-
Notifications
You must be signed in to change notification settings - Fork 26
/
flow_transforms.py
324 lines (251 loc) · 11.5 KB
/
flow_transforms.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
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
from __future__ import division
import torch
import math
import random
from PIL import Image, ImageOps
import numpy as np
import numbers
import types
import scipy.ndimage as ndimage
'''Set of tranform random routines that takes both input and target as arguments,
in order to have random but coherent transformations.
inputs are PIL Image pairs and targets are ndarrays'''
class Compose(object):
""" Composes several co_transforms together.
For example:
>>> co_transforms.Compose([
>>> co_transforms.CenterCrop(10),
>>> co_transforms.ToTensor(),
>>> ])
"""
def __init__(self, co_transforms):
self.co_transforms = co_transforms
def __call__(self, input, target_depth, target_label=None):
for i, t in enumerate(self.co_transforms):
# print('After transform {}'.format(i))
# print(np.max(target_label))
if target_label == None:
input, target_depth, _ = t(input, target_depth, target_depth)
return input, target_depth
else:
input, target_depth, target_label = t(input, target_depth, target_label)
return input, target_depth, target_label
class ArrayToTensor(object):
"""Converts a numpy.ndarray (H x W x C) to a torch.FloatTensor of shape (C x H x W)."""
def __call__(self, array):
assert (isinstance(array, np.ndarray))
# handle numpy array
try:
tensor = torch.from_numpy(array).permute(2, 0, 1)
except:
tensor = torch.from_numpy(np.expand_dims(array, axis=2)).permute(2, 0, 1)
# put it from HWC to CHW format
return tensor.float()
class Lambda(object):
"""Applies a lambda as a transform"""
def __init__(self, lambd):
assert type(lambd) is types.LambdaType
self.lambd = lambd
def __call__(self, input, target):
return self.lambd(input, target)
class CenterCrop(object):
"""Crops the given inputs and target arrays at the center to have a region of
the given size. size can be a tuple (target_height, target_width)
or an integer, in which case the target will be of a square shape (size, size)
Careful, img1 and img2 may not be the same size
"""
def __init__(self, size):
if isinstance(size, numbers.Number):
self.size = (int(size), int(size))
else:
self.size = size
def __call__(self, inputs, target):
h1, w1, _ = inputs[0].shape
h2, w2, _ = inputs[1].shape
th, tw = self.size
x1 = int(round((w1 - tw) / 2.))
y1 = int(round((h1 - th) / 2.))
x2 = int(round((w2 - tw) / 2.))
y2 = int(round((h2 - th) / 2.))
inputs[0] = inputs[0][y1: y1 + th, x1: x1 + tw]
inputs[1] = inputs[1][y2: y2 + th, x2: x2 + tw]
target = target[y1: y1 + th, x1: x1 + tw]
return inputs, target
class Scale_Single(object):
""" Rescales a single object, for example only the ground truth dpeth map """
def __init__(self, size, order=2):
self.size = size
self.order = order
def __call__(self, inputs):
h, w = inputs.shape
if (w <= h and w == self.size) or (h <= w and h == self.size):
return inputs
if w < h:
ratio = self.size / w
else:
ratio = self.size / h
inputs = ndimage.interpolation.zoom(inputs, ratio, order=self.order)
return inputs
class Scale(object):
""" Rescales the inputs and target arrays to the given 'size'.
'size' will be the size of the smaller edge.
For example, if height > width, then image will be
rescaled to (size * height / width, size)
size: size of the smaller edge
interpolation order: Default: 2 (bilinear)
"""
def __init__(self, size, order=2):
self.size = size
self.order = order
def __call__(self, inputs, target_depth=None, target_label=None):
h, w, _ = inputs.shape
if (w <= h and w == self.size) or (h <= w and h == self.size):
if target_depth is not None and target_labels is not None:
return inputs, target_depth, target_labels
elif target_depth is not None:
return inputs, target_depth
elif target_labels is not None:
return inputs, target_labels
if w < h:
ratio = self.size / w
else:
ratio = self.size / h
inputs = np.stack((ndimage.interpolation.zoom(inputs[:, :, 0], ratio, order=self.order),
ndimage.interpolation.zoom(inputs[:, :, 1], ratio, order=self.order), \
ndimage.interpolation.zoom(inputs[:, :, 2], ratio, order=self.order)), axis=2)
if target_label is not None and target_depth is not None:
target_label = ndimage.interpolation.zoom(target_label, ratio, order=self.order)
target_depth = ndimage.interpolation.zoom(target_depth, ratio, order=self.order)
return inputs, target_depth, target_label
elif target_depth is not None:
target_depth = ndimage.interpolation.zoom(target_depth, ratio, order=self.order)
return inputs, target_depth
elif target_label is not None:
target_label = ndimage.interpolation.zoom(target_label, ratio, order=self.order)
return inputs, target_label
else:
return inputs
class RandomCrop(object):
"""Crops the given PIL.Image at a random location to have a region of
the given size. size can be a tuple (target_height, target_width)
or an integer, in which case the target will be of a square shape (size, size)
"""
def __init__(self, size):
if isinstance(size, numbers.Number):
self.size = (int(size), int(size))
else:
self.size = size
def __call__(self, inputs, target_depth, target_label):
h, w, _ = inputs.shape
th, tw = self.size
if w == tw and h == th:
return inputs, target_depth, target_label
x1 = random.randint(0, w - tw)
y1 = random.randint(0, h - th)
inputs = inputs[y1: y1 + th, x1: x1 + tw]
return inputs, target_depth[y1: y1 + th, x1: x1 + tw], target_label[y1: y1 + th, x1: x1 + tw]
class RandomHorizontalFlip(object):
"""Randomly horizontally flips the given PIL.Image with a probability of 0.5
"""
def __call__(self, inputs, target_depth, target_label):
if random.random() < 0.5:
inputs = np.flip(inputs, axis=0).copy()
target_depth = np.flip(target_depth, axis=0).copy()
target_label = np.flip(target_label, axis=0).copy()
return inputs, target_depth, target_label
class RandomVerticalFlip(object):
"""Randomly horizontally flips the given PIL.Image with a probability of 0.5
"""
def __call__(self, inputs, target):
if random.random() < 0.5:
inputs[0] = np.flipud(inputs[0])
inputs[1] = np.flipud(inputs[1])
target = np.flipud(target)
target[:, :, 1] *= -1
return inputs, target
class RandomRotate(object):
"""Random rotation of the image from -angle to angle (in degrees)
This is useful for dataAugmentation, especially for geometric problems such as FlowEstimation
angle: max angle of the rotation
interpolation order: Default: 2 (bilinear)
reshape: Default: false. If set to true, image size will be set to keep every pixel in the image.
diff_angle: Default: 0. Must stay less than 10 degrees, or linear approximation of flowmap will be off.
"""
def __init__(self, angle, diff_angle=0, order=2, reshape=False):
self.angle = angle
self.reshape = reshape
self.order = order
def __call__(self, inputs, target_depth, target_label):
applied_angle = random.uniform(-self.angle, self.angle)
angle1 = applied_angle
angle1_rad = angle1 * np.pi / 180
inputs = ndimage.interpolation.rotate(inputs, angle1, reshape=self.reshape, order=self.order)
target_depth = ndimage.interpolation.rotate(target_depth, angle1, reshape=self.reshape, order=self.order)
target_label = ndimage.interpolation.rotate(target_label, angle1, reshape=self.reshape, order=self.order)
return inputs, target_depth, target_label
class RandomCropRotate(object):
"""Random rotation of the image from -angle to angle (in degrees)
A crop is done to keep same image ratio, and no black pixels
angle: max angle of the rotation, cannot be more than 180 degrees
interpolation order: Default: 2 (bilinear)
"""
def __init__(self, angle, size, diff_angle=0, order=2):
self.angle = angle
self.order = order
self.diff_angle = diff_angle
self.size = size
def __call__(self, inputs, target):
applied_angle = random.uniform(-self.angle, self.angle)
diff = random.uniform(-self.diff_angle, self.diff_angle)
angle1 = applied_angle - diff / 2
angle2 = applied_angle + diff / 2
angle1_rad = angle1 * np.pi / 180
angle2_rad = angle2 * np.pi / 180
h, w, _ = inputs[0].shape
def rotate_flow(i, j, k):
return -k * (j - w / 2) * (diff * np.pi / 180) + (1 - k) * (i - h / 2) * (diff * np.pi / 180)
rotate_flow_map = np.fromfunction(rotate_flow, target.shape)
target += rotate_flow_map
inputs[0] = ndimage.interpolation.rotate(inputs[0], angle1, reshape=True, order=self.order)
inputs[1] = ndimage.interpolation.rotate(inputs[1], angle2, reshape=True, order=self.order)
target = ndimage.interpolation.rotate(target, angle1, reshape=True, order=self.order)
# flow vectors must be rotated too!
target_ = np.array(target, copy=True)
target[:, :, 0] = np.cos(angle1_rad) * target_[:, :, 0] - np.sin(angle1_rad) * target_[:, :, 1]
target[:, :, 1] = np.sin(angle1_rad) * target_[:, :, 0] + np.cos(angle1_rad) * target_[:, :, 1]
# keep angle1 and angle2 within [0,pi/2] with a reflection at pi/2: -1rad is 1rad, 2rad is pi - 2 rad
angle1_rad = np.pi / 2 - np.abs(angle1_rad % np.pi - np.pi / 2)
angle2_rad = np.pi / 2 - np.abs(angle2_rad % np.pi - np.pi / 2)
c1 = np.cos(angle1_rad)
s1 = np.sin(angle1_rad)
c2 = np.cos(angle2_rad)
s2 = np.sin(angle2_rad)
c_diag = h / np.sqrt(h * h + w * w)
s_diag = w / np.sqrt(h * h + w * w)
ratio = c_diag / max(c1 * c_diag + s1 * s_diag, c2 * c_diag + s2 * s_diag)
crop = CenterCrop((int(h * ratio), int(w * ratio)))
scale = Scale(self.size)
inputs, target = crop(inputs, target)
return scale(inputs, target)
class RandomTranslate(object):
def __init__(self, translation):
if isinstance(translation, numbers.Number):
self.translation = (int(translation), int(translation))
else:
self.translation = translation
def __call__(self, inputs, target):
h, w, _ = inputs[0].shape
th, tw = self.translation
tw = random.randint(-tw, tw)
th = random.randint(-th, th)
if tw == 0 and th == 0:
return inputs, target
# compute x1,x2,y1,y2 for img1 and target, and x3,x4,y3,y4 for img2
x1, x2, x3, x4 = max(0, tw), min(w + tw, w), max(0, -tw), min(w - tw, w)
y1, y2, y3, y4 = max(0, th), min(h + th, h), max(0, -th), min(h - th, h)
inputs[0] = inputs[0][y1:y2, x1:x2]
inputs[1] = inputs[1][y3:y4, x3:x4]
target = target[y1:y2, x1:x2]
target[:, :, 0] += tw
target[:, :, 1] += th
return inputs, target