag == 'train':
if rgb:
filenames = read_from_pair_txt(root['NYUD-V2'], 'image-train.lst')
else:
filenames = read_from_pair_txt(root['NYUD-V2'], 'hha-train.lst')
self.im_list = [im_name[0] for im_name in filenames]
self.gt_list = [im_name[1] for im_name in filenames]
elif flag == 'test':
if rgb:
self.im_list = glob.glob(os.path.join(root['NYUD-V2'], 'test/Images/*.png'))
else:
self.im_list = glob.glob(os.path.join(root['NYUD-V2'], 'test/HHA/*.png'))
self.gt_list = [path.split('/')[-1][:-4] for path in self.im_list]
self.length = self.im_list.__len__()
self.transform = transform
self.flag = flag
def __len__(self):
return self.length
def __getitem__(self, item):
image = Image.open(self.im_list[item])
if self.flag == 'train':
label = np.array(Image.open(self.gt_list[item]).convert('L'))
label = Image.fromarray(label.astype(np.float32) / 255.0)
elif self.flag == 'test':
label = Image.open(self.im_list[item])
sample = {'images': image, 'labels': label}
if self.transform:
sample = self.transform(sample)
return sample
'''
#######################################################################################
MultiCue
#######################################################################################
'''
class MultiCue(Dataset):
def __init__(self, root, flag='train', edge=True, transform=None, seq=1):
filenames = read_from_pair_txt(root['multicue'], 'seq' + str(seq) + '.txt')
self.im_list = [im_name[0] for im_name in filenames]
self.gt_list = [im_name[1] for im_name in filenames]
if flag == 'train':
self.im_list = self.im_list[:80]
self.gt_list = self.gt_list[:80]
elif flag == 'test':
self.im_list = self.im_list[80:]
self.gt_list = [gt.split('/')[-1].split('.')[0] for gt in self.gt_list[80:]]
self.length = self.im_list.__len__()
self.transform = transform
self.flag = flag
self.edge = 'edges' if edge == True else 'boundaries'
def __len__(self):
return self.length
def __getitem__(self, item):
image = Image.open(self.im_list[item])
if self.flag == 'train':
h5 = h5py.File(self.gt_list[item], 'r')
label = np.array(h5[self.edge]).astype(np.float32)
label = Image.fromarray(np.mean(label, axis=0))
h5.close()
elif self.flag == 'test':
label = Image.open(self.im_list[item])
sample = {'images': image, 'labels': label}
if self.transform:
sample = self.transform(sample)
return sample
def make_random_multicue(MultiCue_path):
lists = glob.glob(os.path.join(MultiCue_path, 'ground-truth/hdf5', '*.h5'))
lists = [pth.replace(MultiCue_path, '') for pth in lists]
for i in range(10):
random.shuffle(lists)
dset = [path.replace('ground-truth', 'images').replace('h5', 'png').replace('hdf5/', '') + ' ' + path
for path in lists]
with open(os.path.join(MultiCue_path, 'seq' + str(i + 1) + '.txt'), 'a') as file_handle:
file_handle.write('\n'.join(dset))
'''
#######################################################################################
PASCAL_VOC12
#######################################################################################
'''
class PASCAL_VOC12(Dataset):
def __init__(self, root, flag='train', transform=None):
if flag == 'train':
filenames = self.read_file(root['PASCAL-VOC12'], 'train.txt')
self.im_list = [im_name[0] for im_name in filenames]
self.gt_list = [im_name[1] for im_name in filenames]
elif flag == 'test':
filenames = self.read_file(root['PASCAL-VOC12'], 'val.txt')
self.im_list = [im_name[0] for im_name in filenames]
self.gt_list = [im_name[1].split('.')[0].split('/')[-1] for im_name in filenames]
self.length = self.im_list.__len__()
self.transform = transform
self.flag = flag
def read_file(self, path, filename):
with open(os.path.join(path, 'ImageSets', 'Segmentation', filename)) as pfile:
filenames = pfile.readlines()
img_root = os.path.join(path, 'JPEGImages')
gt_root = os.path.join(path, 'boundaries')
filenames = [(os.path.join(img_root, f.strip() + '.jpg'), os.path.join(gt_root, f.strip() + '.png')) for f in
filenames]
return filenames
def __len__(self):
return self.length
def __getitem__(self, item):
image = Image.open(self.im_list[item])
if self.flag == 'train':
label = np.array(Image.open(self.gt_list[item]).convert('L'))
label = Image.fromarray(label.astype(np.float32) / 255.0)
elif self.flag == 'test':
label = Image.open(self.im_list[item])
sample = {'images': image, 'labels': label}
if self.transform:
sample = self.transform(sample)
return sample
'''
#######################################################################################
PASCAL_Context
#######################################################################################
'''
class PASCAL_Context(Dataset):
def __init__(self, root, flag='train', transform=None):
if flag == 'train':
filenames = self.read_file(root['PASCAL-VOC12'], 'train_new.txt')
self.im_list = [im_name[0] for im_name in filenames]
self.gt_list = [im_name[1] for im_name in filenames]
elif flag == 'test':
filenames = self.read_file(root['PASCAL-VOC12'], 'test_new.txt')
self.im_list = [im_name[0] for im_name in filenames]
self.gt_list = [im_name[1].split('.')[0].split('/')[-1] for im_name in filenames]
self.length = self.im_list.__len__()
self.transform = transform
self.flag = flag
def read_file(self, path, filename):
with open(os.path.join(path, 'ImageSets', 'Context', filename)) as pfile:
filenames = pfile.readlines()
img_root = os.path.join(path, 'JPEGImages')
gt_root = os.path.join(path, 'Context_label')
filenames = [(os.path.join(img_root, f.strip() + '.jpg'), os.path.join(gt_root, f.strip() + '.png')) for f in
filenames]
return filenames
def __len__(self):
return self.length
def __getitem__(self, item):
image = Image.open(self.im_list[item])
if self.flag == 'train':
label = np.array(Image.open(self.gt_list[item]).convert('L'))
label = Image.fromarray(label.astype(np.float32) / 255.0)
elif self.flag == 'test':
label = Image.open(self.im_list[item])
sample = {'images': image, 'labels': label}
if self.transform:
sample = self.transform(sample)
return sample