-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
146 lines (136 loc) · 5.71 KB
/
train.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
import numpy as np
import tensorflow as tf
import os
import time
import ops
import data_buscrowd
import matplotlib.pyplot as plt
# from model import model
# from networks.mymodel_with_offset import model as model
from networks.mymodel import model as model
# from networks.resnet import resnet_v2_152 as model
# from networks.google_v3 import inception_v3 as model
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
os.environ['CUDA_VISIBLE_DEVICES'] = '0'
MAX_STEP=10000
BATCH_SIZE=32
logs_train_dir='output/'
def save_list(data_list,name='result.txt'):
fp = open(name, 'w+')
for i in range(len(data_list)):
fp.write(str(data_list[i]))
fp.write(" ")
fp.close()
def remove_output():
path = 'output/'
for i in os.listdir(path):
path_file = os.path.join(path, i)
if os.path.isfile(path_file):
os.remove(path_file)
def check_weights(sess,scope_name,tensor_name):
with tf.variable_scope(scope_name,reuse=True):
tensor = tf.get_variable(tensor_name)
val=sess.run(tensor)
print(tensor_name,val.shape,val)
return val
def train():
remove_output()
is_training = tf.placeholder(tf.bool, shape=())
train_img, train_lable = data_buscrowd.get_train_data(batch_size=BATCH_SIZE)
test_img, test_lable = data_buscrowd.get_test_data(batch_size=BATCH_SIZE)
x = tf.cond(is_training, lambda: train_img, lambda: test_img)
y = tf.cond(is_training, lambda: train_lable, lambda: test_lable)
logits=model(x,num_classes=4)
losses=ops.loss_sparse_softmax_cross_entropy(logits,y)
acc=ops.evaluation(logits,y)
step_ = tf.Variable(tf.constant(0))
learing_rate = tf.train.exponential_decay(
learning_rate=0.001, global_step=step_, decay_steps=100, decay_rate=0.8, staircase=True)
train_op=ops.optimize_adam(losses,learing_rate)
saver = tf.train.Saver()
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
max_acc=0
try:
total_loss=[]
total_acc=[]
total_test_acc=[]
plt.ion() # 开启interactive mode 成功的关键函数
for step in np.arange(MAX_STEP):
if coord.should_stop():
break
_,val_loss,val_acc,val_logits,val_lable,val_lr = sess.run([train_op,losses,acc,logits,y,learing_rate],feed_dict={is_training: True,step_:step})
if step % 10 == 0:
val_loss_test, val_acc_test = sess.run([losses, acc], feed_dict={is_training: False})
total_loss.append(val_loss)
total_acc.append(val_acc)
total_test_acc.append(val_acc_test)
print('\n',step,' lr:',val_lr,' loss:',val_loss)
print(' ','acc:',val_acc)
print('eg:',val_logits[0],val_lable[0])
val_logits=np.argmax(val_logits,1)
print(val_logits-val_lable)
plt.subplot(211)
plt.plot(total_loss,'-r')
plt.subplot(212)
plt.plot(total_acc,'-y')
plt.plot(total_test_acc,'-g')
plt.pause(0.1)
# check_weights( sess,'conv1','weight')
if step % 500 == 0 or (step + 1) == MAX_STEP:
if(val_acc>max_acc):
max_acc=val_acc
checkpoint_path = os.path.join(logs_train_dir, 'model_'+str(max_acc)+'.ckpt')
saver.save(sess, checkpoint_path, global_step=step)
save_list(total_acc,'acc_googlenet.txt')
save_list(total_test_acc,'test_acc_googlenet.txt')
save_list(total_loss,'loss_googlenet.txt')
except tf.errors.OutOfRangeError:
print('Done training epoch limit reached')
finally:
coord.request_stop()
coord.join(threads)
sess.close()
def test():
log_dir='output/'
is_training = tf.placeholder(tf.bool, shape=())
train_img, train_lable = data_buscrowd.get_train_data(batch_size=BATCH_SIZE)
test_img, test_lable = data_buscrowd.get_test_data(batch_size=BATCH_SIZE)
x = tf.cond(is_training, lambda: train_img, lambda: test_img)
y = tf.cond(is_training, lambda: train_lable, lambda: test_lable)
logits = model(x, num_classes=4)
losses = ops.loss_sparse_softmax_cross_entropy(logits, y)
acc = ops.evaluation(logits, y)
saver = tf.train.Saver()
with tf.Session() as sess:
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
ckpt = tf.train.get_checkpoint_state(log_dir)
if ckpt and ckpt.model_checkpoint_path:
global_step = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1]
saver.restore(sess, ckpt.model_checkpoint_path)
else:
print('No checkpoint')
acc_total=[]
cls_matrix=np.zeros([4,4])
try:
for i in range(20):
val_loss,val_acc,val_logits,val_y = sess.run([losses,acc,logits,y], feed_dict={is_training: False})
max_index = np.argmax(val_logits,1)
for j in range(32):
cls_matrix[val_y[j],max_index[j]]+=1
acc_total.append(val_acc)
print(val_acc)
except tf.errors.OutOfRangeError:
print('Done test epoch limit reached')
finally:
coord.request_stop()
coord.join(threads)
sess.close()
return np.average(acc_total),cls_matrix
if __name__=='__main__':
train()
# print(test())