-
Notifications
You must be signed in to change notification settings - Fork 2
/
utils.py
executable file
·217 lines (178 loc) · 6.96 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
"""
Scipy version > 0.18 is needed, due to 'mode' option from scipy.misc.imread function
"""
import os
import glob
import h5py
import random
import matplotlib.pyplot as plt
from PIL import Image # for loading images as YCbCr format
import scipy.misc
import scipy.ndimage
import numpy as np
import tensorflow as tf
FLAGS = tf.app.flags.FLAGS
def read_data(path):
"""
Read h5 format data file
Args:
path: file path of desired file
data: '.h5' file format that contains train data values
label: '.h5' file format that contains train label values
"""
with h5py.File(path, 'r') as hf:
data = np.array(hf.get('data'))
label = np.array(hf.get('label'))
return data, label
def preprocess(path, scale=3):
"""
Preprocess single image file
(1) Read original image as YCbCr format (and grayscale as default)
(2) Normalize
(3) Apply image file with bicubic interpolation
Args:
path: file path of desired file
input_: image applied bicubic interpolation (low-resolution)
label_: image with original resolution (high-resolution)
"""
image = imread(path, is_grayscale=True)
label_ = modcrop(image, scale)
# Must be normalized
input_ = label_ / 255.
label_ = label_ / 255.
#input_ = scipy.ndimage.interpolation.zoom(label_, (1./scale), prefilter=False)
#input_ = scipy.ndimage.interpolation.zoom(input_, (scale/1.), prefilter=False)
return input_, label_
def prepare_data(sess, dataset):
"""
Args:
dataset: choose train dataset or test dataset
For train dataset, output data would be ['.../t1.bmp', '.../t2.bmp', ..., '.../t99.bmp']
"""
if FLAGS.is_train:
filenames = os.listdir(dataset)
data_dir = os.path.join(os.getcwd(), dataset)
data = glob.glob(os.path.join(data_dir, "*.bmp"))
else:
# modify the paras if you want to change test pics
data_dir = os.path.join(os.sep, (os.path.join(os.getcwd(), dataset)), "Set5")
data = glob.glob(os.path.join(data_dir, "*.bmp"))
return data
def make_data(sess, data, label):
"""
Make input data as h5 file format
Depending on 'is_train' (flag value), savepath would be changed.
"""
if FLAGS.is_train:
savepath = os.path.join(os.getcwd(), 'checkpoint/train.h5')
else:
savepath = os.path.join(os.getcwd(), 'checkpoint/test.h5')
with h5py.File(savepath, 'w') as hf:
hf.create_dataset('data', data=data)
hf.create_dataset('label', data=label)
def imread(path, is_grayscale=True):
"""
Read image using its path.
Default value is gray-scale, and image is read by YCbCr format as the paper said.
"""
if is_grayscale:
return scipy.misc.imread(path, flatten=True, mode='YCbCr').astype(np.float)
else:
return scipy.misc.imread(path, mode='YCbCr').astype(np.float)
def modcrop(image, scale=3):
"""
To scale down and up the original image, first thing to do is to have no remainder while scaling operation.
We need to find modulo of height (and width) and scale factor.
Then, subtract the modulo from height (and width) of original image size.
There would be no remainder even after scaling operation.
"""
if len(image.shape) == 3:
h, w, _ = image.shape
h = h - np.mod(h, scale)
w = w - np.mod(w, scale)
image = image[0:h, 0:w, :]
else:
h, w = image.shape
h = h - np.mod(h, scale)
w = w - np.mod(w, scale)
image = image[0:h, 0:w]
return image
def input_setup(sess, config):
"""
Read image files and make their sub-images and saved them as a h5 file format.
"""
# Load data path
if config.is_train:
data = prepare_data(sess, dataset="Train")
else:
data = prepare_data(sess, dataset="Test")
sub_input_sequence = []
sub_label_sequence = []
padding = abs(config.image_size - config.label_size) / 2 # 0
if config.is_train:
for i in xrange(len(data)):
input_, label_ = preprocess(data[i], config.scale)
if len(input_.shape) == 3:
h, w, _ = input_.shape
else:
h, w = input_.shape
for x in range(0, h-config.image_size+1, config.stride):
for y in range(0, w-config.image_size+1, config.stride):
sub_input = input_[x:x+config.image_size, y:y+config.image_size] # [33 x 33]
sub_label = label_[x+padding:x+padding+config.label_size, y+padding:y+padding+config.label_size] # [33 x 33]
# Make channel value
sub_input = sub_input.reshape([config.image_size ,config.image_size, 1])
sub_label = sub_label.reshape([config.label_size, config.label_size, 1])
sub_input_sequence.append(sub_input)
sub_label_sequence.append(sub_label)
# shuffle
order = np.arange(len(sub_input_sequence))
random.shuffle(order)
sub_input_sequence = np.array(sub_input_sequence)[order]
sub_label_sequence = np.array(sub_label_sequence)[order]
else:
input_init, label_init = preprocess(data[2], config.scale) # decide which pic to restore
if len(input_init.shape) == 3:
h, w, _ = input_init.shape
else:
h, w = input_init.shape
# Do padding for the grayscale pictures in rows and cols
pad_h = config.image_size - divmod(h, config.image_size)[1]
pad_w = config.image_size - divmod(w, config.image_size)[1]
input_ = np.pad(input_init,((0,pad_h),(0,pad_w)),'symmetric')
label_ = input_
h = h + pad_h
w = w + pad_w
# Numbers of sub-images in height and width of image are needed to compute merge operation.
# pad_h and pad_w are needed to crop the pic after being processed
nx = ny = 0
for x in range(0, h-config.image_size+1, config.stride):
nx += 1; ny = 0
for y in range(0, w-config.image_size+1, config.stride):
ny += 1
sub_input = input_[x:x+config.image_size, y:y+config.image_size] # [33 x 33]
sub_label = label_[x+padding:x+padding+config.label_size, y+padding:y+padding+config.label_size] # [33 x 33]
sub_input = sub_input.reshape([config.image_size,config.image_size,1])
sub_label = sub_label.reshape([config.label_size, config.label_size, 1])
sub_input_sequence.append(sub_input)
sub_label_sequence.append(sub_label)
"""
len(sub_input_sequence) : the number of sub_input (33 x 33 x ch) in one image
(sub_input_sequence[0]).shape : (33, 33, 1)
"""
# Make list to numpy array. With this transform
arrdata = np.asarray(sub_input_sequence) # [?, 33, 33, 1]
arrlabel = np.asarray(sub_label_sequence) # [?, 33, 33, 1]
make_data(sess, arrdata, arrlabel)
if not config.is_train:
return nx, ny, pad_h, pad_w
def imsave(image, path):
return scipy.misc.imsave(path, image)
def merge(images, size):
h, w = images.shape[1], images.shape[2]
img = np.zeros((h*size[0], w*size[1], 1))
for idx, image in enumerate(images):
i = idx % size[1]
j = idx // size[1]
img[j*h:j*h+h, i*w:i*w+w, :] = image
return img