-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathmodel.py
211 lines (156 loc) · 9.91 KB
/
model.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
import os
import time
from glob import glob
import tensorflow as tf
import numpy as np
from collections import namedtuple
from module import *
from utils import *
def load_weights(saver, sess, model_dir):
ckpt = tf.train.get_checkpoint_state(model_dir)
if ckpt and ckpt.model_checkpoint_path:
ckpt_name = os.path.basename(ckpt.model_checkpoint_path)
saver.restore(sess, os.path.join(model_dir, ckpt_name))
return True
else:
return False
def Save(saver, sess, checkpoint_dir, step):
dir = os.path.join(checkpoint_dir, "model")
saver.save(sess, dir)
class cyclegan(object):
def __init__(self, sess):
self.sess = sess
self.batch_size = 1
self.image_size = 256
self.load_size = 286 # scale images to this size
self.fine_size = 256 # then crop to this size
self.train_size = 1e8
self.input_c_dim = 3 # of input image channels
self.output_c_dim = 3 # of output image channels
self.L1_lambda = 10.0
self.dataset_dir = 'flower_images'
self.ndf = 64 # number of generator filters in first conv layer
self.ngf = 64 # number of discriminator filters in first conv layer
self.phase = 'train'
self.discriminator = discriminator
self.generator = generator_resnet
self.criterionGAN = mae_criterion
OPTIONS = namedtuple('OPTIONS', 'batch_size image_size \
gf_dim df_dim output_c_dim is_training')
self.options = OPTIONS._make((self.batch_size, self.fine_size,
self.ngf, self.ndf, self.output_c_dim, self.phase == 'train'))
self._build_model()
self.saver = tf.train.Saver()
self.pool = ImagePool(50)
def _build_model(self):
self.real_data = tf.placeholder(tf.float32,
[None, self.image_size, self.image_size,
self.input_c_dim + self.output_c_dim],
name='real_A_and_B_images')
self.real_A = self.real_data[:, :, :, :self.input_c_dim]
self.real_B = self.real_data[:, :, :, self.input_c_dim:self.input_c_dim + self.output_c_dim]
self.fake_B = self.generator(self.real_A, self.options, False, name="generatorA2B")
self.fake_A_ = self.generator(self.fake_B, self.options, False, name="generatorB2A")
self.fake_A = self.generator(self.real_B, self.options, True, name="generatorB2A")
self.fake_B_ = self.generator(self.fake_A, self.options, True, name="generatorA2B")
self.DB_fake = self.discriminator(self.fake_B, self.options, reuse=False, name="discriminatorB")
self.DA_fake = self.discriminator(self.fake_A, self.options, reuse=False, name="discriminatorA")
self.g_loss_a2b = self.criterionGAN(self.DB_fake, tf.ones_like(self.DB_fake)) \
+ self.L1_lambda * abs_criterion(self.real_A, self.fake_A_) \
+ self.L1_lambda * abs_criterion(self.real_B, self.fake_B_)
self.g_loss_b2a = self.criterionGAN(self.DA_fake, tf.ones_like(self.DA_fake)) \
+ self.L1_lambda * abs_criterion(self.real_A, self.fake_A_) \
+ self.L1_lambda * abs_criterion(self.real_B, self.fake_B_)
self.g_loss = self.criterionGAN(self.DA_fake, tf.ones_like(self.DA_fake)) \
+ self.criterionGAN(self.DB_fake, tf.ones_like(self.DB_fake)) \
+ self.L1_lambda * abs_criterion(self.real_A, self.fake_A_) \
+ self.L1_lambda * abs_criterion(self.real_B, self.fake_B_)
self.fake_A_sample = tf.placeholder(tf.float32,
[None, self.image_size, self.image_size,
self.input_c_dim], name='fake_A_sample')
self.fake_B_sample = tf.placeholder(tf.float32,
[None, self.image_size, self.image_size,
self.output_c_dim], name='fake_B_sample')
self.DB_real = self.discriminator(self.real_B, self.options, reuse=True, name="discriminatorB")
self.DA_real = self.discriminator(self.real_A, self.options, reuse=True, name="discriminatorA")
self.DB_fake_sample = self.discriminator(self.fake_B_sample, self.options, reuse=True, name="discriminatorB")
self.DA_fake_sample = self.discriminator(self.fake_A_sample, self.options, reuse=True, name="discriminatorA")
self.db_loss_real = self.criterionGAN(self.DB_real, tf.ones_like(self.DB_real))
self.db_loss_fake = self.criterionGAN(self.DB_fake_sample, tf.zeros_like(self.DB_fake_sample))
self.db_loss = (self.db_loss_real + self.db_loss_fake) / 2
self.da_loss_real = self.criterionGAN(self.DA_real, tf.ones_like(self.DA_real))
self.da_loss_fake = self.criterionGAN(self.DA_fake_sample, tf.zeros_like(self.DA_fake_sample))
self.da_loss = (self.da_loss_real + self.da_loss_fake) / 2
self.d_loss = self.da_loss + self.db_loss
self.g_loss_a2b_sum = tf.summary.scalar("g_loss_a2b", self.g_loss_a2b)
self.g_loss_b2a_sum = tf.summary.scalar("g_loss_b2a", self.g_loss_b2a)
self.g_loss_sum = tf.summary.scalar("g_loss", self.g_loss)
self.g_sum = tf.summary.merge([self.g_loss_a2b_sum, self.g_loss_b2a_sum, self.g_loss_sum])
self.db_loss_sum = tf.summary.scalar("db_loss", self.db_loss)
self.da_loss_sum = tf.summary.scalar("da_loss", self.da_loss)
self.d_loss_sum = tf.summary.scalar("d_loss", self.d_loss)
self.db_loss_real_sum = tf.summary.scalar("db_loss_real", self.db_loss_real)
self.db_loss_fake_sum = tf.summary.scalar("db_loss_fake", self.db_loss_fake)
self.da_loss_real_sum = tf.summary.scalar("da_loss_real", self.da_loss_real)
self.da_loss_fake_sum = tf.summary.scalar("da_loss_fake", self.da_loss_fake)
self.d_sum = tf.summary.merge([self.da_loss_sum, self.da_loss_real_sum, self.da_loss_fake_sum,
self.db_loss_sum, self.db_loss_real_sum, self.db_loss_fake_sum,
self.d_loss_sum])
t_vars = tf.trainable_variables()
self.d_vars = [var for var in t_vars if 'discriminator' in var.name]
self.g_vars = [var for var in t_vars if 'generator' in var.name]
for var in t_vars: print(var.name)
def train(self):
"""Train cyclegan"""
self.lr = tf.placeholder(tf.float32, None, name='learning_rate')
self.d_optim = tf.train.AdamOptimizer(self.lr, beta1=0.5) \
.minimize(self.d_loss, var_list=self.d_vars)
self.g_optim = tf.train.AdamOptimizer(self.lr, beta1=0.5) \
.minimize(self.g_loss, var_list=self.g_vars)
sum1 = tf.summary.image("realA", self.real_A, max_outputs=1)
sum2 = tf.summary.image("fakeB", self.fake_B, max_outputs=1)
sum3 = tf.summary.image("fakeA_", self.fake_A_, max_outputs=1)
sum4 = tf.summary.image("realB", self.real_B, max_outputs=1)
sum5 = tf.summary.image("fakeA", self.fake_A, max_outputs=1)
sum6 = tf.summary.image("fakeB_", self.fake_B_, max_outputs=1)
self.image = tf.summary.merge([sum1, sum2, sum3, sum4, sum5, sum6])
init_op = tf.global_variables_initializer()
self.sess.run(init_op)
saver = tf.train.Saver()
load_weights(self.saver, self.sess, '/content/drive/My Drive/weights/')
self.writer = tf.summary.FileWriter("./logs", self.sess.graph)
counter = 1
start_time = time.time()
lr = 0.0002
self.epochs = 100
for epoch in range(self.epochs):
dataA = glob('./datasets/{}/*.*'.format(self.dataset_dir + '/trainA'))
dataB = glob('./datasets/{}/*.*'.format(self.dataset_dir + '/trainB'))
np.random.shuffle(dataA)
np.random.shuffle(dataB)
batch_idxs = min(min(len(dataA), len(dataB)), self.train_size) // self.batch_size
for idx in range(0, batch_idxs):
batch_files = list(zip(dataA[idx * self.batch_size:(idx + 1) * self.batch_size],
dataB[idx * self.batch_size:(idx + 1) * self.batch_size]))
batch_images = [load_train_data(batch_file, self.load_size, self.fine_size) for batch_file in batch_files]
batch_images = np.array(batch_images).astype(np.float32)
# Update G network and record fake outputs
fake_A, fake_B, _, summary_str, summary_str2 = self.sess.run(
[self.fake_A, self.fake_B, self.g_optim, self.g_sum, self.image],
feed_dict={self.real_data: batch_images, self.lr: lr})
self.writer.add_summary(summary_str, counter)
self.writer.add_summary(summary_str2, counter)
[fake_A, fake_B] = self.pool([fake_A, fake_B])
# Update D network
_, summary_str = self.sess.run(
[self.d_optim, self.d_sum],
feed_dict={self.real_data: batch_images,
self.fake_A_sample: fake_A,
self.fake_B_sample: fake_B,
self.lr: lr})
self.writer.add_summary(summary_str, counter)
counter += 1
print(("Epoch: [%2d] [%4d/%4d] time: %4.4f" % (
epoch, idx, batch_idxs, time.time() - start_time)))
if idx % 1000 == 0:
Save(self.saver, self.sess, '/content/drive/My Drive/weights/', epoch)