-
Notifications
You must be signed in to change notification settings - Fork 4
/
augment_data.py
208 lines (183 loc) · 7.38 KB
/
augment_data.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
import os
import numpy as np
import glob
import nibabel as nib
from myconfig import *
def rotate_1(img,lab):
# img,lab must be [channels,height,width,depth]
c1,h,w,d = img.shape
c2 = lab.shape[0]
img_rot = np.zeros((c1,h,w,d),dtype=np.float32)
lab_rot = np.zeros((c2,h,w,d),dtype=np.uint8)
for i in range(d):
for j in range(c1):
img_rot[j,:,:,i] = np.rot90(img[j,:,:,i],1)
for j in range(c2):
lab_rot[j,:,:,i] = np.rot90(lab[j,:,:,i],1)
return img_rot, lab_rot
def rotate_2(img,lab):
# img,lab must be [channels,height,width,depth]
c1,h,w,d = img.shape
c2 = lab.shape[0]
img_rot = np.zeros((c1,h,w,d),dtype=np.float32)
lab_rot = np.zeros((c2,h,w,d),dtype=np.uint8)
for i in range(d):
for j in range(c1):
img_rot[j,:,:,i] = np.rot90(img[j,:,:,i],2)
for j in range(c2):
lab_rot[j,:,:,i] = np.rot90(lab[j,:,:,i],2)
return img_rot, lab_rot
def rotate_3(img,lab):
# img,lab must be [channels,height,width,depth]
c1,h,w,d = img.shape
c2 = lab.shape[0]
img_rot = np.zeros((c1,h,w,d),dtype=np.float32)
lab_rot = np.zeros((c2,h,w,d),dtype=np.uint8)
for i in range(d):
for j in range(c1):
img_rot[j,:,:,i] = np.rot90(img[j,:,:,i],3)
for j in range(c2):
lab_rot[j,:,:,i] = np.rot90(lab[j,:,:,i],3)
return img_rot, lab_rot
def rotate_4(img,lab):
# img,lab must be [channels,height,width,depth]
c1,h,w,d = img.shape
c2 = lab.shape[0]
img_rot = np.zeros((c1,h,w,d),dtype=np.float32)
lab_rot = np.zeros((c2,h,w,d),dtype=np.uint8)
for i in range(d):
for j in range(c1):
img_rot[j,:,:,i] = np.fliplr(img[j,:,:,i])
for j in range(c2):
lab_rot[j,:,:,i] = np.fliplr(lab[j,:,:,i])
return img_rot, lab_rot
def rotate_5(img,lab):
# img,lab must be [channels,height,width,depth]
c1,h,w,d = img.shape
c2 = lab.shape[0]
img_rot = np.zeros((c1,h,w,d),dtype=np.float32)
lab_rot = np.zeros((c2,h,w,d),dtype=np.uint8)
for i in range(d):
for j in range(c1):
img_rot[j,:,:,i] = np.flipud(img[j,:,:,i])
for j in range(c2):
lab_rot[j,:,:,i] = np.flipud(lab[j,:,:,i])
return img_rot, lab_rot
def rotate_6(img,lab):
# img,lab must be [channels,height,width,depth]
c1,h,w,d = img.shape
c2 = lab.shape[0]
img_rot = np.zeros((c1,h,w,d),dtype=np.float32)
lab_rot = np.zeros((c2,h,w,d),dtype=np.uint8)
for i in range(d):
for j in range(c1):
img_rot[j,:,:,i] = np.flipud(np.rot90(img[j,:,:,i],1))
for j in range(c2):
lab_rot[j,:,:,i] = np.flipud(np.rot90(lab[j,:,:,i],1))
return img_rot, lab_rot
def rotate_7(img,lab):
# img,lab must be [channels,height,width,depth]
c1,h,w,d = img.shape
c2 = lab.shape[0]
img_rot = np.zeros((c1,h,w,d),dtype=np.float32)
lab_rot = np.zeros((c2,h,w,d),dtype=np.uint8)
for i in range(d):
for j in range(c1):
img_rot[j,:,:,i] = np.flipud(np.rot90(img[j,:,:,i],3))
for j in range(c2):
lab_rot[j,:,:,i] = np.flipud(np.rot90(lab[j,:,:,i],3))
return img_rot, lab_rot
def rotate_8(img,lab):
# img,lab must be [channels,height,width,depth]
c1,h,w,d = img.shape
c2 = lab.shape[0]
img_rot = np.zeros((c1,h,w,d),dtype=np.float32)
lab_rot = np.zeros((c2,h,w,d),dtype=np.uint8)
for i in range(d):
for j in range(c1):
img_rot[j,:,:,i] = np.flipud(np.rot90(img[j,:,:,i],2))
for j in range(c2):
lab_rot[j,:,:,i] = np.flipud(np.rot90(lab[j,:,:,i],2))
return img_rot, lab_rot
def rotate_9(img,lab):
# img,lab must be [channels,height,width,depth]
c1,h,w,d = img.shape
c2 = lab.shape[0]
img_rot = np.zeros((c1,h,w,d),dtype=np.float32)
lab_rot = np.zeros((c2,h,w,d),dtype=np.uint8)
for i in range(d):
for j in range(c1):
img_rot[j,:,:,i] = np.fliplr(np.rot90(img[j,:,:,i],1))
for j in range(c2):
lab_rot[j,:,:,i] = np.fliplr(np.rot90(lab[j,:,:,i],1))
return img_rot, lab_rot
def rotate_10(img,lab):
# img,lab must be [channels,height,width,depth]
c1,h,w,d = img.shape
c2 = lab.shape[0]
img_rot = np.zeros((c1,h,w,d),dtype=np.float32)
lab_rot = np.zeros((c2,h,w,d),dtype=np.uint8)
for i in range(d):
for j in range(c1):
img_rot[j,:,:,i] = np.fliplr(np.rot90(img[j,:,:,i],3))
for j in range(c2):
lab_rot[j,:,:,i] = np.fliplr(np.rot90(lab[j,:,:,i],3))
return img_rot, lab_rot
def rotate_11(img,lab):
# img,lab must be [channels,height,width,depth]
c1,h,w,d = img.shape
c2 = lab.shape[0]
img_rot = np.zeros((c1,h,w,d),dtype=np.float32)
lab_rot = np.zeros((c2,h,w,d),dtype=np.uint8)
for i in range(d):
for j in range(c1):
img_rot[j,:,:,i] = np.fliplr(np.rot90(img[j,:,:,i],2))
for j in range(c2):
lab_rot[j,:,:,i] = np.fliplr(np.rot90(lab[j,:,:,i],2))
return img_rot, lab_rot
HEIGHT = 96
WIDTH = 96
DEPTH = 48
data_root = DATA_ROOT
sub = 'train'
names = glob.glob(data_root + '/' + sub + '/A-*')
for name in names:
# [channels, depth, height, width]
image = np.zeros((3, HEIGHT, WIDTH, DEPTH),dtype=np.float32)
label = np.zeros((2, HEIGHT, WIDTH, DEPTH),dtype=np.uint8)
image_prefixes = ['InputCT_ROI', 'InputPET_ROI', 'InputPET_SUV_ROI']
label_prefixes = ['GTV_Primary_ROI_CT{}'.format(GT_POSTFIX), 'GTV_Primary_ROI_PET{}'.format(GT_POSTFIX)]
for i, prefix in enumerate(image_prefixes):
f = nib.load('{}/{}.nii.gz'.format(name, prefix))
image[i, ...] = f.get_data().astype(np.float32)
for i, prefix in enumerate(label_prefixes):
f = nib.load('{}/{}.nii.gz'.format(name, prefix))
label[i, ...] = f.get_data().astype(np.uint8)
c1 = 3
c2 = 2
num = 7
image_raw_rot = np.zeros((num,c1,HEIGHT, WIDTH, DEPTH),dtype=np.float32)
label_raw_rot = np.zeros((num,c2,HEIGHT, WIDTH, DEPTH),dtype=np.uint8)
image_raw_rot[0,...], label_raw_rot[0,...] = rotate_1(image, label)
image_raw_rot[1,...], label_raw_rot[1,...] = rotate_2(image, label)
image_raw_rot[2,...], label_raw_rot[2,...] = rotate_3(image, label)
image_raw_rot[3,...], label_raw_rot[3,...] = rotate_4(image, label)
image_raw_rot[4,...], label_raw_rot[4,...] = rotate_5(image, label)
image_raw_rot[5,...], label_raw_rot[5,...] = rotate_6(image, label)
image_raw_rot[6,...], label_raw_rot[6,...] = rotate_7(image, label)
# image_raw_rot[7,...], label_raw_rot[7,...] = rotate_8(image, label) # equal to rotate_4
# image_raw_rot[8,...], label_raw_rot[8,...] = rotate_9(image, label) # equal to rotate_7
# image_raw_rot[9,...], label_raw_rot[9,...] = rotate_10(image, label) # equal to rotate_6
# image_raw_rot[10,...], label_raw_rot[10,...] = rotate_11(image, label) # equal to rotate_5
affine = np.eye(4)
for i in range(num):
save_dir = name+'_rot'+str(i)
os.makedirs(save_dir)
for j, prefix in enumerate(image_prefixes):
img = nib.Nifti1Image(image_raw_rot[i,j,:,:,:], affine)
nib.save(img, '{}/{}.nii.gz'.format(save_dir, prefix))
for j, prefix in enumerate(label_prefixes):
img = nib.Nifti1Image(label_raw_rot[i,j,:,:,:], affine)
nib.save(img, '{}/{}.nii.gz'.format(save_dir, prefix))
print(name)
# break