-
Notifications
You must be signed in to change notification settings - Fork 27
/
Copy pathtemporal_transforms.py
269 lines (203 loc) · 8.36 KB
/
temporal_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
import random
import math
import numpy as np
class LoopPadding(object):
def __init__(self, size, downsample):
self.size = size
self.downsample = downsample
def __call__(self, frame_indices):
vid_duration = len(frame_indices)
clip_duration = self.size * self.downsample
out = frame_indices
for index in out:
if len(out) >= clip_duration:
break
out.append(index)
selected_frames = [out[i] for i in range(0, clip_duration, self.downsample)]
return out
class TemporalBeginCrop(object):
"""Temporally crop the given frame indices at a beginning.
If the number of frames is less than the size,
loop the indices as many times as necessary to satisfy the size.
Args:
size (int): Desired output size of the crop.
"""
def __init__(self, size, downsample):
self.size = size
self.downsample = downsample
def __call__(self, frame_indices):
vid_duration = len(frame_indices)
clip_duration = self.size * self.downsample
out = frame_indices[:clip_duration]
for index in out:
if len(out) >= clip_duration:
break
out.append(index)
selected_frames = [out[i] for i in range(0, clip_duration, self.downsample)]
return selected_frames
class TemporalCenterCrop(object):
"""Temporally crop the given frame indices at a center.
If the number of frames is less than the size,
loop the indices as many times as necessary to satisfy the size.
Args:
size (int): Desired output size of the crop.
"""
def __init__(self, size, downsample):
self.size = size
self.downsample = downsample
def __call__(self, frame_indices):
"""
Args:
frame_indices (list): frame indices to be cropped.
Returns:
list: Cropped frame indices.
"""
vid_duration = len(frame_indices)
clip_duration = self.size * self.downsample
center_index = len(frame_indices) // 2
begin_index = max(0, center_index - (clip_duration // 2))
end_index = min(begin_index + clip_duration, vid_duration)
out = frame_indices[begin_index:end_index]
for index in out:
if len(out) >= clip_duration:
break
out.append(index)
selected_frames = [out[i] for i in range(0, clip_duration, self.downsample)]
return selected_frames
class TemporalRandomCrop(object):
"""Temporally crop the given frame indices at a random location.
If the number of frames is less than the size,
loop the indices as many times as necessary to satisfy the size.
Args:
size (int): Desired output size of the crop.
"""
def __init__(self, size, downsample):
self.size = size
self.downsample = downsample
def __call__(self, frame_indices):
"""
Args:
frame_indices (list): frame indices to be cropped.
Returns:
list: Cropped frame indices.
"""
vid_duration = len(frame_indices)
clip_duration = self.size * self.downsample
rand_end = max(0, vid_duration - clip_duration - 1)
begin_index = random.randint(0, rand_end)
end_index = min(begin_index + clip_duration, vid_duration)
out = frame_indices[begin_index:end_index]
for index in out:
if len(out) >= clip_duration:
break
out.append(index)
selected_frames = [out[i] for i in range(0, clip_duration, self.downsample)]
return selected_frames
class TemporalSelectCrop(object):
def __init__(self, size, downsample, number_clips=6, clip_interval=0):
self.size = size
self.downsample = downsample
self.clip_duration = self.size * self.downsample
self.number_clips = number_clips
self.clip_interval = clip_interval
def __call__(self, frame_indices):
vid_duration = len(frame_indices)
outs = []
if (self.clip_duration + self.clip_interval) * self.number_clips - self.clip_interval <= vid_duration:
center_index = len(frame_indices) // 2
begin_index = max(0, center_index -
(self.clip_duration // 2 +
self.number_clips // 2 * (self.clip_duration + self.clip_interval)))
for i in range(self.number_clips):
end_index = min(begin_index + self.clip_duration, vid_duration)
out = frame_indices[begin_index:end_index]
outs.append(out)
begin_index = end_index + self.clip_interval
else:
for i in range(self.number_clips):
rand_end = max(0, vid_duration - self.clip_duration - 1)
begin_index = random.randint(0, rand_end)
end_index = min(begin_index + self.clip_duration, vid_duration)
out = frame_indices[begin_index:end_index]
outs.append(out)
total_frames = []
for out in outs:
for index in out:
if len(out) >= self.clip_duration:
break
out.append(index)
frames = [out[i] for i in range(0, self.clip_duration, self.downsample)]
total_frames.append(frames)
return total_frames
class TemporalBeginEndCrop(object):
def __init__(self, size, downsample):
self.size = size
self.downsample = downsample
self.clip_duration = self.size * self.downsample
def __call__(self, frame_indices):
vid_duration = len(frame_indices)
outs = []
begin = frame_indices[:self.clip_duration]
end = frame_indices[-self.clip_duration:]
outs.append(begin)
outs.append(end)
total_frames = []
for out in outs:
for index in out:
if len(out) >= self.clip_duration:
break
out.append(index)
frames = [out[i] for i in range(0, self.clip_duration, self.downsample)]
total_frames.append(frames)
return total_frames
class TemporalRandomMultipleCrop(object):
def __init__(self, size, downsample, number_clips=4, clip_interval=-1):
self.size = size
self.downsample = downsample
self.clip_duration = self.size * self.downsample
self.number_clips = number_clips
self.clip_interval = clip_interval
def __call__(self, frame_indices):
vid_duration = len(frame_indices)
outs = []
if self.clip_interval < 0:
# randomly choose clips
for i in range(self.number_clips):
rand_end = max(0, vid_duration - self.clip_duration - 1)
begin_index = random.randint(0, rand_end)
end_index = min(begin_index + self.clip_duration, vid_duration)
out = frame_indices[begin_index : end_index]
outs.append(out)
else:
rand_begin = vid_duration - self.clip_duration * (self.number_clips-1) - self.clip_interval * (self.number_clips-1)
if rand_begin < 0:
begin_index = random.randint(0, rand_begin)
else:
begin_index = 0
for i in range(self.number_clips):
end_index = min(begin_index + self.clip_duration, vid_duration)
out = frame_indices[begin_index : end_index]
outs.append(out)
begin_index = min(begin_index)
total_frames = []
for out in outs:
for index in out:
if len(out) >= self.clip_duration:
break
out.append(index)
frames = [out[i] for i in range(0, self.clip_duration, self.downsample)]
total_frames.append(frames)
return total_frames
class TemporalSequentialCrop(object):
def __init__(self, duration=32, downsample=2):
self.duration = duration
self.downsample = downsample
if self.duration % self.downsample != 0:
print('Error! Sample duration should be be an integral multiple of downsample!')
assert 0
def __call__(self, frame_indices):
help = []
step = self.downsample
for i in range(0, self.duration, step):
help.append(frame_indices[i])
return help