-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
259 lines (232 loc) · 11.3 KB
/
utils.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
import pandas as pd
import numpy as np
import tensorflow as tf
def bn(x, is_training, scope):
return tf.contrib.layers.batch_norm(x,
decay=0.9,
updates_collections=None,
epsilon=1e-5,
scale=True,
is_training=is_training,
scope=scope)
def concat(tensors, axis, *args, **kwargs):
return tf.concat(tensors, axis, *args, **kwargs)
def like_rgb_label(x, y):
"""Concatenate conditioning vector on feature map axis."""
x_shapes = x.get_shape().as_list()
y_shapes = y.get_shape().as_list()
y = tf.reshape(y, [y_shapes[0],1, 1, y_shapes[-1]])#[batch_size, 1, 1, num_classes]
y_ = tf.ones([x_shapes[0], x_shapes[1], x_shapes[2], y_shapes[-1]])#[batch_size, width, height, num_classes]
return y*y_
def conv2d(input_, output_dim, kernel=(3,3), stride=(2,2),padding='SAME', activation='',use_bn=False, is_training=False, name="conv2d"):
with tf.variable_scope(name):
w = tf.get_variable('w', [kernel[0], kernel[1], input_.get_shape()[-1], output_dim],
initializer=tf.truncated_normal_initializer(stddev=0.02))
conv = tf.nn.conv2d(input_, w, strides=[1, stride[0], stride[1], 1], padding=padding)
biases = tf.get_variable('biases', [output_dim], initializer=tf.constant_initializer(0.0))
hidden = tf.reshape(tf.nn.bias_add(conv, biases), conv.get_shape())
if use_bn:
hidden = bn(x=hidden, is_training=is_training, scope='bn')
if activation == 'relu':
hidden = tf.nn.relu(hidden)
elif activation == 'lrelu':
hidden = lrelu(hidden)
elif activation == 'sigmoid':
hidden = tf.nn.sigmoid(hidden)
return hidden
def deconv2d(input_, output_size, output_channel, kernel=(3,3), stride=(2,2),padding='SAME',
activation='',use_bn=False, is_training=False, name="deconv2d"):
with tf.variable_scope(name):
# filter : [height, width, output_channels, in_channels]
batch_size = input_.get_shape().as_list()[0]
w = tf.get_variable('w', [kernel[0], kernel[1], output_channel, input_.get_shape()[-1]],
initializer=tf.random_normal_initializer(stddev=0.02))
deconv = tf.nn.conv2d_transpose(input_, w, output_shape=[batch_size,output_size,output_size,output_channel],
strides=[1, stride[0], stride[1], 1], padding=padding)
biases = tf.get_variable('biases', [output_channel], initializer=tf.constant_initializer(0.0))
hidden = tf.reshape(tf.nn.bias_add(deconv, biases), deconv.get_shape())
if use_bn:
hidden = bn(x=hidden, is_training=is_training, scope='bn')
if activation == 'relu':
hidden = tf.nn.relu(hidden)
elif activation == 'lrelu':
hidden = lrelu(hidden)
elif activation == 'sigmoid':
hidden = tf.nn.sigmoid(hidden)
return hidden
def lrelu(x, leak=0.2, name="lrelu"):
return tf.maximum(x, leak*x)
def linear(input_, output_size, scope=None):
shape = input_.get_shape().as_list()
with tf.variable_scope(scope or "Linear"):
matrix = tf.get_variable("Matrix", [shape[1], output_size], tf.float32,
tf.random_normal_initializer(stddev=0.02))
bias = tf.get_variable("bias", [output_size],
initializer=tf.constant_initializer(0.0))
return tf.matmul(input_, matrix) + bias
class data_iter(object):
# filelist: 要生成batch的图像路径和标签的filelist
# batch_size: 每个batch有多少张图片
def __init__(self,filelist,batch_size):
self.batch_size = batch_size
self.filelist = filelist
self.num_batches = 0
self.pointer = 0
self.x_batches = []
self.y_batches = []
self.creat_batches = self.get_csv_batch()
# 生成相同大小的批次 CSV文件
def get_csv_batch(self):
with tf.variable_scope('input'):
data = pd.read_csv(self.filelist, header=0, dtype=np.int)
x, y = np.asarray(data.iloc[:, 1:]), data['label']
self.num_batches = int(len(y) / self.batch_size)
samples = self.num_batches * self.batch_size
# for i in range(self.num_batches):
# self.x_batches.append(x[i])
self.x_batches = np.split(x[:samples].reshape(self.batch_size, -1), self.num_batches, 1)
self.y_batches = np.split(y[:samples].reshape(self.batch_size, -1), self.num_batches, 1)
def next_batch(self):
x, y = self.x_batches[self.pointer], self.y_batches[self.pointer]
self.pointer += 1
return x, np.squeeze(y)
def reset_batch_pointer(self):
self.pointer = 0
def load_model(sess, saver, restore_checkpoint):
print('Reading checkpoints...')
ckpt = tf.train.get_checkpoint_state(restore_checkpoint)
try:
if ckpt and tf.train.checkpoint_exists(ckpt.model_checkpoint_path):
saver.restore(sess, ckpt.model_checkpoint_path)
print('Sucessful loading checkpoing...%s' % ckpt.model_checkpoint_path)
else:
sess.run(tf.global_variables_initializer())
raise TypeError('no checkpoint in %s' % restore_checkpoint)
except Exception as e:
print(e)
def prams_summaries_all():
# Add summaries for variables.
prams0 = tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES, 'gen')
prams1 = tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES, 'dis')
prams = prams0 + prams1
sum_list = []
for p in prams:
name = tf.summary.histogram(p.op.name, p)
sum_list.append(name)
return sum_list
def count_trainable_params():
total_parameters = 0
a = []
for variable in tf.trainable_variables():
a.append(variable)
shape = variable.get_shape()
variable_parametes = 1
for dim in shape:
variable_parametes *= dim.value
total_parameters += variable_parametes
print("Total training params: %.1fM" % (total_parameters / 1e6))
def model_identifier(model_type):
print ("training model is : %s " % model_type)
def save_model(saver, sess, logdir, global_stepe, gl, dl):
save_path = logdir + 'model.ckpt'
saver.save(sess, save_path, global_step=global_stepe)
print('\nGen_loss {:.9f} and Dis_loss {:.9f} in step : {})'
'\ncheckpoint has been saved in : {}'.format(gl, dl, global_stepe, logdir))
def view_samples(epoch, samples,shape, output_dir):
"""
#用于可视化epoch后输出图片
epoch代表第几次迭代的图像
samples为我们的采样结果
"""
import matplotlib
# Force matplotlib to not use any Xwindows backend.
matplotlib.use('Agg')
import matplotlib.pyplot as plt
rows, cols = shape[0],shape[1]
map = []
for i in range(rows):
tmp = []
for j in range(cols):
tmp.append(((samples[i*cols + j])))
# tmp = np.hstack(tmp)
map.append(np.hstack(tmp))
map = np.asarray(np.vstack(map))
plt.imshow(map)
plt.title('Epoch: %s' % epoch)
out_file = output_dir + '%s.png' % epoch
plt.savefig(out_file, dpi=300)
return out_file
def gen_gif(show_images_path, output_dir):
from PIL import Image
im = Image.open(show_images_path[0])
images = []
for i in range(len(show_images_path)):
images.append(Image.open(show_images_path[i]))
im.save(output_dir+'mnist.gif', save_all=True, append_images=images, loop=1, duration=1, comment=b"gen_images")
def gen_pair(file):
import glob
data_path = glob.glob(file+'*.jpg')
label = np.ones(len(data_path),np.int32)
return data_path,label
def get_image_label_pair(filelist_path):
# 解析文本文件
label_image = lambda x: x.strip().split(' ')
with open(filelist_path) as f:
label = [int(label_image(line)[0]) for line in f.readlines()]
with open(filelist_path) as f:
image_path_list = [label_image(line)[1] for line in f.readlines()]
return image_path_list, label
def pre_processing(img, Isize, crop_size, method):
# resize and crop
img = tf.image.convert_image_dtype(img, dtype=tf.float32)
img = tf.image.resize_images(img, [Isize, Isize], method=tf.image.ResizeMethod.NEAREST_NEIGHBOR)
# img = tf.random_crop(img, [crop_size, crop_size, 3])
# preprocess
if method == 'default':
# img = tf.image.random_flip_left_right(img)
# img = tf.image.random_brightness(img, max_delta=63)
# img = tf.image.random_contrast(img, lower=0.2, upper=1.8)
# img = tf.image.per_image_standardization(img)
# img = tf.cast(img, tf.float32) * (1. / 255)
img = tf.reshape(img, [Isize, Isize, 3])
# keras preprocess module
return img
# 生成相同大小的批次
def get_batch(image, label, image_W=256, image_H=256, batch_size=32, capacity=256,min_after_dequeue=None, is_training=True):
# image, label: 要生成batch的图像路径和标签list
# image_W, image_H: 图片的宽高
# batch_size: 每个batch有多少张图片
# capacity: 队列容量
# return: 图像和标签的batch
# 将python.list类型转换成tf能够识别的格式
with tf.variable_scope('input'):
image = tf.cast(image, tf.string)
label = tf.cast(label, tf.int64)
# 生成队列
input_queue = tf.train.slice_input_producer([image, label])
image_contents = tf.read_file(input_queue[0])
label = input_queue[1]
image = tf.image.decode_jpeg(image_contents, channels=3)
image = pre_processing(image, image_H, image_H, 'default')
# 统一图片大小
# image = tf.image.resize_images(image, [image_H, image_W], method=tf.image.ResizeMethod.NEAREST_NEIGHBOR)
# image = tf.cast(image, tf.float32)
# image = tf.image.per_image_standardization(image) # 标准化数据
if is_training:
if not min_after_dequeue:
image_batch, label_batch, filename = tf.train.batch([image, label, input_queue[0]],
batch_size=batch_size,
num_threads=64, # 线程
capacity=capacity)
else:
image_batch, label_batch, filename = tf.train.shuffle_batch([image, label, input_queue[0]],
batch_size=batch_size,
num_threads=64, # 线程
capacity=capacity + min_after_dequeue,
min_after_dequeue=min_after_dequeue)
else:
image_batch, label_batch, filename = tf.train.batch([image, label, input_queue[0]],
batch_size=batch_size,
num_threads=64, # 线程
capacity=capacity)
return image_batch, label_batch, filename