-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathglimpse.py
172 lines (136 loc) · 5.99 KB
/
glimpse.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
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import tensorflow as tf
from utils import weight_variable, bias_variable
class GlimpseNet(object):
"""Glimpse network.
Take glimpse location input and output features for RNN.
"""
def __init__(self, config, images_ph):
self.original_size = config.original_size
self.num_channels = config.num_channels
self.sensor_size = config.sensor_size
self.win_size = config.win_size
self.minRadius = config.minRadius
self.depth = config.depth
self.glimpse_size = config.glimpse_size
#self.batch_size = config.batch_size
self.minRadius = config.minRadius
self.hg_size = config.hg_size
self.hl_size = config.hl_size
self.g_size = config.g_size
self.loc_dim = config.loc_dim
self.images_ph = images_ph
self.init_weights()
def init_weights(self):
""" Initialize all the trainable weights."""
self.w_g0 = weight_variable((self.sensor_size, self.hg_size))
self.b_g0 = bias_variable((self.hg_size,))
self.w_l0 = weight_variable((self.loc_dim, self.hl_size))
self.b_l0 = bias_variable((self.hl_size,))
self.w_g1 = weight_variable((self.hg_size, self.g_size))
self.b_g1 = bias_variable((self.g_size,))
self.w_l1 = weight_variable((self.hl_size, self.g_size))
self.b_l1 = weight_variable((self.g_size,))
def get_glimpse(self, loc):
"""Take glimpse on the original images."""
imgs = tf.reshape(self.images_ph, [
tf.shape(self.images_ph)[0], self.original_size, self.original_size,
self.num_channels
])
glimpses=[]
'''
for k in xrange(batch_size):
imgZooms = []
one_img = img[k,:,:,:]
max_radius = self.minRadius * (2 ** (self.depth - 1))
offset = 2 * max_radius
# pad image with zeros
one_img = tf.image.pad_to_bounding_box(one_img, offset, offset, \
max_radius * 4 + self.original_size, max_radius * 4 + self.original_size)
for i in xrange(depth):
r = int(minRadius * (2 ** (i)))
d_raw = 2 * r
d = tf.constant(d_raw, shape=[1])
d = tf.tile(d, [2])
loc_k = loc[k,:]
adjusted_loc = offset + loc_k - r
one_img2 = tf.reshape(one_img, (one_img.get_shape()[0].value, one_img.get_shape()[1].value))
# crop image to (d x d)
zoom = tf.slice(one_img2, adjusted_loc, d)
# resize cropped image to (sensorBandwidth x sensorBandwidth)
zoom = tf.image.resize_bilinear(tf.reshape(zoom, (1, d_raw, d_raw, 1)), (self.glimpse_size, self.glimpse_size))
zoom = tf.reshape(zoom, (self.glimpse_size, self.glimpse_size))
imgZooms.append(zoom)
glimpses.append(tf.stack(imgZooms))
glimpses = tf.stack(glimpses)
'''
max_radius = self.minRadius * (2 ** (self.depth - 1))
offset = 2 * max_radius
loc = loc * self.original_size / (max_radius * 2 + self.original_size)
padding_imgs = tf.image.pad_to_bounding_box(imgs, max_radius, max_radius, \
max_radius * 2 + self.original_size, max_radius * 2 + self.original_size)
for i in xrange(self.depth):
size = int( self.win_size * (2 ** (i)) )
glimpse_imgs = tf.image.extract_glimpse(padding_imgs, [size, size], loc, uniform_noise=False)
glimpse_imgs_resize = tf.image.resize_bilinear(glimpse_imgs, [self.glimpse_size, self.glimpse_size])
glimpse_imgs_flatten = tf.reshape(glimpse_imgs_resize, [
tf.shape(loc)[0], self.glimpse_size * self.glimpse_size * self.num_channels
])
glimpses.append(glimpse_imgs_flatten)
#loc = tf.round(((loc + 1) / 2.0) * self.original_size)
#loc = tf.cast(loc, tf.int32)
loc = (loc + 1) / 2.0
loc1 = loc - 1./2 * self.glimpse_size/(max_radius * 2 + self.original_size)
loc2 = loc + 1./2 * self.glimpse_size/(max_radius * 2 + self.original_size)
loc = tf.concat([loc1, loc2], axis=1)
loc = tf.expand_dims(loc, axis=1)
#loc = tf.clip_by_value(loc, 0., 1.)
summary_imgs = tf.image.draw_bounding_boxes(padding_imgs, loc)
tf.summary.image('glimpse/imgs_' + str(i), glimpse_imgs, max_outputs = 1)
tf.summary.image('bbox/imgs_' + str(i), summary_imgs, max_outputs = 1)
glimpses = tf.stack(glimpses, axis=1)
return glimpses
def __call__(self, loc):
glimpse_input = self.get_glimpse(loc)
glimpse_input = tf.reshape(glimpse_input,
(tf.shape(loc)[0], self.sensor_size))
g = tf.nn.relu(tf.nn.xw_plus_b(glimpse_input, self.w_g0, self.b_g0))
g = tf.nn.xw_plus_b(g, self.w_g1, self.b_g1)
l = tf.nn.relu(tf.nn.xw_plus_b(loc, self.w_l0, self.b_l0))
l = tf.nn.xw_plus_b(l, self.w_l1, self.b_l1)
g = tf.nn.relu(g + l)
return g
class LocNet(object):
"""Location network.
Take output from other network and produce and sample the next location.
"""
def __init__(self, config):
self.loc_dim = config.loc_dim
self.input_dim = config.cell_output_size
self.loc_std = config.loc_std
self._sampling = True
self.init_weights()
def init_weights(self):
self.w = weight_variable((self.input_dim, self.loc_dim))
self.b = bias_variable((self.loc_dim,))
def __call__(self, input):
mean = tf.nn.xw_plus_b(input, self.w, self.b)
#mean = tf.tanh(mean)
mean = tf.clip_by_value(tf.nn.xw_plus_b(input, self.w, self.b), -1., 1.)
#mean = tf.stop_gradient(mean)
if self._sampling:
loc = mean + tf.random_normal(
(tf.shape(input)[0], self.loc_dim), stddev=self.loc_std)
loc = tf.clip_by_value(loc, -1., 1.)
else:
loc = mean
loc = tf.stop_gradient(loc)
return loc, mean
@property
def sampling(self):
return self._sampling
@sampling.setter
def sampling(self, sampling):
self._sampling = sampling