-
Notifications
You must be signed in to change notification settings - Fork 84
/
Copy pathgrasp_img_proc.py
160 lines (132 loc) · 6.19 KB
/
grasp_img_proc.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
import tensorflow as tf
import numpy as np
FLAGS = tf.app.flags.FLAGS
tf.app.flags.DEFINE_integer('image_size', 224,
"""Provide square images of this size.""")
tf.app.flags.DEFINE_integer('num_preprocess_threads', 12,
"""Number of preprocessing threads per tower. """
"""Please make this a multiple of 4.""")
tf.app.flags.DEFINE_integer('num_readers', 12,
"""Number of parallel readers during train.""")
tf.app.flags.DEFINE_integer('input_queue_memory_factor', 12,
"""Size of the queue of preprocessed images. """
"""Default is ideal but try smaller values, e.g. """
"""4, 2 or 1, if host memory is constrained. See """
"""comments in code for more details.""")
def parse_example_proto(examples_serialized):
feature_map={
'image/encoded': tf.FixedLenFeature([], dtype=tf.string,
default_value=''),
'bboxes': tf.VarLenFeature(dtype=tf.float32)
}
features=tf.parse_single_example(examples_serialized, feature_map)
bboxes = tf.sparse_tensor_to_dense(features['bboxes'])
r = 8*tf.random_uniform((1,), minval=0, maxval=tf.size(bboxes, out_type=tf.int32)/8, dtype=tf.int32)
bbox = tf.gather_nd(bboxes, [r,r+1,r+2,r+3,r+4,r+5,r+6,r+7])
return features['image/encoded'], bbox
def eval_image(image, height, width):
image = tf.image.central_crop(image, central_fraction=0.875)
image = tf.expand_dims(image, 0)
image = tf.image.resize_bilinear(image, [height, width],
align_corners=False)
image = tf.squeeze(image, [0])
return image
def distort_color(image, thread_id):
color_ordering = thread_id % 2
if color_ordering == 0:
image = tf.image.random_brightness(image, max_delta=32. / 255.)
image = tf.image.random_saturation(image, lower=0.5, upper=1.5)
image = tf.image.random_hue(image, max_delta=0.2)
image = tf.image.random_contrast(image, lower=0.5, upper=1.5)
elif color_ordering == 1:
image = tf.image.random_brightness(image, max_delta=32. / 255.)
image = tf.image.random_contrast(image, lower=0.5, upper=1.5)
image = tf.image.random_saturation(image, lower=0.5, upper=1.5)
image = tf.image.random_hue(image, max_delta=0.2)
image = tf.clip_by_value(image, 0.0, 1.0)
return image
def distort_image(image, height, width, thread_id):
distorted_image = tf.image.random_flip_left_right(image)
distorted_image = distort_color(distorted_image, thread_id)
return distorted_image
def image_preprocessing(image_buffer, train, thread_id=0):
height = FLAGS.image_size
width = FLAGS.image_size
image = tf.image.decode_png(image_buffer, channels=3)
image = tf.image.convert_image_dtype(image, dtype=tf.float32)
image = tf.image.resize_images(image, [height,width])
if train:
image = distort_image(image, height, width, thread_id)
#else:
# image = eval_image(image, height, width)
image = tf.subtract(image, 0.5)
image = tf.multiply(image, 2.0)
return image
def batch_inputs(data_files, train, num_epochs, batch_size,
num_preprocess_threads, num_readers):
print(train)
if train:
filename_queue = tf.train.string_input_producer(data_files,
num_epochs,
shuffle=True,
capacity=16)
else:
filename_queue = tf.train.string_input_producer(data_files,
num_epochs,
shuffle=False,
capacity=1)
examples_per_shard = 1024
min_queue_examples = examples_per_shard * FLAGS.input_queue_memory_factor
if train:
print('pass')
examples_queue = tf.RandomShuffleQueue(
capacity=min_queue_examples+3*batch_size,
min_after_dequeue=min_queue_examples,
dtypes=[tf.string])
else:
examples_queue = tf.FIFOQueue(
capacity=examples_per_shard + 3 * batch_size,
dtypes=[tf.string])
if num_readers > 1:
enqueue_ops = []
for _ in range(num_readers):
reader = tf.TFRecordReader()
_, value = reader.read(filename_queue)
enqueue_ops.append(examples_queue.enqueue([value]))
tf.train.queue_runner.add_queue_runner(
tf.train.queue_runner.QueueRunner(examples_queue,enqueue_ops))
examples_serialized = examples_queue.dequeue()
else:
reader = tf.TFRecordReader()
_, examples_serialized = reader.read(filename_queue)
images_and_bboxes=[]
for thread_id in range(num_preprocess_threads):
image_buffer, bbox = parse_example_proto(examples_serialized)
image = image_preprocessing(image_buffer, train, thread_id)
images_and_bboxes.append([image, bbox])
images, bboxes = tf.train.batch_join(
images_and_bboxes,
batch_size=batch_size,
capacity=2*num_preprocess_threads*batch_size)
height = FLAGS.image_size
width = FLAGS.image_size
depth = 3
images = tf.cast(images, tf.float32)
images = tf.reshape(images, shape=[batch_size, height, width, depth])
return images, bboxes
def distorted_inputs(data_files, num_epochs, train=True, batch_size=None):
with tf.device('/cpu:0'):
print(train)
images, bboxes = batch_inputs(
data_files, train, num_epochs, batch_size,
num_preprocess_threads=FLAGS.num_preprocess_threads,
num_readers=FLAGS.num_readers)
return images, bboxes
def inputs(data_files, num_epochs=1, train=False, batch_size=1):
with tf.device('/cpu:0'):
print(train)
images, bboxes = batch_inputs(
data_files, train, num_epochs, batch_size,
num_preprocess_threads=FLAGS.num_preprocess_threads,
num_readers=1)
return images, bboxes