-
Notifications
You must be signed in to change notification settings - Fork 1
/
bigfile.py
170 lines (128 loc) · 5.72 KB
/
bigfile.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
# coding=utf-8
import os, sys, array
import numpy as np
import util
class BigFile:
def __init__(self, datadir, bin_file="feature.bin"):
self.nr_of_images, self.ndims = list(map(int, open(os.path.join(datadir, 'shape.txt')).readline().split()))
id_file = os.path.join(datadir, "id.txt")
self.names = open(id_file, 'r').read().strip().split('\n') # 所有 video 文件名
if len(self.names) != self.nr_of_images:
self.names = open(id_file, 'r').read().strip().split(' ')
assert(len(self.names) == self.nr_of_images)
self.name2index = dict(list(zip(self.names, list(range(self.nr_of_images))))) # 给每一个文件名弄一个编号
self.binary_file = os.path.join(datadir, bin_file)
print(("[%s] %dx%d instances loaded from %s" % (self.__class__.__name__, self.nr_of_images, self.ndims, datadir)))
def readall(self, isname=True):
# requested = set(requested)
# if isname:
# index_name_array = [(self.name2index[x], x) for x in requested if x in self.name2index]
# else:
# assert(min(requested)>=0)
# assert(max(requested)<len(self.names))
# index_name_array = [(x, self.names[x]) for x in requested]
# if len(index_name_array) == 0:
# return [], []
index_name_array = [(self.name2index[x], x) for x in set(self.names) if x in self.name2index]
index_name_array.sort(key=lambda v:v[0])
sorted_index = [x[0] for x in index_name_array]
nr_of_images = len(index_name_array)
vecs = [None] * nr_of_images
offset = np.float32(1).nbytes * self.ndims
res = array.array('f')
fr = open(self.binary_file, 'rb')
fr.seek(index_name_array[0][0] * offset)
res.fromfile(fr, self.ndims)
previous = index_name_array[0][0]
for next in sorted_index[1:]:
move = (next-1-previous) * offset
#print next, move
fr.seek(move, 1)
res.fromfile(fr, self.ndims)
previous = next
fr.close()
return [x[1] for x in index_name_array], [ res[i*self.ndims:(i+1)*self.ndims].tolist() for i in range(nr_of_images) ]
def read(self, requested, isname=True):
"""
根据文件名读取文件，具体是从 bin 文件中读取numpy 矩阵，这里主要是视频名字和 feature vector
:param requested:
:param isname:
:return: 这里主要是视频名字和 feature vector
"""
requested = set(requested)
if isname:
index_name_array = [(self.name2index[x], x) for x in requested if x in self.name2index]
else:
assert(min(requested)>=0)
assert(max(requested)<len(self.names))
index_name_array = [(x, self.names[x]) for x in requested]
if len(index_name_array) == 0:
return [], []
index_name_array.sort(key=lambda v:v[0])
sorted_index = [x[0] for x in index_name_array]
nr_of_images = len(index_name_array)
vecs = [None] * nr_of_images
offset = np.float32(1).nbytes * self.ndims
res = array.array('f')
fr = open(self.binary_file, 'rb')
fr.seek(index_name_array[0][0] * offset)
res.fromfile(fr, self.ndims)
previous = index_name_array[0][0]
for next in sorted_index[1:]:
move = (next-1-previous) * offset
#print next, move
fr.seek(move, 1)
res.fromfile(fr, self.ndims)
previous = next
fr.close()
return [x[1] for x in index_name_array], [ res[i*self.ndims:(i+1)*self.ndims].tolist() for i in range(nr_of_images) ]
def read_one(self, name):
renamed, vectors = self.read([name])
return vectors[0]
def shape(self):
return [self.nr_of_images, self.ndims]
class StreamFile:
def __init__(self, datadir):
self.feat_dir = datadir
self.nr_of_images, self.ndims = list(map(int, open(os.path.join(datadir,'shape.txt')).readline().split()))
id_file = os.path.join(datadir, "id.txt")
self.names = open(id_file, 'r').read().strip().split('\n') # 所有 video 文件名
if len(self.names) != self.nr_of_images:
self.names = open(id_file, 'r').read().strip().split(' ')
assert(len(self.names) == self.nr_of_images)
self.name2index = dict(list(zip(self.names, list(range(self.nr_of_images)))))
self.binary_file = os.path.join(datadir, "feature.bin")
print(("[%s] %dx%d instances loaded from %s" % (self.__class__.__name__, self.nr_of_images, self.ndims, datadir)))
self.fr = None
self.current = 0
def open(self):
self.fr = open(os.path.join(self.feat_dir,'feature.bin'), 'rb')
self.current = 0
def close(self):
if self.fr:
self.fr.close()
self.fr = None
def __iter__(self):
return self
def __next__(self):
if self.current >= self.nr_of_images:
self.close()
raise StopIteration
else:
res = array.array('f')
res.fromfile(self.fr, self.ndims)
_id = self.names[self.current]
self.current += 1
return _id, res.tolist()
if __name__ == '__main__':
feat_dir = "/data2/hf/VisualSearch/toydata/FeatureData/f1"
bigfile = BigFile(feat_dir)
imset = str.split('b z a a b c')
renamed, vectors = bigfile.read(imset)
for name,vec in zip(renamed, vectors):
print(name, vec)
bigfile = StreamFile(feat_dir)
bigfile.open()
for name, vec in bigfile:
print(name, vec)
bigfile.close()