-
Notifications
You must be signed in to change notification settings - Fork 1
/
train.py
231 lines (198 loc) · 7.99 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
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
from __future__ import print_function
import argparse
import math
import os
import time
import numpy as np
import tensorflow as tf
import dataset.reader as reader
import models.resnet_deeplab as resnet_deeplab
from utils.util import INFO, WARN, FAIL
def get_arguments():
"""
define all configurable params
"""
parser = argparse.ArgumentParser(description="Resnet-deeplab train model.")
parser.add_argument(
'--batch_size',
type=int,
default=2,
help='The batch size of each iteration.')
parser.add_argument(
'--epoch_size', type=int, default=50, help='The epoch size of train.')
parser.add_argument(
'--print_step', type=int, default=50, help='The number of print step.')
parser.add_argument(
'--data_dir',
type=str,
default='dataset',
help='The directory of dataset')
parser.add_argument(
'--pretrain_model_path',
type=str,
default=os.path.join('pretrain_model', 'model.ckpt'),
help='The path of pretrained model.')
parser.add_argument(
'--saved_model_dir',
type=str,
default=os.path.join('saved_model'),
help='The path of saved model.')
parser.add_argument(
'--log_dir', type=str, default='logs', help='Directory of log.')
parser.add_argument(
'--num_classes',
type=int,
default=150,
help='The number of class in the dataset.')
parser.add_argument(
'--input_size',
type=str,
default='512x512',
help='The size of input image.')
parser.add_argument(
'--is_training',
action='store_true',
help=
'Whether to update the mean and variance in batch normalization layer.'
)
parser.add_argument(
'--not_restore_fc',
action='store_true',
help='Whether to restore the last fully connected layer.')
parser.add_argument(
'--weight_decay',
type=float,
default='0.0005',
help='Regularisation parameter for L2-loss.')
parser.add_argument(
'--lr', type=float, default='1e-4', help='The base learning rate.')
parser.add_argument(
'--power', type=float, default='0.8', help='Decay for learning rate.')
parser.add_argument(
'--momentum',
type=float,
default='0.9',
help='Momentum component of the optimiser.')
args = parser.parse_args()
return args
def print_arguments(args):
INFO('*' * 15, 'arguments', '*' * 15)
for key, value in vars(args).items():
INFO(key, ":", value)
INFO('*' * 40)
def main():
args = get_arguments()
print_arguments(args)
height, width = map(int, args.input_size.split('x'))
with tf.device('/cpu:0'):
with tf.name_scope('image_reader'):
dataset = reader.Dataset(args.data_dir, subset='train')
dataset.make_batch(
args.batch_size, input_size=(height, width), shuffle=True)
images, labels = dataset.next_batch()
model = resnet_deeplab.Resnet101Deeplab(images, args.num_classes,
args.is_training)
predictions = model.prediction
restore_var = [
v for v in tf.global_variables()
if 'fc' not in v.name or not args.not_restore_fc
]
all_trainable = [
v for v in tf.global_variables()
if 'beta' not in v.name or 'gamma' not in v.name
]
fc_trainable = [v for v in all_trainable if 'fc' in v.name]
conv_trainable = [v for v in all_trainable
if 'fc' not in v.name] # lr * 1.0
fc_w_trainable = [v for v in fc_trainable
if 'weights' in v.name] # lr * 10.0
fc_b_trainable = [v for v in fc_trainable if 'bias' in v.name] # lr *20.0
with tf.name_scope('loss'):
prediction_size = tf.stack(predictions.shape[1:3])
labels = tf.image.resize_nearest_neighbor(labels, prediction_size)
labels = tf.squeeze(
labels, squeeze_dims=[3]) # squeeze the last (color) channel.
labels = tf.reshape(labels, [-1])
# filter the indices which labels exceeds than num class.
boolean_mask = tf.less(labels, args.num_classes)
labels = tf.boolean_mask(labels, boolean_mask)
labels = tf.cast(labels, tf.int32)
predictions = tf.reshape(predictions, [-1, args.num_classes])
predictions = tf.boolean_mask(predictions, boolean_mask)
entropy_loss = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=predictions, labels=labels))
l2_loss = tf.add_n([
args.weight_decay * tf.nn.l2_loss(v)
for v in tf.trainable_variables()
if 'weights' in v.name or 'biases' in v.name
])
loss = entropy_loss + l2_loss
tf.summary.scalar('entropy_loss', entropy_loss)
tf.summary.scalar('l2_loss', l2_loss)
tf.summary.scalar('loss', loss)
num_steps = math.ceil(
dataset.num_examples / args.batch_size) * args.epoch_size
with tf.name_scope('optimization'):
step_ph = tf.placeholder(dtype=tf.float32, shape=())
lr = tf.scalar_mul(args.lr,
tf.pow((1 - step_ph / num_steps), args.power))
opt_conv = tf.train.MomentumOptimizer(lr, args.momentum)
opt_fc_w = tf.train.MomentumOptimizer(lr * 10.0, args.momentum)
opt_fc_b = tf.train.MomentumOptimizer(lr * 20.0, args.momentum)
grads = tf.gradients(loss,
conv_trainable + fc_w_trainable + fc_b_trainable)
grads_conv = grads[:len(conv_trainable)]
grads_fc_w = grads[len(conv_trainable):(
len(conv_trainable) + len(fc_w_trainable))]
grads_fc_b = grads[(len(conv_trainable) + len(fc_w_trainable)):]
train_op_conv = opt_conv.apply_gradients(
zip(grads_conv, conv_trainable))
train_op_fc_w = opt_fc_w.apply_gradients(
zip(grads_fc_w, fc_w_trainable))
train_op_fc_b = opt_fc_b.apply_gradients(
zip(grads_fc_b, fc_b_trainable))
optimization = tf.group(train_op_conv, train_op_fc_w, train_op_fc_b)
if tf.gfile.Exists(args.log_dir):
tf.gfile.DeleteRecursively(args.log_dir)
tf.gfile.MakeDirs(args.log_dir)
# saver for saving and restoring model.
loader = tf.train.Saver(var_list=restore_var)
saver = tf.train.Saver()
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
with tf.Session(config=config) as sess:
sess.run(tf.global_variables_initializer())
loader.restore(sess, args.pretrain_model_path)
merged = tf.summary.merge_all()
train_writer = tf.summary.FileWriter(args.log_dir + '/train',
sess.graph)
start_time = time.time()
for i in range(num_steps):
summary, _ = sess.run(
[merged, optimization], feed_dict={
step_ph: i
})
train_writer.add_summary(summary, i)
if i % args.print_step == 0 and i > 0:
pred, entropy_loss_val, l2_loss_val = sess.run(
[predictions, entropy_loss, l2_loss])
duration = (time.time() - start_time) / i
INFO(
'step {:d} entropy_loss = {:.3f} l2_loss= {:.3f} total_loss= {:.3f} ({:.3f} sec/step)'.
format(i, entropy_loss_val, l2_loss_val,
entropy_loss_val + l2_loss_val, duration))
if np.isnan(entropy_loss_val + l2_loss_val):
for var in all_trainable:
val = sess.run(var)
if np.sum(np.isnan(val)) > 0:
FAIL(var)
FAIL(val)
if i % 1000 == 0:
saver.save(
sess,
os.path.join(args.saved_model_dir, "model.ckpt"),
global_step=i)
train_writer.close()
if __name__ == '__main__':
main()