-
Notifications
You must be signed in to change notification settings - Fork 1
/
data_loader.py
120 lines (84 loc) · 4.21 KB
/
data_loader.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
import tensorflow as tf
import numpy as np
import argparse
import os
class data_loader():
def __init__(self, conf):
self.batch_size = args.batch_size
self.in_memory = args.in_memory
self.channel = args.channel
self.tf_records = args.tf_records
if self.in_memory:
self.width = args.width
self.height = args.height
self.image_arr = tf.placeholder(shape = [None, self.height, self.width, self.channel], dtype = tf.uint8)
if not self.in_memory or self.tf_records :
self.image_arr = np.array([os.path.join(args.data_direc, ele) for ele in sorted(os.listdir(args.data_direc))])
def build_loader(self):
if not self.tf_records:
self.tr_dataset = tf.data.Dataset.from_tensor_slices(self.image_arr)
if not self.in_memory:
self.tr_dataset = self.tr_dataset.map(self._parse, num_parallel_calls = 4).prefetch(32)
else:
self.tr_dataset = tf.data.TFRecordDataset(self.image_arr)
self.tr_dataset = self.tr_dataset.map(self._tf_record_parse, num_parallel_calls = 4).prefetch(32)
self.tr_dataset = self.tr_dataset.shuffle(32)
self.tr_dataset = self.tr_dataset.repeat()
self.tr_dataset = self.tr_dataset.batch(self.batch_size)
iterator = tf.data.Iterator.from_structure(self.tr_dataset.output_types, self.tr_dataset.output_shapes)
self.next_batch = iterator.get_next()
self.init_op = iterator.make_initializer(self.tr_dataset)
def _parse(self, image):
image = tf.read_file(image)
image = tf.image.decode_png(image, channels = self.channel)
return image
def _tf_record_parse(self, example):
feature = {'img' : tf.FixedLenFeature([], tf.string),
'height' : tf.FixedLenFeature([], tf.int64),
'width' : tf.FixedLenFeature([], tf.int64)}
parsed_feature = tf.parse_single_example(example, feature)
img = tf.decode_raw(parsed_feature['img'], tf.uint8)
height = tf.cast(parsed_feature['height'], tf.int32)
width = tf.cast(parsed_feature['width'], tf.int32)
img = tf.reshape(img, (height, width, self.channel))
return img
if __name__ == '__main__':
from PIL import Image
def str2bool(v):
return v.lower() in ('true')
parser = argparse.ArgumentParser()
parser.add_argument("--batch_size", type = int, default = 4)
parser.add_argument("--in_memory", type = str2bool, default = False)
parser.add_argument("--channel", type = int, default = 3)
parser.add_argument("--width", type = int, default = 600)
parser.add_argument("--height", type = int, default = 600)
parser.add_argument("--data_direc", type = str, default = './test_data/')
parser.add_argument("--result_path", type = str, default = './result/')
parser.add_argument("--tf_records", type = str2bool, default = False)
parser.add_argument("--test_num", type = int, default = 3)
args = parser.parse_args()
if not os.path.exists(args.result_path):
os.makedirs(args.result_path)
data_loader = data_loader(args)
data_loader.build_loader()
tf_output = data_loader.next_batch
tf_output = tf_output[:,:50,:50,:]
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
sess = tf.Session(config = config)
sess.run(tf.global_variables_initializer())
if args.in_memory and not args.tf_records:
img_list = []
for ele in sorted(os.listdir(args.data_direc)):
img_list.append(np.array(Image.open(os.path.join(args.data_direc, ele))))
img_list = np.array(img_list)
sess.run(data_loader.init_op, feed_dict = {data_loader.image_arr : img_list})
else:
sess.run(data_loader.init_op)
count = 0
for i in range(args.test_num):
output = sess.run(tf_output)
for ele in output:
img = Image.fromarray(ele)
img.save(os.path.join(args.result_path, '%02d_result.png'%count))
count += 1