-
Notifications
You must be signed in to change notification settings - Fork 0
/
loss.py
603 lines (329 loc) · 12.4 KB
/
loss.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
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
from __future__ import print_function
import tensorflow as tf
'''
def normalize(x, axis=-1):
"""Normalizing to unit length along the specified dimension.
Args:
x: pytorch Variable
Returns:
x: pytorch Variable, same shape as input
"""
#torch.norm,返回输入张量input 的p 范数。 x = 1. * x / (torch.norm(x, 2, axis, keepdim=True).expand_as(x) + 1e-12)
return x'''
def normalize(x, axis=-1):
x=1.*x/(tf.sqrt(tf.reduce_sum(tf.pow(x,2),keep_dims=True))+1e-12)
return x
'''def euclidean_dist(x, y):
"""
Args:
x: pytorch Variable, with shape [m, d]
y: pytorch Variable, with shape [n, d]
Returns:
dist: pytorch Variable, with shape [m, n]
"""
m, n = x.size(0), y.size(0) #tf.shape()
#x.expand(m,n)将矩阵扩充为m,n;x.t()矩阵的转置
xx = torch.pow(x, 2).sum(1, keepdim=True).expand(m, n) #xx=tf.pow(x,2)
#tf.reduce_sum()
yy = torch.pow(y, 2).sum(1, keepdim=True).expand(n, m).t()#.t()矩阵的转置
dist = xx + yy
# \(out = (beta * M) + (alpha * mat1 @ mat2)\)
dist.addmm_(1, -2, x, y.t()) #执行矩阵乘法
#clamp:小于min的数置为min,大于max的置为max
dist = dist.clamp(min=1e-12).sqrt() # for numerical stability
return dist'''
def euclidean_dist(x,y):
m=tf.shape(x)[0]
n=tf.shape(y)[0]
xx=tf.pow(x,2)
xx=tf.reduce_sum(xx,1,keep_dims=True)
xx=tf.tile(xx,(1,n))
yy=tf.pow(y,2)
yy=tf.reduce_sum(yy,1,keep_dims=True)
yy=tf.tile(yy,(1,n))
yy=tf.transpose(yy)
dist=xx+yy
dist=tf.add(dist,tf.matmul(-2*x,tf.transpose(y)))
dist=tf.clip_by_value(dist,le-12,le+12)
return dist
'''def batch_euclidean_dist(x, y):
"""
Args:
x: pytorch Variable, with shape [N, m, d]
y: pytorch Variable, with shape [N, n, d]
Returns:
dist: pytorch Variable, with shape [N, m, n]
"""
assert len(x.size()) == 3
assert len(y.size()) == 3
assert x.size(0) == y.size(0)
assert x.size(-1) == y.size(-1)
N, m, d = x.size()
N, n, d = y.size()
# shape [N, m, n]
xx = torch.pow(x, 2).sum(-1, keepdim=True).expand(N, m, n)
yy = torch.pow(y, 2).sum(-1, keepdim=True).expand(N, n, m).permute(0, 2, 1)
dist = xx + yy
#baddbmm_()批矩阵的乘法
dist.baddbmm_(1, -2, x, y.permute(0, 2, 1))
dist = dist.clamp(min=1e-12).sqrt() # for numerical stability
return dist
'''
def batch_euclidean_dist(x, y):
assert tf.size(tf.shape(x)) == tf.constant(3)
assert tf.size(tf.shape(y)) == 3
assert tf.shape(x)[0] == tf.shape(y)[0]
assert tf.shape(x)[2] == tf.shape(y)[2]
N, m, d = tf.shape(x)[0],tf.shape(x)[1],tf.shape(x)[2]
xx=tf.reduce_sum(tf.pow(x,2),-1,keep_dims=True)
xx=tf.tile(xx,[1,1,n])
yy=tf.reduce_sum(tf.pow(y,2),-1,keep_dims=True).transpose(0,2,1)
yy=tf.tile(yy,[1,1,m])
dist=xx+yy
dist=tf.add(dist,tf.matmul(-2*x,tf.transpose(0,2,1)))
dist=tf.clip_by_value(dist,le-12,le+12)
return dist
'''def shortest_dist(dist_mat):
"""Parallel version.
Args:
dist_mat: pytorch Variable, available shape:
1) [m, n]
2) [m, n, N], N is batch size
3) [m, n, *], * can be arbitrary additional dimensions
Returns:
dist: three cases corresponding to `dist_mat`:
1) scalar
2) pytorch Variable, with shape [N]
3) pytorch Variable, with shape [*]
"""
m, n = dist_mat.size()[:2]
# Just offering some reference for accessing intermediate distance.
dist = [[0 for _ in range(n)] for _ in range(m)]
for i in range(m):
for j in range(n):
if (i == 0) and (j == 0):
dist[i][j] = dist_mat[i, j]
elif (i == 0) and (j > 0):
dist[i][j] = dist[i][j - 1] + dist_mat[i, j]
elif (i > 0) and (j == 0):
dist[i][j] = dist[i - 1][j] + dist_mat[i, j]
else:
dist[i][j] = torch.min(dist[i - 1][j], dist[i][j - 1]) + dist_mat[i, j]
dist = dist[-1][-1]
return dist'''
def shortest_dist(dist_mat):
m,n=tf.shape(dist_mat)[0],tf.shape(dist_mat)[1]
sess=tf.Session()
init_op=tf.initialize_all_variables()
m,n=sess.run(m),sess.run(n)
dist_mat=sess.run(dist_mat)
dist = [[0 for _ in range(n)] for _ in range(m)]
for i in range(m):
for j in range(n):
if (i == 0) and (j == 0):
dist[i][j] = dist_mat[i, j]
elif (i == 0) and (j > 0):
dist[i][j] = dist[i][j - 1] + dist_mat[i, j]
elif (i > 0) and (j == 0):
dist[i][j] = dist[i - 1][j] + dist_mat[i, j]
else:
dist[i][j] = tf.minimum(dist[i - 1][j], dist[i][j - 1]) + dist_mat[i, j]
dist = dist[-1][-1]
return dist
'''def local_dist(x, y):
"""
Args:
x: pytorch Variable, with shape [M, m, d]
y: pytorch Variable, with shape [N, n, d]
Returns:
dist: pytorch Variable, with shape [M, N]
"""
M, m, d = x.size()
N, n, d = y.size()
#返回一个内存连续的有相同数据的 tensor, 如果原 tensor 内存连续则返回原 tensor.
#返回一个有相同数据但大小不同的新的 tensor.
#返回的 tensor 与原 tensor 共享相同的数据,一定有相同数目的元素,但大小不同. 一个 tensor 必须是连续的 ( contiguous() ) 才能被查看.
x = x.contiguous().view(M * m, d)
y = y.contiguous().view(N * n, d)
# shape [M * m, N * n]
dist_mat = euclidean_dist(x, y)
dist_mat = (torch.exp(dist_mat) - 1.) / (torch.exp(dist_mat) + 1.)
# shape [M * m, N * n] -> [M, m, N, n] -> [m, n, M, N]
dist_mat = dist_mat.contiguous().view(M, m, N, n).permute(1, 3, 0, 2) #将tensor的维度换位
# shape [M, N]
dist_mat = shortest_dist(dist_mat)
return dist_mat'''
def local_dist(x, y):
M,m,d= tf.shape(x)[0],tf.shape(x)[1],tf.shape(x)[2]
N,n,d= tf.shape(x)[0],tf.shape(x)[1],tf.shape(x)[2]
x=tf.reshape(x,[M*m,d])
y=tf.reshape(y,[N*n,d])
dist_mat=euclidean_dist(x,y)
dist_mat=(tf.exp(dist_mat)-1.)/(tf.exp(dist_mat)+1)
dist_mat=tf.transpose(dist_mat,perm=[1,3,0,2])
dist_mat=shortest_dist(dist_mat)
return dist_mat
'''def batch_local_dist(x, y):
"""
Args:
x: pytorch Variable, with shape [N, m, d]
y: pytorch Variable, with shape [N, n, d]
Returns:
dist: pytorch Variable, with shape [N]
"""
assert len(x.size()) == 3
assert len(y.size()) == 3
assert x.size(0) == y.size(0)
assert x.size(-1) == y.size(-1)
# shape [N, m, n]
dist_mat = batch_euclidean_dist(x, y)
dist_mat = (torch.exp(dist_mat) - 1.) / (torch.exp(dist_mat) + 1.)
# shape [N]
dist = shortest_dist(dist_mat.permute(1, 2, 0))
return dist'''
def batch_local_list(x,y):
dist_mat=batch_euclidean_dist(x,y)
dist_mat=(tf.exp(dist_mat)-1.)/(tf.exp(dist_mat)+1.)
dist=shortest_dist(tf.transpose(dist_mat,perm=[1,2,0]))
return dist
'''def hard_example_mining(dist_mat, labels, return_inds=False):
"""For each anchor, find the hardest positive and negative sample.
Args:
dist_mat: pytorch Variable, pair wise distance between samples, shape [N, N]
labels: pytorch LongTensor, with shape [N]
return_inds: whether to return the indices. Save time if `False`(?)
Returns:
dist_ap: pytorch Variable, distance(anchor, positive); shape [N]
dist_an: pytorch Variable, distance(anchor, negative); shape [N]
p_inds: pytorch LongTensor, with shape [N];
indices of selected hard positive samples; 0 <= p_inds[i] <= N - 1
n_inds: pytorch LongTensor, with shape [N];
indices of selected hard negative samples; 0 <= n_inds[i] <= N - 1
NOTE: Only consider the case in which all labels have same num of samples,
thus we can cope with all anchors in parallel.
"""
assert len(dist_mat.size()) == 2 #是否是二维的
assert dist_mat.size(0) == dist_mat.size(1)
N = dist_mat.size(0)
# shape [N, N]
is_pos = labels.expand(N, N).eq(labels.expand(N, N).t())
is_neg = labels.expand(N, N).ne(labels.expand(N, N).t())
# `dist_ap` means distance(anchor, positive)
# both `dist_ap` and `relative_p_inds` with shape [N, 1]
dist_ap, relative_p_inds = torch.max(
dist_mat[is_pos].contiguous().view(N, -1), 1, keepdim=True)
# `dist_an` means distance(anchor, negative)
# both `dist_an` and `relative_n_inds` with shape [N, 1]
dist_an, relative_n_inds = torch.min(
dist_mat[is_neg].contiguous().view(N, -1), 1, keepdim=True)
# shape [N]
dist_ap = dist_ap.squeeze(1) #去掉size为1的维度
dist_an = dist_an.squeeze(1)
if return_inds:
# shape [N, N]
ind = (labels.new().resize_as_(labels)
.copy_(torch.arange(0, N).long())
.unsqueeze( 0).expand(N, N))
# shape [N, 1]
p_inds = torch.gather(
ind[is_pos].contiguous().view(N, -1), 1, relative_p_inds.data)
n_inds = torch.gather(
ind[is_neg].contiguous().view(N, -1), 1, relative_n_inds.data)
# shape [N]
p_inds = p_inds.squeeze(1)
n_inds = n_inds.squeeze(1)
return dist_ap, dist_an, p_inds, n_inds
return dist_ap, dist_an'''
def hard_example_mining(dist_mat, labels, return_inds=False):
lables=tf.expand_dims(labels,0)
sess=tf.Session()
sess.run(tf.initialize_all_variables())
N=tf.shape(dist_mat)[0]
is_pos=tf.equal(tf.tile(lables,(N,1)),tf.transpose(tf.tile(lables,(N,1))))
is_neg=tf.not_equal(tf.tile(lables,[N,1]),tf.transpose(tf.tile(lables,[N,1])))
temp_ap=sess.run(dist_mat)
temp_ap[sess.run(is_neg)]=0
temp_an=sess.run(dist_mat)
temp_an[sess.run(is_pos)]=0
dist_ap=tf.reduce_max(tf.Variable(temp_ap),1,keep_dims=True)
dist_an=tf.reduce_max(tf.Variable(temp_an),1,keep_dims=True)
dist_ap=tf.squeeze(dist_ap)
dist_an=tf.squeeze(dist_an)
return dist_ap,dist_an
def global_loss(tri_loss, global_feat, labels, normalize_feature=True):
"""
Args:
tri_loss: a `TripletLoss` object
global_feat: pytorch Variable, shape [N, C]
labels: pytorch LongTensor, with shape [N]
normalize_feature: whether to normalize feature to unit length along the
Channel dimension
Returns:
loss: pytorch Variable, with shape [1]
p_inds: pytorch LongTensor, with shape [N];
indices of selected hard positive samples; 0 <= p_inds[i] <= N - 1
n_inds: pytorch LongTensor, with shape [N];
indices of selected hard negative samples; 0 <= n_inds[i] <= N - 1
=============
For Debugging
=============
dist_ap: pytorch Variable, distance(anchor, positive); shape [N]
dist_an: pytorch Variable, distance(anchor, negative); shape [N]
===================
For Mutual Learning
===================
dist_mat: pytorch Variable, pairwise euclidean distance; shape [N, N]
"""
if normalize_feature:
global_feat = normalize(global_feat, axis=-1)
# shape [N, N]
dist_mat = euclidean_dist(global_feat, global_feat)
dist_ap, dist_an= hard_example_mining(
dist_mat, labels, return_inds=True)
loss = tri_loss(dist_ap, dist_an)
return loss,dist_ap, dist_an, dist_mat
def local_loss(
tri_loss,
local_feat,
p_inds=None,
n_inds=None,
labels=None,
normalize_feature=True):
"""
Args:
tri_loss: a `TripletLoss` object
local_feat: pytorch Variable, shape [N, H, c] (NOTE THE SHAPE!)
p_inds: pytorch LongTensor, with shape [N];
indices of selected hard positive samples; 0 <= p_inds[i] <= N - 1
n_inds: pytorch LongTensor, with shape [N];
indices of selected hard negative samples; 0 <= n_inds[i] <= N - 1
labels: pytorch LongTensor, with shape [N]
normalize_feature: whether to normalize feature to unit length along the
Channel dimension
If hard samples are specified by `p_inds` and `n_inds`, then `labels` is not
used. Otherwise, local distance finds its own hard samples independent of
global distance.
Returns:
loss: pytorch Variable,with shape [1]
=============
For Debugging
=============
dist_ap: pytorch Variable, distance(anchor, positive); shape [N]
dist_an: pytorch Variable, distance(anchor, negative); shape [N]
===================
For Mutual Learning
===================
dist_mat: pytorch Variable, pairwise local distance; shape [N, N]
"""
if normalize_feature:
local_feat = normalize(local_feat, axis=-1)
if p_inds is None or n_inds is None:
dist_mat = local_dist(local_feat, local_feat)
dist_ap, dist_an = hard_example_mining(dist_mat, labels, return_inds=False)
loss = tri_loss(dist_ap, dist_an)
return loss, dist_ap, dist_an, dist_mat
else:
dist_ap = batch_local_dist(local_feat, local_feat[p_inds])
dist_an = batch_local_dist(local_feat, local_feat[n_inds])
loss = tri_loss(dist_ap, dist_an)
return loss, dist_ap, dist_an