-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmain_gan_train.py
273 lines (221 loc) · 12.5 KB
/
main_gan_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
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
#! /usr/bin/python
# -*- coding: utf8 -*-
import matplotlib.pyplot as plt
import os, time, pickle, random
import numpy as np
import tensorflow as tf
import tensorlayer as tl
from model import *
from utils import *
from config import config
## Adam
batch_size = config.TRAIN.batch_size
lr_init = config.TRAIN.lr_init
beta1 = config.TRAIN.beta1
## initialize G
n_epoch_init = 20
## adversarial learning
n_epoch = 4
lr_decay = config.TRAIN.lr_decay
decay_every = 2
patch_size_l = 74
patch_size_h = 296
ni = int(np.sqrt(batch_size))
def modcrop(imgs, modulo):
tmpsz = imgs.shape
sz = tmpsz[0:2]
h = sz[0] - sz[0]%modulo
w = sz[1] - sz[1]%modulo
imgs = imgs[0:h+1, 0:w+1,:]
return imgs
def read_all_imgs(img_list, path='', n_threads=32):
""" Returns all images in array by given path and name of each image file. """
imgs = []
for idx in range(0, len(img_list), n_threads):
b_imgs_list = img_list[idx : idx + n_threads]
b_imgs = tl.prepro.threading_data(b_imgs_list, fn=get_imgs_fn, path=path)
# print(b_imgs.shape)
imgs.extend(b_imgs)
print('read %d from %s' % (len(imgs), path))
return imgs
def train():
## create folders to save result images and trained model
save_dir_gan = config.TRAIN.save_valid_results
tl.files.exists_or_mkdir(save_dir_gan)
checkpoint_dir = config.TRAIN.checkpoint
tl.files.exists_or_mkdir(checkpoint_dir)
###====================== PRE-LOAD DATA ===========================###
train_hr_img_list = sorted(tl.files.load_file_list(path=config.TRAIN.hr_img_path, regx='.*.png', printable=False))
train_lr_img_list = sorted(tl.files.load_file_list(path=config.TRAIN.lr_img_path, regx='.*.png', printable=False))
# valid_hr_img_list = sorted(tl.files.load_file_list(path=config.VALID.hr_img_path, regx='.*.bmp', printable=False))
# valid_lr_img_list = sorted(tl.files.load_file_list(path=config.VALID.lr_img_path, regx='.*.bmp', printable=False))
## If your machine have enough memory, please pre-load the whole train set.
# train_hr_imgs = read_all_imgs(train_hr_img_list, path=config.TRAIN.hr_img_path, n_threads=32)
# for im in train_hr_imgs:
# print(im.shape)
# valid_lr_imgs = read_all_imgs(valid_lr_img_list, path=config.VALID.lr_img_path, n_threads=32)
# for im in valid_lr_imgs:
# print(im.shape)
# valid_hr_imgs = read_all_imgs(valid_hr_img_list, path=config.VALID.hr_img_path, n_threads=32)
# for im in valid_hr_imgs:
# print(im.shape)
# exit()
###========================== DEFINE MODEL ============================###
## train inference
t_image = tf.placeholder('float32', [batch_size, patch_size_l, patch_size_l, 3], name='t_image_input_to_generator')
t_target_image = tf.placeholder('float32', [batch_size, patch_size_h, patch_size_h, 3], name='t_target_image')
# Generator
net_g= SRGAN_g(t_image, is_train=False, reuse=False)
# Discriminator
net_d, logits_real = SRGAN_d(t_target_image, is_train=True, reuse=False)
_, logits_fake = SRGAN_d(net_g.outputs, is_train=True, reuse=True)
# VGG network
net_vgg, vgg_target_emb= Vgg19_simple_api((t_target_image+1)/2, reuse=False)
_, vgg_predict_emb = Vgg19_simple_api((net_g.outputs+1)/2, reuse=True)
# Feature Discriminator
vgg_scale5 = 1/12.75
# vgg_scale4 = 1.4e-3
# vgg_scale2 = 1e-3
net_vgg_d, logits_vgg_real = SRGAN_vgg_d(vgg_scale5*vgg_target_emb.outputs, is_train=True, reuse=False)
_, logits_vgg_fake = SRGAN_vgg_d(vgg_scale5*vgg_predict_emb.outputs, is_train=True, reuse=True)
#########
## validation
net_g_test = SRGAN_g(t_image, is_train=False, reuse=True)
# ###========================== DEFINE TRAIN OPS ==========================###
#Discriminator loss
d_loss1 = 1e-3 *(tl.cost.sigmoid_cross_entropy(logits_real, tf.ones_like(logits_real), name='d1'))
d_loss2 = 1e-3 *(tl.cost.sigmoid_cross_entropy(logits_fake, tf.zeros_like(logits_fake), name='d2'))
d_vgg_loss1 = 1e-3 *(tl.cost.sigmoid_cross_entropy(logits_vgg_real, tf.ones_like(logits_vgg_real), name='d_vgg_1'))
d_vgg_loss2 = 1e-3 *(tl.cost.sigmoid_cross_entropy(logits_vgg_fake, tf.zeros_like(logits_vgg_fake), name='d_vgg_2'))
d_loss = d_loss1 + d_loss2 + d_vgg_loss1 + d_vgg_loss2
##
d_loss1_summary = tf.summary.scalar('d_loss1', d_loss1)
d_loss2_summary = tf.summary.scalar('d_loss2', d_loss2)
d_vgg_loss1_summary = tf.summary.scalar('d_vgg_loss1', d_vgg_loss1)
d_vgg_loss2_summary = tf.summary.scalar('d_vgg_loss2', d_vgg_loss2)
merged_d = tf.summary.merge([d_loss1_summary, d_loss2_summary,d_vgg_loss1_summary,d_vgg_loss2_summary])
##
#GAN loss
g_gan_loss = 1e-3 * tl.cost.sigmoid_cross_entropy(logits_fake, tf.ones_like(logits_fake), name='g')
g_gan_vgg_loss = 1e-3 * tl.cost.sigmoid_cross_entropy(logits_vgg_fake, tf.ones_like(logits_vgg_fake), name='g')
#vgg loss
vgg_loss = tl.cost.mean_squared_error(vgg_scale5*vgg_predict_emb.outputs, vgg_scale5*vgg_target_emb.outputs, is_mean=True) # weight..? feature map rescale?
##
vgg_summary = tf.summary.scalar('vgg_loss', vgg_loss)
g_gan_summary = tf.summary.scalar('g_gan_loss', g_gan_loss)
g_gan_vgg_summary = tf.summary.scalar('g_gan_vgg_loss', g_gan_vgg_loss)
merged_g = tf.summary.merge([vgg_summary, g_gan_summary, g_gan_vgg_summary])
##
#Total loss
g_loss = vgg_loss + g_gan_loss + g_gan_vgg_loss
g_vars = tl.layers.get_variables_with_name('SRGAN_g', True, True)
d_vars = tl.layers.get_variables_with_name('SRGAN_d', True, True)
d_vgg_vars = tl.layers.get_variables_with_name('SRGAN_vgg_d', True, True)
with tf.variable_scope('learning_rate'):
lr_v = tf.Variable(lr_init, trainable=False)
## SRGAN
g_optim = tf.train.AdamOptimizer(lr_v, beta1=beta1, epsilon=1e-10).minimize(g_loss, var_list=g_vars)
d_optim = tf.train.AdamOptimizer(lr_v, beta1=beta1, epsilon=1e-10).minimize(d_loss, var_list=[d_vars, d_vgg_vars])
###========================== RESTORE MODEL =============================###
sess = tf.Session(config=tf.ConfigProto(allow_soft_placement=True, log_device_placement=False))
summary_writer = tf.summary.FileWriter(config.TRAIN.summary_adv,sess.graph)
summary_writer.add_graph(sess.graph)
tl.layers.initialize_global_variables(sess)
if tl.files.load_and_assign_npz(sess=sess, name=checkpoint_dir+'/g_{}.npz'.format('SRFeat'), network=net_g) is False:
tl.files.load_and_assign_npz(sess=sess, name=checkpoint_dir+'/g_{}_init.npz'.format('SRFeat'), network=net_g)
tl.files.load_and_assign_npz(sess=sess, name=checkpoint_dir+'/d_{}.npz'.format('SRFeat'), network=net_d)
###============================= LOAD VGG ===============================###
vgg19_npy_path = "vgg19.npy"
if not os.path.isfile(vgg19_npy_path):
print("Please download vgg19.npz from : https://github.com/machrisaa/tensorflow-vgg")
exit()
npz = np.load(vgg19_npy_path, encoding='latin1').item()
#
params = []
count_layers =0
for val in sorted( npz.items() ):
if(count_layers<16):
W = np.asarray(val[1][0])
b = np.asarray(val[1][1])
print(" Loading %s: %s, %s" % (val[0], W.shape, b.shape))
params.extend([W, b])
count_layers += 1
tl.files.assign_params(sess, params, net_vgg)
###============================= TRAINING ===============================###
## use first `batch_size` of train set to have a quick test during training
sample_list = [0,10001,20001,30001,40001,50001,60001,70001,80001]
sample_imgs = read_all_imgs([train_hr_img_list[k] for k in sample_list], path=config.TRAIN.hr_img_path, n_threads=batch_size) # if no pre-load train set
sample_imgs_384 = tl.prepro.threading_data(sample_imgs, fn=scale_imgs_fn)
sample_imgs_96 = read_all_imgs([train_lr_img_list[k] for k in sample_list], path=config.TRAIN.lr_img_path, n_threads=batch_size)
sample_imgs_96 = tl.prepro.threading_data(sample_imgs_96, fn=scale_imgs_fn)
tl.vis.save_images(sample_imgs_96, [ni, ni], save_dir_gan+'/_train_sample_96.png')
tl.vis.save_images(sample_imgs_384, [ni, ni], save_dir_gan+'/_train_sample_384.png')
###========================= train GAN (SRGAN) =========================###
iters =0
lr_hr_list = list(zip(train_hr_img_list,train_lr_img_list))
random.shuffle(lr_hr_list)
train_hr_img_list, train_lr_img_list = zip(*lr_hr_list)
for epoch in range(0, n_epoch+1):
## update learning rate
if epoch !=0 and (epoch % decay_every == 0):
new_lr_decay = lr_decay ** (epoch // decay_every)
sess.run(tf.assign(lr_v, lr_init * new_lr_decay))
log = " ** new learning rate: %f (for GAN)" % (lr_init * new_lr_decay)
print(log)
elif epoch == 0:
sess.run(tf.assign(lr_v, lr_init))
log = " ** init lr: %f decay_every_init: %d, lr_decay: %f (for GAN)" % (lr_init, decay_every, lr_decay)
print(log)
epoch_time = time.time()
total_d_loss, total_g_loss, n_iter = 0, 0, 0
## If your machine cannot load all images into memory, you should use
## this one to load batch of images while training.
# random.shuffle(train_hr_img_list)
for idx in range(0, len(train_hr_img_list)-batch_size, batch_size):
iters = iters+1
b_imgs_list = train_hr_img_list[idx : idx + batch_size]
b_imgs_list_lr = train_lr_img_list[idx : idx + batch_size]
b_imgs = tl.prepro.threading_data(b_imgs_list, fn=get_imgs_fn, path=config.TRAIN.hr_img_path)
b_imgs_96 = tl.prepro.threading_data(b_imgs_list_lr, fn=get_imgs_fn, path=config.TRAIN.lr_img_path)
b_imgs_384 = tl.prepro.threading_data(b_imgs, fn=scale_imgs_fn)
b_imgs_96 = tl.prepro.threading_data(b_imgs_96, fn=scale_imgs_fn)
## If your machine have enough memory, please pre-load the whole train set.
# for idx in range(0, len(train_hr_imgs), batch_size):
step_time = time.time()
# b_imgs_384 = tl.prepro.threading_data(
# train_hr_imgs[idx : idx + batch_size],
# fn=crop_sub_imgs_fn, is_random=True,is_thread=True)
# b_imgs_96 = tl.prepro.threading_data(b_imgs_384, fn=downsample_fn)
#
summary_d, errD, errD1, errD2, errD3, errD4, _ = sess.run([merged_d, d_loss, d_loss1, d_loss2, d_vgg_loss1, d_vgg_loss2, d_optim], {t_image: b_imgs_96, t_target_image: b_imgs_384})
## update G
summary_g, errG, errV, errA, errA2, _ = sess.run([merged_g, g_loss, vgg_loss, g_gan_loss, g_gan_vgg_loss, g_optim], {t_image: b_imgs_96, t_target_image: b_imgs_384})
## summary
summary_writer.add_summary(summary_d, iters)
summary_writer.add_summary(summary_g, iters)
print("Epoch [%2d/%2d] %4d time: %4.4fs, d_loss: %.8f (d1: %.8f, d2: %.8f, d3_vgg: %.8f, d4_vgg: %.8f), g_loss: %.8f (vgg: %.6f adv: %.6f adv2: %.6f)" % (epoch, n_epoch, n_iter, time.time() - step_time, errD, errD1, errD2, errD3, errD4, errG, errV, errA, errA2))
total_d_loss += errD
total_g_loss += errG
n_iter += 1
## quick evaluation on train set
if (iters != 0) and (iters % 100 == 0):
out = sess.run(net_g_test.outputs, {t_image: sample_imgs_96})#; print('gen sub-image:', out.shape, out.min(), out.max())
print("[*] save images")
tl.vis.save_images(out, [ni, ni], save_dir_gan+'/train_%d.png' % iters)
## save model
if (iters != 0) and (iters % 1000 == 0):
tl.files.save_npz(net_g.all_params, name=checkpoint_dir+'/g_{}.npz'.format('SRFeat'), sess=sess)
if (iters != 0) and (iters % 10000 == 0):
tl.files.save_npz(net_d.all_params, name=checkpoint_dir+'/d_{}.npz'.format('SRFeat'), sess=sess)
log = "[*] Epoch: [%2d/%2d] time: %4.4fs, d_loss: %.8f g_loss: %.8f" % (epoch, n_epoch, time.time() - epoch_time, total_d_loss/n_iter, total_g_loss/n_iter)
print(log)
summary_writer.close()
if __name__ == '__main__':
import argparse
parser = argparse.ArgumentParser()
parser.add_argument('--model', type=str, default='srgan')
parser.add_argument('--set',type=str,default='Set5')
args = parser.parse_args()
tl.global_flag['model'] = args.model
tl.global_flag['set'] = args.set
train()