-
Notifications
You must be signed in to change notification settings - Fork 2
/
train_gd_nir2nir.py
356 lines (305 loc) · 15.4 KB
/
train_gd_nir2nir.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
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
# -*- coding: utf-8 -*-
"""
Created on 2018.11.18
@author: DELL
"""
'''
用于生成器的训练,用于对 nir 进行自编码学习
nir --> nir
'''
import numpy as np
import math
import time
import model_gd_nir as model
import tensorflow as tf
import read_data_map as read
import os
from datetime import datetime
import logging
import cv2
BATCH_SIZE_matching = 100
EPOCH = 81
learning_rate = 2e-4
image_width = 64
image_height = 64
#checkpoint_dir = 'ckpt_gd_r2r'
#checkpoint_dir_en = 'ckpt_en_r2r'
#output_dir = 'out_r2r'
#checkpoint_file = os.path.join(checkpoint_dir, 'model.ckpt')
#checkpoint_file_en = os.path.join(checkpoint_dir_en, 'model.ckpt')
train_dir='summary_gd'
os.environ['CUDA_VISIBLE_DEVICES']='0'
def initLogging(logFilename='record_gd_r2r.log'):
"""Init for logging
"""
logging.basicConfig(
level = logging.DEBUG,
format='%(asctime)s-%(levelname)s-%(message)s',
datefmt = '%y-%m-%d %H:%M',
filename = logFilename,
filemode = 'w');
console = logging.StreamHandler()
console.setLevel(logging.INFO)
formatter = logging.Formatter('%(asctime)s-%(levelname)s-%(message)s')
console.setFormatter(formatter)
logging.getLogger('').addHandler(console)
def preprocess(image):
# with tf.name_scope("preprocess"):
# [0, 1] => [-1, 1]
return image * 2.0 - 1.0
def deprocess(image):
# with tf.name_scope("deprocess"):
# [-1, 1] => [0, 1]
return (image + 1.0) / 2.0
def combine_images(generated_images):
num = generated_images.shape[0]
width = int(math.sqrt(num))
height = int(math.ceil(float(num)/width))
shape = generated_images.shape[1:3]
image = np.zeros((height*shape[0], width*shape[1]),
dtype=generated_images.dtype)
for index, img in enumerate(generated_images):
i = int(index/width)
j = index % width
image[i*shape[0]:(i+1)*shape[0], j*shape[1]:(j+1)*shape[1]] = img[:,:,0]
return image
def gd_train(f_name):
checkpoint_dir = 'ckpt_gd_r2r_'+f_name
checkpoint_dir_en = 'ckpt_en_r2r_'+f_name
output_dir = 'out_r2r_'+f_name
checkpoint_file = os.path.join(checkpoint_dir, 'model.ckpt')
checkpoint_file_en = os.path.join(checkpoint_dir_en, 'model.ckpt')
initLogging()
# if FLAGS.load_model is not None:
# checkpoints_dir = 'checkpoints/' + FLAGS.load_model
# else:
current_time = datetime.now().strftime('%Y%m%d-%H%M')
try:
os.makedirs(checkpoint_dir_en)
os.makedirs(checkpoint_dir)
except os.error:
pass
try:
os.makedirs(output_dir)
except os.error:
pass
##
data1 = np.load('/home/ws/文档/wrj/mapping_data_gray/map_' + f_name + '.npz')
train_matching_y = data1['arr_0'][:,np.newaxis]
numbers_train = train_matching_y.shape[0] #训练集总数
epoch_steps = np.int(numbers_train / BATCH_SIZE_matching) +1 # 一个epoch有多少个steps
all_loss = np.array([])
graph = tf.Graph()
with graph.as_default():
inputs_p1 = tf.placeholder(tf.float32, [BATCH_SIZE_matching, image_height, image_width,1], name='inputs_gray')
inputs_p2 = tf.placeholder(tf.float32, [BATCH_SIZE_matching, image_height, image_width,1], name='inputs_nir')
# test_opt = tf.placeholder(tf.float32, [None, image_height, image_width, 1], name='test_opt')
# 训练 G
inputs_p2_ = model.preprocess(inputs_p2)
gen_loss ,nir_layers = model.gd_model_r2r(inputs_p2_)
g_train_opt = tf.train.AdamOptimizer(learning_rate).minimize(gen_loss)
filename = '/home/ws/文档/wrj/mapping_data_gray/map_data/map_' + f_name + '.tfrecord'
filename_queue = tf.train.string_input_producer([filename],num_epochs=EPOCH,shuffle=True)
img_batch, label_batch = read.batch_inputs(filename_queue, train = True, batch_size = BATCH_SIZE_matching)
saver = tf.train.Saver(max_to_keep=20)
# 用于保存 编码器的权重参数
en_tvars = [var for var in tf.trainable_variables() if var.name.startswith("nir_encode")]
saver_g = tf.train.Saver(var_list=en_tvars, max_to_keep=20)
init = tf.global_variables_initializer()
#
gpu_options = tf.GPUOptions(allow_growth=True)
sess_config = tf.ConfigProto(gpu_options=gpu_options)
with tf.Session(config=sess_config) as sess:
sess.run(tf.local_variables_initializer())
sess.run(init)
try:
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
step = 0
while not coord.should_stop():
start_time = time.time()
step = step + 1
batch, l_batch = sess.run([img_batch, label_batch])
y_batch = batch[:,:,64:,np.newaxis]
feed_dict = {inputs_p2:y_batch}
_, g_loss = sess.run([ g_train_opt, gen_loss], feed_dict = feed_dict)
if step % 10 ==0:
loss_write = np.array([[step, g_loss]])
if step ==10:
all_loss = loss_write
else:
all_loss = np.concatenate((all_loss, loss_write))
if step % 100 == 0:
duration = time.time() - start_time
logging.info('>> Step %d run_train: g_loss = %.2f (%.3f sec)'
% (step, g_loss, duration))
if (step % epoch_steps == 0) and ((step / epoch_steps)%10 == 0):
current_epoch = int(step / epoch_steps)
if current_epoch >= 50:
logging.info('>> %s Saving in %s' % (datetime.now(), checkpoint_dir))
saver.save(sess, checkpoint_file, global_step=current_epoch)
saver_g.save(sess, checkpoint_file_en, global_step=current_epoch)
mc_test_all(current_epoch)
# if current_epoch >= 50:
# mc_test(f_name, current_epoch)
except KeyboardInterrupt:
print('INTERRUPTED')
coord.request_stop()
except Exception as e:
coord.request_stop(e)
finally:
saver.save(sess, checkpoint_file, global_step=step)
saver_g.save(sess, checkpoint_file_en, global_step=step)
np.save(os.path.join(checkpoint_dir,'ckpt_map_gd_nir_'+f_name), all_loss)
np.save(os.path.join(checkpoint_dir_en,'ckpt_map_en_nir_'+f_name), all_loss)
print('Model saved in file :%s'%checkpoint_dir)
# When done, ask the threads to stop.
coord.request_stop()
coord.join(threads)
def mc_draw(file_name, number):
# data = np.load('/media/ws/98E62262E622413C/zzzz_wrj/wrj/mapping_data_all/map_rgb_'+file_name+'.npz')
# train_matching_y = data['arr_0'][:]
# numbers_train = train_matching_y.shape[0] #训练集总数
BATCH_SIZE_map = 40
graph = tf.Graph()
with graph.as_default():
inputs_p1 = tf.placeholder(tf.float32, [BATCH_SIZE_map, image_height, image_width,1], name='inputs_gray')
inputs_p2 = tf.placeholder(tf.float32, [BATCH_SIZE_map, image_height, image_width,1], name='inputs_nir')
inputs_p2_ = model.preprocess(inputs_p2)
# gen_loss, all_layers = model.gd_model_r2r(inputs_p2_)
gen,all_layers = model.create_generator(inputs_p2_, 1, reuse=True)
gen = model.deprocess(gen)
filename = '/home/ws/文档/wrj/data_all/country/'+file_name+'.tfrecord'
filename_queue = tf.train.string_input_producer([filename],num_epochs=1,shuffle=False)
img_batch, label_batch = read.batch_inputs(filename_queue, train = False, batch_size = BATCH_SIZE_map)
saver = tf.train.Saver()
with tf.Session() as sess:
sess.run(tf.local_variables_initializer())
saver.restore(sess, tf.train.latest_checkpoint('ckpt_gd_r2r_all'))
# saver.restore(sess, 'ckpt_tensor/model.ckpt-4000')
# num = 0
try:
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
step_test = 0
while not coord.should_stop():
if step_test < 1:
step_test = step_test + 1
batch, l_batch = sess.run([img_batch, label_batch])
y_batch = batch[:,:,64:,np.newaxis]
feed_dict = {inputs_p2:y_batch}
gen_out, layers_out = sess.run([gen,all_layers], feed_dict = feed_dict)
gen_out_dir = 'out_r2r_country/epoch_' + str(number)
try:
os.makedirs('feature_map')
os.makedirs(gen_out_dir)
except os.error:
pass
show_images = np.concatenate((y_batch, gen_out), axis=1)
show_images = show_images*255
for i in range(BATCH_SIZE_map):
# cv2.imwrite(gen_out_dir+'/{}.png'.format(file_name+'_'+str(i+1)+"_opt"), np.squeeze(x_batch[i,:,:,:]*255))
cv2.imwrite(gen_out_dir+'/{}.png'.format(file_name+'_'+str(i+1)+"_nir"), np.squeeze(show_images[i,:64,:,:]))
cv2.imwrite(gen_out_dir+'/{}.png'.format(file_name+'_'+str(i+1)+"_fnir"), np.squeeze(show_images[i,64:,:,:]))
layers = np.squeeze(layers_out[2][i,:,:,:])
m=0;n=0;s=8
max_ = np.max(layers)
min_ = np.min(layers)
layers_ = (layers - min_)/(max_ - min_)
features = np.ones((11*s,12*s),np.uint8)*255
for j in range(128):
feature = np.squeeze(np.uint8(layers_[:,:,j]*255))
features[m*s:m*s+s, n*s:n*s+s] = feature
n+=1
if(n>=12):
n=0
m+=1
cv2.imwrite('feature_map/{}.png'.format(file_name+'_'+str(i+1)), features)
else:
break
except KeyboardInterrupt:
print('INTERRUPTED')
coord.request_stop()
except Exception as e:
coord.request_stop(e)
finally:
coord.request_stop()
coord.join(threads)
'''
test the network
'''
def mc_test(file_name, number):
# data = np.load('/media/ws/98E62262E622413C/zzzz_wrj/wrj/mapping_data_all/map_rgb_'+file_name+'.npz')
# train_matching_y = data['arr_0'][:]
# numbers_train = train_matching_y.shape[0] #训练集总数
BATCH_SIZE_map = 40
graph = tf.Graph()
with graph.as_default():
inputs_p1 = tf.placeholder(tf.float32, [BATCH_SIZE_map, image_height, image_width,1], name='inputs_gray')
inputs_p2 = tf.placeholder(tf.float32, [BATCH_SIZE_map, image_height, image_width,1], name='inputs_nir')
inputs_p2_ = model.preprocess(inputs_p2)
# gen_loss, all_layers = model.gd_model_r2r(inputs_p2_)
gen,all_layers = model.create_generator(inputs_p2_, 1, reuse=True)
gen = model.deprocess(gen)
filename = '/home/ws/文档/wrj/data_all/country/'+file_name+'.tfrecord'
filename_queue = tf.train.string_input_producer([filename],num_epochs=1,shuffle=False)
img_batch, label_batch = read.batch_inputs(filename_queue, train = False, batch_size = BATCH_SIZE_map)
saver = tf.train.Saver()
with tf.Session() as sess:
sess.run(tf.local_variables_initializer())
saver.restore(sess, tf.train.latest_checkpoint('ckpt_gd_r2r_all'))
# saver.restore(sess, 'ckpt_tensor/model.ckpt-4000')
# num = 0
try:
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
step_test = 0
while not coord.should_stop():
if step_test < 1:
step_test = step_test + 1
batch, l_batch = sess.run([img_batch, label_batch])
y_batch = batch[:,:,64:,np.newaxis]
feed_dict = {inputs_p2:y_batch}
gen_out = sess.run(gen, feed_dict = feed_dict)
gen_out_dir = 'out_r2r_all/epoch_' + str(number)
try:
os.makedirs(gen_out_dir)
except os.error:
pass
show_images = np.concatenate((y_batch, gen_out), axis=1)
show_images = show_images*255
for i in range(BATCH_SIZE_map):
# cv2.imwrite(gen_out_dir+'/{}.png'.format(file_name+'_'+str(i+1)+"_opt"), np.squeeze(x_batch[i,:,:,:]*255))
cv2.imwrite(gen_out_dir+'/{}.png'.format(file_name+'_'+str(i+1)+"_nir"), np.squeeze(show_images[i,:64,:,:]))
cv2.imwrite(gen_out_dir+'/{}.png'.format(file_name+'_'+str(i+1)+"_fnir"), np.squeeze(show_images[i,64:,:,:]))
else:
break
except KeyboardInterrupt:
print('INTERRUPTED')
coord.request_stop()
except Exception as e:
coord.request_stop(e)
finally:
coord.request_stop()
coord.join(threads)
'''
test all datasets
'''
def mc_draw_all(number):
test_data = ['country', 'field', 'forest','indoor','mountain','oldbuilding','street','urban','water']
for i in range(9):
mc_draw(test_data[i],number)
def mc_test_all(number):
test_data = ['country', 'field', 'forest','indoor','mountain','oldbuilding','street','urban','water']
for i in range(9):
mc_test(test_data[i],number)
def mc_train_all():
train_data = ['country', 'field', 'forest','indoor','mountain','oldbuilding','street','urban','water']
for i in range(1):
gd_train(train_data[i])
if __name__ == '__main__':
# mc_train_all()
# gd_train('all')
# mc_test_all(1)
# mc_test('country', 25)
mc_draw_all(25)