-
Notifications
You must be signed in to change notification settings - Fork 63
/
Copy pathtrain.py
204 lines (169 loc) · 9.39 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
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
import os
import Net
import numpy as np
import Data
import config as cfg
import tensorflow as tf
slim = tf.contrib.slim
import matplotlib
matplotlib.use('TkAgg')
class Solver(object):
def __init__(self, net, data, is_training=True):
self.net = net
self.data = data
self.max_iter = cfg.Max_iter
self.save_iter = cfg.Save_iter
self.summary_iter = cfg.Summary_iter
self.output_dir = './Output'
if not os.path.exists(self.output_dir):
os.makedirs(self.output_dir)
self.ckpt_file = os.path.join(self.output_dir, 'save.ckpt')
self.weights_file = cfg.Weights_file
self.summary_op = tf.summary.merge_all()
self.writer = tf.summary.FileWriter(self.output_dir)
exclude = ['fc_6/weights', 'fc_7/weights', 'fc_8/weights', 'fc_9/weights',
'fc_6/biases', 'fc_7/biases', 'fc_8/biases',
'fc_9/biases', 'is_training']
self.variable_to_restore = slim.get_variables_to_restore(exclude=exclude)
self.restorer = tf.train.Saver(self.variable_to_restore, max_to_keep=1)
self.saver = tf.train.Saver(max_to_keep=1)
if is_training:
self.initial_learning_rate = cfg.Initial_learning_rate
self.decay_rate = cfg.Decay_rate
self.decay_iter = cfg.Decay_iter
self.staircase = cfg.Staircase
self.global_step = tf.get_variable('global_step', [], initializer=tf.constant_initializer(0),
trainable=False)
self.learning_rate = tf.train.exponential_decay(
self.initial_learning_rate,
self.global_step,
self.decay_iter,
self.decay_rate,
self.staircase,
name='learning_rate'
)
self.optimizer = tf.train.AdamOptimizer(learning_rate=self.learning_rate).minimize(
self.net.total_loss, global_step=self.global_step)
self.ema = tf.train.ExponentialMovingAverage(0.9)
self.average_op = self.ema.apply(tf.trainable_variables())
with tf.control_dependencies([self.optimizer]):
self.train_op = tf.group(self.average_op)
self.sess = tf.Session()
self.sess.run(tf.global_variables_initializer())
if self.weights_file is not None:
self.restorer.restore(self.sess, self.weights_file)
print('loading the model weight...')
self.writer.add_graph(self.sess.graph)
self.save_cfg()
def save_cfg(self):
with open(os.path.join(self.output_dir, 'config.txt'), 'w') as f:
cfg_dict = cfg.__dict__
for key in sorted(cfg_dict.keys()):
if key[0].isupper():
cfg_str = '{}: {}\n'.format(key, cfg_dict[key])
f.write(cfg_str)
def train(self):
for step in range(1, self.max_iter + 1):
images, rois, labels = self.data.get_batch()
feed_dict = {self.net.images: images, self.net.rois: rois, self.net.labels: labels}
if step % self.summary_iter == 0:
summary, loss, _ = self.sess.run([self.summary_op, self.net.total_loss, self.train_op],
feed_dict=feed_dict)
self.writer.add_summary(summary, step)
print("Data_epoch:" + str(self.data.epoch) + " " * 5 + "training_step:" + str(
step) + " " * 5 + "batch_loss:" + str(loss))
else:
self.sess.run([self.train_op], feed_dict=feed_dict)
if step % self.save_iter == 0:
print("saving the model as : " + self.ckpt_file)
self.saver.save(self.sess, self.ckpt_file, global_step=self.global_step)
def predict(self):
class_num = cfg.Class_num
image_path = cfg.Images_path
with open(r'./Data/test_list.txt','r') as f:
test_index_collect = f.readlines()
for test_index in test_index_collect:
test_index = test_index.strip()
regions = []
results_dic = {}
for cls_index in range(1, class_num+1):
results_dic[cls_index] = []
images, roises, labels = self.data.get_valid_batch(test_index)
feed_dict = {self.net.images: images, self.net.rois: roises}
results, bbox_ = self.sess.run([self.net.logits, self.net.bbox], feed_dict=feed_dict)
print('********************************************')
for index in range(len(results)):
if (np.argmax(results[index][0:class_num])) != 0 and np.max(results[index][0:class_num] > 0.5):
rois = roises[index][1:5] * 16
regions_old = [(rois[0] + rois[2]) / 2.0, (rois[1] + rois[3]) / 2.0, rois[2] - rois[0],
rois[3] - rois[1]]
ind = np.argmax(results[index]) - 1
score = np.max(results[index][0:5])
x_rate, y_rate, w_rate, h_rate = bbox_[index][ind * 4], bbox_[index][ind * 4 + 1], \
bbox_[index][ind * 4 + 2], bbox_[index][ind * 4 + 3]
region_new = [regions_old[0] + regions_old[2] * x_rate, regions_old[1] + regions_old[3] * y_rate,
regions_old[2] * np.exp(w_rate), regions_old[3] * np.exp(h_rate)]
results_dic[ind + 1].append(
[results[index][ind + 1], region_new[0] - region_new[2] / 2.0, region_new[1] - region_new[3] / 2.0,
region_new[0] + region_new[2] / 2.0, region_new[1] + region_new[3] / 2.0])
regions.append([region_new[0] - region_new[2] / 2.0, region_new[1] - region_new[3] / 2.0, region_new[2],
region_new[3], ind + 1, score])
if len(regions) != 0:
print(image_path+test_index+'.jpg')
Data.show_rect(image_path+test_index+'.jpg', regions, test_index)
else:
print('There is no target')
NMS_results = self.NMS(results_dic)
if len(NMS_results) != 0:
Data.show_rect(image_path+test_index+'.jpg', NMS_results, test_index )
NMS_average_results = self.NMS_average(results_dic)
if len(NMS_average_results) != 0:
Data.show_rect(image_path+test_index+'.jpg', NMS_average_results, test_index )
def NMS_IOU(self, vertice1, vertice2): # verticle:[pro,xin,ymin,xmax,ymax]
lu = np.maximum(vertice1[1:3], vertice2[1:3])
rd = np.minimum(vertice1[3:], vertice2[3:])
intersection = np.maximum(0.0, rd - lu)
inter_square = intersection[0] * intersection[1]
square1 = (vertice1[3] - vertice1[1]) * (vertice1[4] - vertice1[2])
square2 = (vertice2[3] - vertice2[1]) * (vertice2[4] - vertice2[2])
union_square = np.maximum(square1 + square2 - inter_square, 1e-10)
return np.clip(inter_square / union_square, 0.0, 1.0)
def NMS(self, result_dic):
final_result = []
for cls_ind, cls_collect in result_dic.items():
cls_collect = sorted(cls_collect, reverse=True)
for i in range(len(cls_collect) - 1):
for j in range(len(cls_collect) - 1, i, -1):
if self.NMS_IOU(cls_collect[i], cls_collect[j]) > cfg.NMS_threshold:
del cls_collect[j]
for each_result in cls_collect:
final_result.append(
[each_result[1], each_result[2], each_result[3] - each_result[1], each_result[4] - each_result[2],
cls_ind, each_result[0]])
return final_result
def NMS_average(self, result_dic):
final_result = []
for cls_ind, cls_collect in result_dic.items():
cls_collect = sorted(cls_collect, reverse=True)
for i in range(len(cls_collect) - 1):
for j in range(len(cls_collect) - 1, i, -1):
if self.NMS_IOU(cls_collect[i], cls_collect[j]) > cfg.NMS_threshold:
cls_collect[i] = [(x + y) / 2.0 for (x, y) in zip(cls_collect[i], cls_collect[j])]
del cls_collect[j]
for each_result in cls_collect:
final_result.append(
[each_result[1], each_result[2], each_result[3] - each_result[1], each_result[4] - each_result[2],
cls_ind, each_result[0]])
return final_result
if __name__ == '__main__':
train = True
if train:
net = Net.Alexnet(is_training = True)
data = Data.data()
solver = Solver(net, data, is_training = True)
solver.train()
else:
net = Net.Alexnet(is_training=False)
data = Data.data()
solver = Solver(net, data, is_training=False)
solver.predict()