-
Notifications
You must be signed in to change notification settings - Fork 5
/
train_SR.py
204 lines (166 loc) · 9.76 KB
/
train_SR.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 time
import numpy as np
import tensorflow as tf
from tensorflow.contrib import slim
import cv2
from dataset import SR_data_load
import SR_models
import losses
import utils
tf.app.flags.DEFINE_string('run_gpu', '0', 'use single gpu')
tf.app.flags.DEFINE_string('save_path', '/where/your/folder', '')
tf.app.flags.DEFINE_boolean('model_restore', False, '')
tf.app.flags.DEFINE_string('image_path', '/where/your/saved/image/folder', '')
tf.app.flags.DEFINE_integer('batch_size', 32, '')
tf.app.flags.DEFINE_integer('num_readers', 4, '')
tf.app.flags.DEFINE_integer('input_size', 32, '')
tf.app.flags.DEFINE_float('learning_rate', 0.0001, 'define your learing strategy')
tf.app.flags.DEFINE_float('moving_average_decay', 0.997, '')
tf.app.flags.DEFINE_string('vgg_path', None, '/where/your/vgg_19.ckpt')
tf.app.flags.DEFINE_integer('num_workers', 4, '')
tf.app.flags.DEFINE_integer('max_to_keep', 10, 'how many do you want to save models?')
tf.app.flags.DEFINE_integer('save_model_steps', 10000, '')
tf.app.flags.DEFINE_integer('save_summary_steps', 10, '')
tf.app.flags.DEFINE_integer('max_steps', 1000000, '')
tf.app.flags.DEFINE_string('losses', 'perceptual', 'mse,perceptual,texture,adv')
tf.app.flags.DEFINE_string('adv_direction', 'g2d', 'g2d or d2g')
tf.app.flags.DEFINE_float('adv_gen_w', 0.001, '')
tf.app.flags.DEFINE_float('adv_disc_w', 1.0, '')
FLAGS = tf.app.flags.FLAGS
def main(argv=None):
######################### System setup
os.environ['CUDA_VISIBLE_DEVICES'] = FLAGS.run_gpu
utils.prepare_checkpoint_path(FLAGS.save_path, FLAGS.model_restore)
######################### Model setup
low_size = FLAGS.input_size
high_size = int(FLAGS.input_size * FLAGS.SR_scale)
input_low_images = tf.placeholder(tf.float32, shape=[FLAGS.batch_size, low_size, low_size, 3], name='input_low_images')
input_high_images = tf.placeholder(tf.float32, shape=[FLAGS.batch_size, high_size, high_size, 3], name='input_high_images')
model_builder = SR_models.model_builder()
generated_high_images, resized_low_images = model_builder.generator(input_low_images, is_training=True, model=FLAGS.model)
tf.summary.image('1_input_low_images', input_low_images)
tf.summary.image('2_input_high_images', input_high_images)
vis_gen_images = tf.cast(tf.clip_by_value(generated_high_images, 0, 255), tf.uint8)
tf.summary.image('3_generated_images', vis_gen_images)
tf.summary.image('4_bicubic_images', resized_low_images)
vis_high_images = tf.cast(input_high_images, tf.uint8)
vis_gen_images = tf.concat([resized_low_images, vis_gen_images, vis_high_images], axis=2)
tf.summary.image('5_bicubic_gen_gt', vis_gen_images)
######################### Losses setup
loss_builder = losses.loss_builder()
loss_list = utils.loss_parser(FLAGS.losses)
generator_loss = 0.0
if 'mse' in loss_list or 'l2' in loss_list or 'l2_loss' in loss_list:
mse_loss = loss_builder.get_loss(input_high_images, generated_high_images, type='mse')
generator_loss = generator_loss + 1.0 * mse_loss
tf.summary.scalar('mse_loss', mse_loss)
if 'inverse_mse' in loss_list:
inv_mse_loss = loss_builder.get_loss(input_low_images, generated_high_images, type='inverse_mse')
generator_loss = generator_loss + 100.0 * inv_mse_loss
tf.summary.scalar('inv_mse_loss', inv_mse_loss)
if 'fft_mse' in loss_list:
fft_mse_loss = loss_builder.get_loss(input_high_images, generated_high_images, type='fft_mse')
generator_loss = generator_loss + 1.0 * fft_mse_loss
tf.summary.scalar('fft_mse_loss', fft_mse_loss)
if 'l1' in loss_list or 'l1_loss' in loss_list:
l1_loss = loss_builder.get_loss(input_high_images, generated_high_images, type='l1_loss')
generator_loss = generator_loss + 0.01 * l1_loss
tf.summary.scalar('l1_loss', l1_loss)
if 'perceptual' in loss_list:
pl_pool5 = loss_builder.get_loss(input_high_images, generated_high_images, type='perceptual')
pl_pool5 *= 2e-2
generator_loss = generator_loss + pl_pool5
tf.summary.scalar('pl_pool5', pl_pool5)
if 'texture' in loss_list:
tl_conv1, tl_conv2, tl_conv3 = loss_builder.get_loss(input_high_images, generated_high_images, type='texture')
#generator_loss = generator_loss + 1e-2 * tl_conv1 + 1e-2 * tl_conv2 + 1e-2 * tl_conv3
tl_weight = 10.0
#generator_loss = generator_loss + tl_weight * tl_conv1 + tl_weight * tl_conv2 + tl_weight * tl_conv3
generator_loss = generator_loss + tl_weight * tl_conv3
tf.summary.scalar('tl_conv1', tl_conv1)
tf.summary.scalar('tl_conv2', tl_conv2)
tf.summary.scalar('tl_conv3', tl_conv3)
if 'adv' in loss_list:
adv_gen_loss, adv_disc_loss = loss_builder.get_loss(input_high_images, generated_high_images, type='adv')
tf.summary.scalar('adv_gen', adv_gen_loss)
tf.summary.scalar('adv_disc', adv_disc_loss)
discrim_loss = FLAGS.adv_disc_w * adv_disc_loss
generator_loss = generator_loss + FLAGS.adv_gen_w * adv_gen_loss
tf.summary.scalar('generator_loss', generator_loss)
######################### Training setup
global_step = tf.get_variable('global_step', [], dtype=tf.int64, initializer=tf.constant_initializer(0), trainable=False)
train_vars = tf.trainable_variables()
update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS)
learning_rate = utils.configure_learning_rate(FLAGS.learning_rate, global_step)
#gen_optimizer = utils.configure_optimizer(learning_rate)
#gen_gradients = gen_optimizer.compute_gradients(generator_loss, var_list=generator_vars)
#gen_grad_updates = gen_optimizer.apply_gradients(gen_gradients)#, global_step=global_step)
if 'adv' in loss_list:
discrim_vars = [var for var in train_vars if var.name.startswith('discriminator')]
disc_optimizer = utils.configure_optimizer(learning_rate)
disc_gradients = disc_optimizer.compute_gradients(discrim_loss, var_list=discrim_vars)
disc_grad_updates = disc_optimizer.apply_gradients(disc_gradients, global_step=global_step)
with tf.control_dependencies([disc_grad_updates] + update_ops):
generator_vars = [var for var in train_vars if var.name.startswith('generator')]
gen_optimizer = utils.configure_optimizer(learning_rate)
gen_gradients = gen_optimizer.compute_gradients(generator_loss, var_list=generator_vars)
gen_grad_updates = gen_optimizer.apply_gradients(gen_gradients, global_step=global_step)
train_op = gen_grad_updates
else:
generator_vars = [var for var in train_vars if var.name.startswith('generator')]
discrim_vars = generator_vars
gen_optimizer = utils.configure_optimizer(learning_rate)
gen_gradients = gen_optimizer.compute_gradients(generator_loss, var_list=generator_vars)
gen_grad_updates = gen_optimizer.apply_gradients(gen_gradients)#, global_step=global_step)
with tf.control_dependencies([gen_grad_updates] + update_ops):
train_op = tf.no_op(name='train_op')
saver = tf.train.Saver(max_to_keep=FLAGS.max_to_keep)
summary_op = tf.summary.merge_all()
summary_writer = tf.summary.FileWriter(FLAGS.save_path, tf.get_default_graph())
######################### Train process
data_generator = SR_data_load.get_batch(image_path=FLAGS.image_path,
num_workers=FLAGS.num_workers,
batch_size=FLAGS.batch_size,
hr_size=high_size)
## vgg_stop process
#utils.print_vars(train_vars)
#utils.print_vars(generator_vars)
if FLAGS.vgg_path is not None:
variable_restore_op = utils.get_restore_op(FLAGS.vgg_path, train_vars)
#############
with tf.Session(config=tf.ConfigProto(allow_soft_placement=True)) as sess:
if FLAGS.model_restore:
ckpt = tf.train.latest_checkpoint(FLAGS.save_path)
saver.restore(sess, ckpt)
else:
sess.run(tf.global_variables_initializer())
if FLAGS.vgg_path is not None:
variable_restore_op(sess)
start_time = time.time()
for iter_val in range(int(global_step.eval()) + 1, FLAGS.max_steps + 1):
data = next(data_generator)
high_images = np.asarray(data[0])
low_images = np.asarray(data[1])
feed_dict = {input_low_images: low_images,
input_high_images: high_images}
generator_loss_val, _, g_w, d_w = sess.run([generator_loss, train_op, generator_vars[0], discrim_vars[0]], feed_dict=feed_dict)
if iter_val != 0 and iter_val % FLAGS.save_summary_steps == 0:
summary_str = sess.run(summary_op, feed_dict=feed_dict)
summary_writer.add_summary(summary_str, global_step=iter_val)
used_time = time.time() - start_time
avg_time_per_step = used_time / FLAGS.save_summary_steps
avg_examples_per_second = (FLAGS.save_summary_steps * FLAGS.batch_size) / used_time
print('step %d, generator_loss %.4f, weights %.2f, %.2f, %.2f seconds/step, %.2f examples/second'
% (iter_val, generator_loss_val, np.sum(g_w), np.sum(d_w), avg_time_per_step, avg_examples_per_second))
start_time = time.time()
if iter_val != 0 and iter_val % FLAGS.save_model_steps == 0:
checkpoint_fn = os.path.join(FLAGS.save_path, 'model.ckpt')
saver.save(sess, checkpoint_fn, global_step=iter_val)
print('')
print('*' * 30)
print(' Training done!!! ')
print('*' * 30)
print('')
if __name__ == '__main__':
tf.app.run()