-
Notifications
You must be signed in to change notification settings - Fork 4
/
extract_triplet_cross.py
192 lines (139 loc) · 6.3 KB
/
extract_triplet_cross.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
import numpy as np
import os,sys
import tensorflow as tf
import os.path as osp
import multiprocessing as mp
import config
import similarity
import model
import IPython
from data import transform_img
import cv2, pickle
flags = tf.app.flags
flags.DEFINE_string('feature', 'fc6', 'Extract which layer(pool5, fc6, fc7)')
flags.DEFINE_string('model_dir', None, 'Model directory')
flags.DEFINE_integer('batch_size', 128, 'Value of batch size')
flags.DEFINE_integer('p', 200, 'Size of proposals')
FLAGS = flags.FLAGS
#layer_list = ["pool5"]
layer_list = [FLAGS.feature]
output_root = "visual_feature/triplet"
query_dir = ["search", "streetview_clean", "aerial_clean"]
proposal_max = FLAGS.p
#query_dir = ["image_gt"]
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
def write_pkl(pkl, sess, pred):
index = 0
while index*FLAGS.batch_size<len(pkl):
img = []
for p in pkl[index*FLAGS.batch_size:(index+1)*FLAGS.batch_size]:
img.append(p[0])
out = sess.run([pred], feed_dict={img_search: img})
out = np.array(out[0])
p_i = 0
for p in pkl[index*FLAGS.batch_size:(index+1)*FLAGS.batch_size]:
p.append(out[p_i])
p_i += 1
for p in pkl[index*FLAGS.batch_size:(index+1)*FLAGS.batch_size]:
with open(os.path.join(p[1]), 'wb') as ff:
pickle.dump(p[2], ff)
index += 1
print(len(pkl))
def write_cross_pkl(pkl, sess, cross_pred):
index = 0
while index*FLAGS.batch_size<len(pkl):
search = []
street = []
aerial = []
for p in pkl[index*FLAGS.batch_size:(index+1)*FLAGS.batch_size]:
search.append(p[0]['search'])
street.append(p[0]['streetview_clean'])
aerial.append(p[0]['aerial_clean'])
try:
out = sess.run([cross_pred], feed_dict={img_search: search, img_street: street, img_aerial: aerial})
except:
IPython.embed()
out = np.array(out[0])
p_i = 0
for p in pkl[index*FLAGS.batch_size:(index+1)*FLAGS.batch_size]:
p.append(out[p_i])
p_i += 1
for p in pkl[index*FLAGS.batch_size:(index+1)*FLAGS.batch_size]:
with open(os.path.join(p[1]), 'wb') as ff:
pickle.dump(p[2], ff)
index += 1
print(len(pkl))
with tf.Graph().as_default(), tf.Session(config=config) as sess:
img_search = tf.placeholder('float32', shape=(None, 227, 227, 3))
img_street = tf.placeholder('float32', shape=(None, 227, 227, 3))
img_aerial = tf.placeholder('float32', shape=(None, 227, 227, 3))
cross_feature = model.inference_crossview([img_search, img_street, img_aerial]
, 1, FLAGS.feature, False)
norm_cross_pred = model.feature_normalize([cross_feature])
cross_pred = norm_cross_pred[0]
feature = model.inference(img_search, 1, FLAGS.feature)
norm_pred = model.feature_normalize([feature])
pred = norm_pred[0]
saver = tf.train.Saver()
if FLAGS.model_dir:
saver.restore(sess, FLAGS.model_dir)
else:
saver.restore(sess, 'model/{}/model_final'.format(FLAGS.feature))
pkl_list = {}
if True:
output_dir = os.path.join(output_root, "cross")
for img_name in os.listdir(query_dir[0]):
if img_name.find('.jpg')==-1: #is a directory
continue
img_name=img_name.replace(".jpg","").replace(".png","")
print(img_name)
img = {}
for query in query_dir:
img[query] = cv2.imread(os.path.join(query, img_name+'.jpg'), cv2.IMREAD_COLOR)
img[query] = transform_img(img[query], 227,227)
for layer in layer_list:
output_layer = os.path.join(output_dir, layer)
if not layer in pkl_list:
pkl_list[layer] = []
if not os.path.exists(output_layer):
os.makedirs(output_layer)
pkl_list[layer].append([img, os.path.join(output_layer, img_name+".pkl")])
for layer in layer_list:
write_cross_pkl(pkl_list[layer], sess, cross_pred)
if True:
frame_dir="frame/all/"
proposal_list = ["faster_bb"]
pkl_list = {}
for img_name in os.listdir(frame_dir):
img_name=img_name.replace(".jpg","").replace(".png","")
origin_img = cv2.imread( frame_dir+img_name+'.jpg', cv2.IMREAD_COLOR)
for proposal_dir in proposal_list:
with open(os.path.join(proposal_dir, img_name+".txt"), 'r') as ff:
output_dir = os.path.join(output_root, proposal_dir)
proposal_num = 0
for linee in ff:
token=linee.strip().split()
bb=[int(float(token[0])), int(float(token[1])), int(float(token[2])), int(float(token[3]))]
box_score=float(token[4])
[bb_width,bb_height]=[bb[3]-bb[1],bb[2]-bb[0]]
img=origin_img[bb[1]:bb[3],bb[0]:bb[2]]
print(bb)
img = transform_img(img,227,227)
for layer in layer_list:
output_layer = os.path.join(output_dir, layer)
if not os.path.exists(output_dir):
os.makedirs(output_dir)
key = proposal_dir+"_"+layer
if not key in pkl_list:
pkl_list[key] = []
pkl_list[key].append([img, os.path.join(output_layer, img_name+"_"+str(bb[0])+"_"+str(bb[1])+"_"+str(bb[2])+"_"+str(bb[3])+".pkl")])
proposal_num += 1
if proposal_num >= proposal_max:
break
for proposal_dir in proposal_list:
for layer in layer_list:
key = proposal_dir+"_"+layer
if not os.path.exists(output_layer):
os.makedirs(output_layer)
write_pkl(pkl_list[key], sess, pred)