-
Notifications
You must be signed in to change notification settings - Fork 2
/
filelistdataset_splittest.py
197 lines (162 loc) · 7.77 KB
/
filelistdataset_splittest.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
import torch
from PIL import Image
from torch.utils.data import Dataset
from easydl.common.wheel import *
import numpy as np
class BaseImageDataset(Dataset):
"""
base image dataset
for image dataset, ``__getitem__`` usually reads an image from a given file path
the image is guaranteed to be in **RGB** mode
subclasses should fill ``datas`` and ``labels`` as they need.
"""
def __init__(self, transform=None, return_id=False):
self.return_id = return_id
self.transform = transform or (lambda x : x)
self.datas = []
self.labels = []
self.test_datas = []
self.test_labels = []
def __getitem__(self, index):
im = Image.open(self.datas[index]).convert('RGB')
im = self.transform(im)
if not self.return_id:
return im, self.labels[index]
return im, self.labels[index], index
def __len__(self):
return len(self.datas)
class FileListDataset_splittrain(BaseImageDataset):
"""
dataset that consists of a file which has the structure of :
image_path label_id
image_path label_id
......
i.e., each line contains an image path and a label id
"""
def __init__(self, list_path, path_prefix='', transform=None, return_id=True, num_classes=None, filter=None, num_per_class=10):
"""
:param str list_path: absolute path of image list file (which contains (path, label_id) in each line) **avoid space in path!**
:param str path_prefix: prefix to add to each line in image list to get the absolute path of image,
esp, you should set path_prefix if file path in image list file is relative path
:param int num_classes: if not specified, ``max(labels) + 1`` is used
:param int -> bool filter: filter out the data to be used
"""
super(FileListDataset_splittrain, self).__init__(transform=transform, return_id = return_id)
self.list_path = list_path
self.path_prefix = path_prefix
filter = filter or (lambda x : True)
with open(self.list_path, 'r') as f:
data = [[line.split()[0], line.split()[1] if len(line.split()) > 1 else '0'] for line in f.readlines() if
line.strip()] # avoid empty lines
self.datas = [join_path(self.path_prefix, x[0]) for x in data]
try:
self.labels = [int(x[1]) for x in data]
except ValueError as e:
print('invalid label number, maybe there is a space in the image path?')
raise e
classes = np.unique(self.labels)
count = np.zeros((len(classes), ))
i = 0
while(i < len(self.datas)):
count[self.labels[i]] += 1
if count[self.labels[i]] > num_per_class:
self.datas.pop(i)
self.labels.pop(i)
i = i - 1
i = i + 1
ans = [(x, y) for (x, y) in zip(self.datas, self.labels) if filter(y)]
self.datas, self.labels = zip(*ans)
self.num_classes = num_classes or max(self.labels) + 1
class FileListDataset_splittest(BaseImageDataset):
"""
dataset that consists of a file which has the structure of :
image_path label_id
image_path label_id
......
i.e., each line contains an image path and a label id
"""
def __init__(self, list_path, path_prefix='', transform=None, return_id=True, num_classes=None, filter=None, num_per_class=10):
"""
:param str list_path: absolute path of image list file (which contains (path, label_id) in each line) **avoid space in path!**
:param str path_prefix: prefix to add to each line in image list to get the absolute path of image,
esp, you should set path_prefix if file path in image list file is relative path
:param int num_classes: if not specified, ``max(labels) + 1`` is used
:param int -> bool filter: filter out the data to be used
"""
super(FileListDataset_splittest, self).__init__(transform=transform, return_id = return_id)
self.list_path = list_path
self.path_prefix = path_prefix
filter = filter or (lambda x : True)
with open(self.list_path, 'r') as f:
data = [[line.split()[0], line.split()[1] if len(line.split()) > 1 else '0'] for line in f.readlines() if
line.strip()] # avoid empty lines
self.datas = [join_path(self.path_prefix, x[0]) for x in data]
try:
self.labels = [int(x[1]) for x in data]
except ValueError as e:
print('invalid label number, maybe there is a space in the image path?')
raise e
classes = np.unique(self.labels)
count = np.zeros((len(classes), ))
i = 0
while(i < len(self.datas)):
if count[self.labels[i]] < num_per_class:
count[self.labels[i]] += 1
self.datas.pop(i)
self.labels.pop(i)
i = i - 1
i = i + 1
i = 0
count = np.zeros((len(classes), ))
while(i < len(self.datas)):
count[self.labels[i]] += 1
if count[self.labels[i]] > num_per_class:
self.datas.pop(i)
self.labels.pop(i)
i = i - 1
i = i + 1
ans = [(x, y) for (x, y) in zip(self.datas, self.labels) if filter(y)]
self.datas, self.labels = zip(*ans)
self.num_classes = num_classes or max(self.labels) + 1
class FileListDataset_unlabeled(BaseImageDataset):
"""
dataset that consists of a file which has the structure of :
image_path label_id
image_path label_id
......
i.e., each line contains an image path and a label id
"""
def __init__(self, list_path, path_prefix='', transform=None, return_id=False, num_classes=None, filter=None, num_per_class=10):
"""
:param str list_path: absolute path of image list file (which contains (path, label_id) in each line) **avoid space in path!**
:param str path_prefix: prefix to add to each line in image list to get the absolute path of image,
esp, you should set path_prefix if file path in image list file is relative path
:param int num_classes: if not specified, ``max(labels) + 1`` is used
:param int -> bool filter: filter out the data to be used
"""
super(FileListDataset_unlabeled, self).__init__(transform=transform, return_id = return_id)
self.list_path = list_path
self.path_prefix = path_prefix
filter = filter or (lambda x : True)
with open(self.list_path, 'r') as f:
data = [[line.split()[0], line.split()[1] if len(line.split()) > 1 else '0'] for line in f.readlines() if
line.strip()] # avoid empty lines
self.datas = [join_path(self.path_prefix, x[0]) for x in data]
try:
self.labels = [int(x[1]) for x in data]
except ValueError as e:
print('invalid label number, maybe there is a space in the image path?')
raise e
classes = np.unique(self.labels)
count = np.zeros((len(classes), ))
i = 0
while(i < len(self.datas)):
count[self.labels[i]] += 1
if count[self.labels[i]] > num_per_class:
self.datas.pop(i)
self.labels.pop(i)
i = i - 1
i = i + 1
ans = [(x, y) for (x, y) in zip(self.datas, self.labels) if filter(y)]
self.datas, self.labels = zip(*ans)
self.num_classes = num_classes or max(self.labels) + 1