-
Notifications
You must be signed in to change notification settings - Fork 14
/
Copy pathN2V_DataGeneratorTR001.py
266 lines (227 loc) · 11.6 KB
/
N2V_DataGeneratorTR001.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
import numpy as np
from os.path import join
from glob import glob
import tifffile
from matplotlib import image
from csbdeep.utils import _raise
class N2V_DataGenerator2():
"""
The 'N2V_DataGenerator' enables training and validation data generation for Noise2Void.
"""
def load_imgs(self, files, to32bit, dims='YX'):
"""
Helper to read a list of files. The images are not required to have same size,
but have to be of same dimensionality.
Parameters
----------
files : list(String)
List of paths to tiff-files.
dims : String, optional(default='YX')
Dimensions of the images to read. Known dimensions are: 'TZYXC'
to32bit: makes conversion to 32 bit optional, if set to False the original datatype is used
Returns
-------
images : list(array(float))
A list of the read tif-files. The images have dimensionality 'SZYXC' or 'SYXC'
"""
assert 'Y' in dims and 'X' in dims, "'dims' has to contain 'X' and 'Y'."
tmp_dims = dims
for b in ['X', 'Y', 'Z', 'T', 'C']:
assert tmp_dims.count(b) <= 1, "'dims' has to contain {} at most once.".format(b)
tmp_dims = tmp_dims.replace(b, '')
assert len(tmp_dims) == 0, "Unknown dimensions in 'dims'."
if 'Z' in dims:
net_axes = 'ZYXC'
else:
net_axes = 'YXC'
move_axis_from = ()
move_axis_to = ()
for d, b in enumerate(dims):
move_axis_from += tuple([d])
if b == 'T':
move_axis_to += tuple([0])
elif b == 'C':
move_axis_to += tuple([-1])
elif b in 'XYZ':
if 'T' in dims:
move_axis_to += tuple([net_axes.index(b)+1])
else:
move_axis_to += tuple([net_axes.index(b)])
imgs = []
for f in files:
if f.endswith('.tif') or f.endswith('.tiff'):
imread = tifffile.imread
elif f.endswith('.png'):
imread = image.imread
elif f.endswith('.jpg') or f.endswith('.jpeg') or f.endswith('.JPEG') or f.endswith('.JPG'):
_raise(Exception("JPEG is not supported, because it is not loss-less and breaks the pixel-wise independence assumption."))
else:
_raise("Filetype '{}' is not supported.".format(f))
if to32bit:
img = imread(f).astype(np.float32)
else:
img = imread(f)
assert len(img.shape) == len(dims), "Number of image dimensions doesn't match 'dims'."
img = np.moveaxis(img, move_axis_from, move_axis_to)
if not ('T' in dims):
img = img[np.newaxis]
if not ('C' in dims):
img = img[..., np.newaxis]
imgs.append(img)
return imgs
def load_imgs_from_directory(self, directory, filter='*.tif', dims='YX', names_back = False, to32bit = True):
"""
Helper to read all files which match 'filter' from a directory. The images are not required to have same size,
but have to be of same dimensionality.
Parameters
----------
directory : String
Directory from which the data is loaded.
filter : String, optional(default='*.tif')
Filter to match the file names.
dims : String, optional(default='YX')
Dimensions of the images to read. Known dimensions are: 'TZYXC'
names_back: if set to True, the function returns the names of the input files as list
to32bit: makes conversion to 32 bit optional, if set to False the original datatype is used
Returns
-------
images : list(array(float))
A list of the read tif-files. The images have dimensionality 'SZYXC' or 'SYXC'
(optional): files, names of the input files as list
"""
files = glob(join(directory, filter))
files.sort()
if names_back:
return files, self.load_imgs(files, to32bit, dims=dims)
else:
return self.load_imgs(files, to32bit, dims=dims)
def generate_patches_from_list(self, data, num_patches_per_img=None, shape=(256, 256), augment=True, shuffle=False):
"""
Extracts patches from 'list_data', which is a list of images, and returns them in a 'numpy-array'. The images
can have different dimensionality.
Parameters
----------
data : list(array(float))
List of images with dimensions 'SZYXC' or 'SYXC'
num_patches_per_img : int, optional(default=None)
Number of patches to extract per image. If 'None', as many patches as fit i nto the
dimensions are extracted.
shape : tuple(int), optional(default=(256, 256))
Shape of the extracted patches.
augment : bool, optional(default=True)
Rotate the patches in XY-Plane and flip them along X-Axis. This only works if the patches are square in XY.
shuffle : bool, optional(default=False)
Shuffles extracted patches across all given images (data).
Returns
-------
patches : array(float)
Numpy-Array with the patches. The dimensions are 'SZYXC' or 'SYXC'
"""
patches = []
for img in data:
for s in range(img.shape[0]):
p = self.generate_patches(img[s][np.newaxis], num_patches=num_patches_per_img, shape=shape, augment=augment)
patches.append(p)
patches = np.concatenate(patches, axis=0)
if shuffle:
np.random.shuffle(patches)
return patches
def generate_patches(self, data, num_patches=None, shape=(256, 256), augment=True, shuffle_patches = True):
"""
Extracts patches from 'data'. The patches can be augmented, which means they get rotated three times
in XY-Plane and flipped along the X-Axis. Augmentation leads to an eight-fold increase in training data.
Parameters
----------
data : list(array(float))
List of images with dimensions 'SZYXC' or 'SYXC'
num_patches : int, optional(default=None)
Number of patches to extract per image. If 'None', as many patches as fit i nto the
dimensions are extracted.
shape : tuple(int), optional(default=(256, 256))
Shape of the extracted patches.
augment : bool, optional(default=True)
Rotate the patches in XY-Plane and flip them along X-Axis. This only works if the patches are square in XY.
Returns
-------
patches : array(float)
Numpy-Array containing all patches (randomly shuffled along S-dimension).
The dimensions are 'SZYXC' or 'SYXC'
"""
patches = self.__extract_patches__(data, num_patches=num_patches, shape=shape, n_dims=len(data.shape)-2)
if shape[-2] == shape[-1]:
if augment:
patches = self.__augment_patches__(patches=patches)
else:
if augment:
print("XY-Plane is not square. Omit augmentation!")
if shuffle_patches:
np.random.shuffle(patches)
#print('Generated patches:', patches.shape)
return patches
def __extract_patches__(self, data, num_patches=None, shape=(256, 256), n_dims=2):
if num_patches == None:
patches = []
if n_dims == 2:
if data.shape[1] >= shape[0] and data.shape[2] >= shape[1]:
for y in range(0, data.shape[1] - shape[0]+1, shape[0]):
for x in range(0, data.shape[2] - shape[1]+1, shape[1]):
patches.append(data[:, y:y + shape[0], x:x + shape[1]])
return np.concatenate(patches)
elif data.shape[1] == shape[0] and data.shape[2] == shape[1]:
return data
else:
print("'shape' is too big.")
elif n_dims == 3:
if data.shape[1] >= shape[0] and data.shape[2] >= shape[1] and data.shape[3] >= shape[2]:
for z in range(0, data.shape[1] - shape[0]+1, shape[0]):
for y in range(0, data.shape[2] - shape[1]+1, shape[1]):
for x in range(0, data.shape[3] - shape[2]+1, shape[2]):
patches.append(data[:, z:z + shape[0], y:y + shape[1], x:x + shape[2]])
return np.concatenate(patches)
elif data.shape[1] == shape[0] and data.shape[2] == shape[1] and data.shape[3] == shape[
2]:
return data
else:
print("'shape' is too big.")
else:
print('Not implemented for more than 4 dimensional (ZYXC) data.')
else:
patches = []
if n_dims == 2:
for i in range(num_patches):
y, x = np.random.randint(0, data.shape[1] - shape[0] + 1), np.random.randint(0,
data.shape[
2] - shape[
1] + 1)
patches.append(data[0, y:y + shape[0], x:x + shape[1]])
if len(patches) > 1:
return np.stack(patches)
else:
return np.array(patches)[np.newaxis]
elif n_dims == 3:
for i in range(num_patches):
z, y, x = np.random.randint(0, data.shape[1] - shape[0] + 1), np.random.randint(0,
data.shape[
2] - shape[
1] + 1), np.random.randint(
0, data.shape[3] - shape[2] + 1)
patches.append(data[0, z:z + shape[0], y:y + shape[1], x:x + shape[2]])
if len(patches) > 1:
return np.stack(patches)
else:
return np.array(patches)[np.newaxis]
else:
print('Not implemented for more than 4 dimensional (ZYXC) data.')
def __augment_patches__(self, patches):
if len(patches.shape[1:-1]) == 2:
augmented = np.concatenate((patches,
np.rot90(patches, k=1, axes=(1, 2)),
np.rot90(patches, k=2, axes=(1, 2)),
np.rot90(patches, k=3, axes=(1, 2))))
elif len(patches.shape[1:-1]) == 3:
augmented = np.concatenate((patches,
np.rot90(patches, k=1, axes=(2, 3)),
np.rot90(patches, k=2, axes=(2, 3)),
np.rot90(patches, k=3, axes=(2, 3))))
augmented = np.concatenate((augmented, np.flip(augmented, axis=-2)))
return augmented