-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathdataset.py
293 lines (252 loc) · 11 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
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
285
286
287
288
289
290
291
292
293
from scipy.misc import imread, imresize
# import cv2
import os
import numpy as np
import cPickle
import time
from progressbar import ProgressBar
import tensorflow as tf
class imdb(object):
def __init__(self, name, size_image=64, image_value_range=(-1,1), num_categories=10):
self._name = name
self._root_dir = None
self._train_ind = None
self._test_ind = None
self._size_image = size_image
self._image_value_range = image_value_range
self._num_categories = num_categories
@property
def name(self):
return self._name
@property
def train_ind(self):
return self._train_ind
@property
def test_ind(self):
return self._test_ind
@property
def size_image(self):
return self._size_image
@property
def image_value_range(self):
return self._image_value_range
@property
def num_categories(self):
return self._num_categories
def load_image(
self,
image_path,
image_size=64,
image_value_range=(-1, 1),
is_gray=False,
):
if is_gray:
# image = cv2.imread(os.path.join('./data', image_path), 0).astype(np.float32)
image = imread(os.path.join('./data', image_path), mode='L').astype(np.float32)
else:
# image = cv2.imread(os.path.join('./data', image_path), 1).astype(np.float32)
image = imread(os.path.join('./data', image_path), mode='RGB').astype(np.float32)
# im_scale_x = float(image_size) / float(image.shape[1])
# im_scale_y = float(image_size) / float(image.shape[0])
# image = cv2.resize(image, None, None, fx=im_scale_x, fy=im_scale_y,
# interpolation=cv2.INTER_LINEAR)
image = imresize(image, [image_size, image_size])
image = image.astype(np.float32) * (image_value_range[-1] - image_value_range[0]) / 255.0 + image_value_range[0]
return image
# class UTKFace(imdb):
# def __init__(self, size_image=64, image_value_range=(-1,1), num_categories=10, in_memory=False):
# super(UTKFace, self).__init__('UTKFace', size_image, image_value_range, num_categories)
# self._root_dir = os.path.join('./data', self.name)
# self._info = self._load_info() ## file_name in order
# self._namelist_ind = self._load_ind() ## ind -> filename
# self._train_ind = self._namelist_ind[:-100]
# self._test_ind = self._namelist_ind[-100:]
# self._save_test_pkl()
# self._in_memory = in_memory
# if self._in_memory:
# print('\tLoading into memory ...')
# self._imgs = self._load_imgs() ## uint8
# self._name2ind = self._get_name2ind()
# def _load_info(self):
# info_path = sorted(os.listdir(self._root_dir))
# return [os.path.join(self.name, item) for item in info_path]
# def _load_ind(self):
# ind = range(len(self._info))
# np.random.shuffle(ind)
# return ind
# def _load_imgs(self):
# ret = []
# pbar = ProgressBar(maxval=len(self._info))
# pbar.start()
# i = 0
# for i, item in enumerate(self._info):
# ret.append(imread(os.path.join('./data', item), mode='RGB'))
# pbar.update(i)
# pbar.finish()
# return ret
# def _get_name2ind(self):
# ret = {}
# for i in range(len(self._info)):
# ret[self._info[i]] = i
# return ret
# def _save_test_pkl(self):
# cache_file = os.path.join('./data', 'UTK_test_files.pkl')
# with open(cache_file, 'wb') as fid:
# cPickle.dump(self._test_ind, fid, cPickle.HIGHEST_PROTOCOL)
# def get_dataset(self, sample_inds):
# if self._in_memory:
# sample = []
# for sample_ind in sample_inds:
# image = self._imgs[self._name2ind[self._info[sample_ind]]].astype(np.float32)
# image = imresize(image, [self.size_image, self.size_image])
# image = image.astype(np.float32) * (self.image_value_range[-1] - self.image_value_range[0]) / 255.0 + self.image_value_range[0]
# sample.append(image)
# else:
# sample = [self.load_image(
# image_path=self._info[sample_ind],
# image_size=self.size_image,
# image_value_range=self.image_value_range,
# is_gray=False,
# ) for sample_ind in sample_inds]
# sample_images = np.array(sample).astype(np.float32)
# sample_label_age = np.ones(
# shape=(len(sample_inds), self.num_categories),
# dtype=np.float32
# ) * self.image_value_range[0]
# sample_label_gender = np.ones(
# shape=(len(sample_inds), 2),
# dtype=np.float32
# ) * self.image_value_range[0]
# for i, ind in enumerate(sample_inds):
# sample_file = self._info[ind]
# label = int(str(sample_file).split('/')[-1].split('_')[0])
# if 0 <= label <= 5:
# label = 0
# elif 6 <= label <= 10:
# label = 1
# elif 11 <= label <= 15:
# label = 2
# elif 16 <= label <= 20:
# label = 3
# elif 21 <= label <= 30:
# label = 4
# elif 31 <= label <= 40:
# label = 5
# elif 41 <= label <= 50:
# label = 6
# elif 51 <= label <= 60:
# label = 7
# elif 61 <= label <= 70:
# label = 8
# else:
# label = 9
# sample_label_age[i, label] = self.image_value_range[-1]
# gender = int(str(sample_file).split('/')[-1].split('_')[1])
# sample_label_gender[i, gender] = self.image_value_range[-1]
# return sample_images, sample_label_age, sample_label_gender
class UTKFace(object):
def __init__(self, size_image=64, num_categories=10, data_path=None, batch_size=100, num_epochs=300, shuffle_buffer=500, prefetch_buffer_size=100 , map_parallel=1 ,):
self._size_image = size_image
self._num_categories = num_categories
self._batch_size = batch_size
self._num_epochs = num_epochs
self._data_path = os.path.join('data', 'UTKFace_16_tfrecords') if not data_path else data_path
self._shuffle_buffer = shuffle_buffer
self._prefetch_buffer_size = prefetch_buffer_size
self._map_parallel = map_parallel
def parse_fn(self, serial_exmp):
feats = tf.parse_single_example(serial_exmp, features={
'image':tf.FixedLenFeature([], tf.string),
'age':tf.FixedLenFeature([], tf.int64),
'gender':tf.FixedLenFeature([], tf.int64),
})
image = tf.decode_raw(feats['image'], tf.uint8)
image = tf.reshape(image, [128, 128, 3])
image = tf.cast(image, tf.float32)
image = tf.image.resize_images(image, [self._size_image, self._size_image])
image = image / 127.5 - 1.0
age = tf.one_hot(feats['age'], self._num_categories, off_value=-1.0, axis=-1)
gender = tf.one_hot(feats['gender'], 2, off_value=-1.0, axis=-1)
return image, age, gender
def input_pipeline_new(self, tfrecords_list, shuffle=False, prefetch=False):
dataset = tf.data.TFRecordDataset(tfrecords_list)
if prefetch:
dataset = dataset.prefetch(buffer_size=self._prefetch_buffer_size)
if shuffle:
dataset = dataset.apply(tf.contrib.data.shuffle_and_repeat(buffer_size=self._shuffle_buffer, count=self._num_epochs))
dataset = dataset.apply(tf.contrib.data.map_and_batch(
map_func=self.parse_fn,
batch_size=self._batch_size,
num_parallel_batches=self._map_parallel,
drop_remainder=True))
if prefetch:
dataset = dataset.prefetch(buffer_size=tf.contrib.data.AUTOTUNE)
return dataset
def get_dataset(self):
tfrecords_list = sorted(os.listdir(self._data_path))
tfrecords_list = [os.path.join(self._data_path, item) for item in tfrecords_list]
train_dataset = self.input_pipeline_new(tfrecords_list[:-1], shuffle=True, prefetch=True)
test_dataset = self.input_pipeline_new(tfrecords_list[-1], shuffle=False, prefetch=False)
return train_dataset, test_dataset
class IMDBWIKI(imdb):
def __init__(self, size_image=64, image_value_range=(-1,1), num_categories=6):
super(IMDBWIKI, self).__init__('IMDBWIKI', size_image, image_value_range, num_categories)
self._info = self._load_info() ## age, path, gender
self._namelist_ind = self._load_namelist_ind() ## ind_list instead of filename_list
self._train_ind = self._namelist_ind[:100000]
self._test_ind = self._namelist_ind[-100:]
self._save_test_pkl()
def _load_info(self):
info_path = os.path.join('./data', 'imdb_wiki.pkl')
if os.path.exists(info_path):
with open(info_path, 'rb') as fid:
return cPickle.load(fid)
else:
raise Exception("{} not find".format(info_path))
def _load_namelist_ind(self):
ind = range(len(self._info))
np.random.shuffle(ind)
return ind
def _save_test_pkl(self):
cache_file = os.path.join('./data', 'IW_test_files.pkl')
with open(cache_file, 'wb') as fid:
cPickle.dump(self._test_ind, fid, cPickle.HIGHEST_PROTOCOL)
def get_dataset(self, sample_inds):
# st = time.time()
sample = [self.load_image(
image_path=self._info[sample_ind][1],
image_size=self.size_image,
image_value_range=self.image_value_range,
is_gray=False,
) for sample_ind in sample_inds]
# print(time.time()-st)
sample_images = np.array(sample).astype(np.float32)
sample_label_age = np.ones(
shape=(len(sample_inds), self.num_categories),
dtype=np.float32
) * self.image_value_range[0]
sample_label_gender = np.ones(
shape=(len(sample_inds), 2),
dtype=np.float32
) * self.image_value_range[0]
for i, ind in enumerate(sample_inds):
label, _, gender = self._info[ind]
label = int(label)
gender = int(gender)
if gender == -1:
gender = np.random.randint(2)
if 0 <= label <= 18:
label = 0
elif 19 <= label <= 29:
label = 1
elif 30 <= label <= 39:
label = 2
elif 40 <= label <= 49:
label = 3
elif 50 <= label <= 59:
label = 4
else:
label = 5
sample_label_age[i, label] = self.image_value_range[-1]
sample_label_gender[i, gender] = self.image_value_range[-1]
return sample_images, sample_label_age, sample_label_gender