/
my_augmentations.py
212 lines (171 loc) · 6.79 KB
/
my_augmentations.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
import argparse
import os
import numpy as np
from PIL import Image
from scipy.misc import imresize
from scipy.ndimage.interpolation import rotate
def random_crop(image, crop_size):
crop_size = check_size(crop_size)
h, w, _ = image.shape
top = np.random.randint(0, h - crop_size[0])
left = np.random.randint(0, w - crop_size[1])
bottom = top + crop_size[0]
right = left + crop_size[1]
image = image[top:bottom, left:right, :]
return image
def horizontal_flip(image, rate=0.5):
if np.random.rand() < rate:
image = image[:, ::-1, :]
return image
def zero_pad(image, final_size, start_at):
result = np.zeros(final_size)
result[start_at:start_at+image.shape[0],
start_at:start_at+image.shape[1],
:] = image
return result
def read_image(imagefile, dtype=np.float32):
image = np.array(Image.open(imagefile), dtype=dtype)
return image
def save_image(image, imagefile, data_format='channel_last'):
image = np.asarray(image, dtype=np.uint8)
image = Image.fromarray(image)
image.save(imagefile)
def concat_images(images, rows, cols):
_, h, w, _ = images.shape
images = images.reshape((rows, cols, h, w, 3))
images = images.transpose(0, 2, 1, 3, 4)
images = images.reshape((rows * h, cols * w, 3))
return images
def check_size(size):
if type(size) == int:
size = (size, size)
if type(size) != tuple:
raise TypeError('size is int or tuple')
return size
def subtract(image):
image = image / 255
return image
def resize(image, size):
size = check_size(size)
image = imresize(image, size)
return image
def center_crop(image, crop_size):
crop_size = check_size(crop_size)
h, w, _ = image.shape
top = (h - crop_size[0]) // 2
left = (w - crop_size[1]) // 2
bottom = top + crop_size[0]
right = left + crop_size[1]
image = image[top:bottom, left:right, :]
return image
def vertical_flip(image, rate=0.5):
if np.random.rand() < rate:
image = image[::-1, :, :]
return image
def scale_augmentation(image, scale_range, crop_size):
scale_size = np.random.randint(*scale_range)
image = imresize(image, (scale_size, scale_size))
image = random_crop(image, crop_size)
return image
def random_rotation(image, angle_range=(0, 180)):
h, w, _ = image.shape
angle = np.random.randint(*angle_range)
image = rotate(image, angle)
image = resize(image, (h, w))
return image
def cutout(image_origin, mask_size, mask_value='mean'):
image = np.copy(image_origin)
if mask_value == 'mean':
mask_value = image.mean()
elif mask_value == 'random':
mask_value = np.random.randint(0, 256)
h, w, _ = image.shape
top = np.random.randint(0 - mask_size // 2, h - mask_size)
left = np.random.randint(0 - mask_size // 2, w - mask_size)
bottom = top + mask_size
right = left + mask_size
if top < 0:
top = 0
if left < 0:
left = 0
image[top:bottom, left:right, :].fill(mask_value)
return image
def random_erasing(image_origin, p=0.5, s=(0.02, 0.4), r=(0.3, 3), mask_value='random'):
image = np.copy(image_origin)
if np.random.rand() > p:
return image
if mask_value == 'mean':
mask_value = image.mean()
elif mask_value == 'random':
mask_value = np.random.randint(0, 256)
h, w, _ = image.shape
mask_area = np.random.randint(h * w * s[0], h * w * s[1])
mask_aspect_ratio = np.random.rand() * r[1] + r[0]
mask_height = int(np.sqrt(mask_area / mask_aspect_ratio))
if mask_height > h - 1:
mask_height = h - 1
mask_width = int(mask_aspect_ratio * mask_height)
if mask_width > w - 1:
mask_width = w - 1
top = np.random.randint(0, h - mask_height)
left = np.random.randint(0, w - mask_width)
bottom = top + mask_height
right = left + mask_width
image[top:bottom, left:right, :].fill(mask_value)
return image
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Image Data Augmentation')
parser.add_argument('infile')
parser.add_argument('--outdir', '-o', default='./')
parser.add_argument('--n_loop', '-n', type=int, default=1)
parser.add_argument('--concat', '-c', action='store_true')
args = parser.parse_args()
processing_list = ['random_crop', 'horizontal_flip', 'vertical_flip',
'scale_augmentation', 'random_rotation', 'cutout',
'random_erasing']
inimg = read_image(args.infile)
inimg224 = resize(inimg, 224)
if args.concat:
if not os.path.exists(args.outdir):
os.makedirs(args.outdir)
def save_concat_image(outimg_name, func, *func_args):
images = []
for i in range(args.n_loop):
images.append(func(*func_args))
x = int(np.sqrt(args.n_loop))
outimg = concat_images(np.array(images), x, x)
save_image(outimg, os.path.join(args.outdir, outimg_name))
save_concat_image('random_crop.jpg', random_crop, resize(inimg, 400), 224)
save_concat_image('horizontal_flip.jpg', horizontal_flip, inimg224)
save_concat_image('vertical_flip.jpg', vertical_flip, inimg224)
save_concat_image('scale_augmentation.jpg', scale_augmentation, inimg, (256, 480), 224)
save_concat_image('random_rotation.jpg', random_rotation, inimg224)
save_concat_image('cutout.jpg', cutout, inimg224, inimg224.shape[0] // 2)
save_concat_image('random_erasing.jpg', random_erasing, inimg224)
else:
for processing_name in processing_list:
outdir = os.path.join(args.outdir, processing_name)
if not os.path.exists(outdir):
os.makedirs(outdir)
for i in range(args.n_loop):
save_image(
random_crop(resize(inimg, 256), 224),
os.path.join(args.outdir, 'random_crop', '{}.jpg'.format(i)))
save_image(
horizontal_flip(inimg224),
os.path.join(args.outdir, 'horizontal_flip', '{}.jpg'.format(i)))
save_image(
vertical_flip(inimg224),
os.path.join(args.outdir, 'vertical_flip', '{}.jpg'.format(i)))
save_image(
scale_augmentation(inimg, (256, 480), 224),
os.path.join(args.outdir, 'scale_augmentation', '{}.jpg'.format(i)))
save_image(
random_rotation(inimg224),
os.path.join(args.outdir, 'random_rotation', '{}.jpg'.format(i)))
save_image(
cutout(inimg224, inimg224.shape[0] // 2),
os.path.join(args.outdir, 'cutout', '{}.jpg'.format(i)))
save_image(
random_erasing(inimg224),
os.path.join(args.outdir, 'random_erasing', '{}.jpg'.format(i)))