/
detection.py
2223 lines (1927 loc) · 91.3 KB
/
detection.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
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserve.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
All layers just related to the detection neural network.
"""
from __future__ import print_function
from .layer_function_generator import generate_layer_fn
from .layer_function_generator import autodoc, templatedoc
from ..layer_helper import LayerHelper
from ..framework import Variable
from . import tensor
from . import nn
from . import ops
from ... import compat as cpt
import math
import six
import numpy
from functools import reduce
__all__ = [
'prior_box',
'density_prior_box',
'multi_box_head',
'bipartite_match',
'target_assign',
'detection_output',
'ssd_loss',
'detection_map',
'rpn_target_assign',
'anchor_generator',
'roi_perspective_transform',
'generate_proposal_labels',
'generate_proposals',
'generate_mask_labels',
'iou_similarity',
'box_coder',
'polygon_box_transform',
'yolov3_loss',
'box_clip',
'multiclass_nms',
]
def rpn_target_assign(bbox_pred,
cls_logits,
anchor_box,
anchor_var,
gt_boxes,
is_crowd,
im_info,
rpn_batch_size_per_im=256,
rpn_straddle_thresh=0.0,
rpn_fg_fraction=0.5,
rpn_positive_overlap=0.7,
rpn_negative_overlap=0.3,
use_random=True):
"""
**Target Assign Layer for region proposal network (RPN) in Faster-RCNN detection.**
This layer can be, for given the Intersection-over-Union (IoU) overlap
between anchors and ground truth boxes, to assign classification and
regression targets to each each anchor, these target labels are used for
train RPN. The classification targets is a binary class label (of being
an object or not). Following the paper of Faster-RCNN, the positive labels
are two kinds of anchors: (i) the anchor/anchors with the highest IoU
overlap with a ground-truth box, or (ii) an anchor that has an IoU overlap
higher than rpn_positive_overlap(0.7) with any ground-truth box. Note
that a single ground-truth box may assign positive labels to multiple
anchors. A non-positive anchor is when its IoU ratio is lower than
rpn_negative_overlap (0.3) for all ground-truth boxes. Anchors that are
neither positive nor negative do not contribute to the training objective.
The regression targets are the encoded ground-truth boxes associated with
the positive anchors.
Args:
bbox_pred(Variable): A 3-D Tensor with shape [N, M, 4] represents the
predicted locations of M bounding bboxes. N is the batch size,
and each bounding box has four coordinate values and the layout
is [xmin, ymin, xmax, ymax].
cls_logits(Variable): A 3-D Tensor with shape [N, M, 1] represents the
predicted confidence predictions. N is the batch size, 1 is the
frontground and background sigmoid, M is number of bounding boxes.
anchor_box(Variable): A 2-D Tensor with shape [M, 4] holds M boxes,
each box is represented as [xmin, ymin, xmax, ymax],
[xmin, ymin] is the left top coordinate of the anchor box,
if the input is image feature map, they are close to the origin
of the coordinate system. [xmax, ymax] is the right bottom
coordinate of the anchor box.
anchor_var(Variable): A 2-D Tensor with shape [M,4] holds expanded
variances of anchors.
gt_boxes (Variable): The ground-truth boudding boxes (bboxes) are a 2D
LoDTensor with shape [Ng, 4], Ng is the total number of ground-truth
bboxes of mini-batch input.
is_crowd (Variable): A 1-D LoDTensor which indicates groud-truth is crowd.
im_info (Variable): A 2-D LoDTensor with shape [N, 3]. N is the batch size,
3 is the height, width and scale.
rpn_batch_size_per_im(int): Total number of RPN examples per image.
rpn_straddle_thresh(float): Remove RPN anchors that go outside the image
by straddle_thresh pixels.
rpn_fg_fraction(float): Target fraction of RoI minibatch that is labeled
foreground (i.e. class > 0), 0-th class is background.
rpn_positive_overlap(float): Minimum overlap required between an anchor
and ground-truth box for the (anchor, gt box) pair to be a positive
example.
rpn_negative_overlap(float): Maximum overlap allowed between an anchor
and ground-truth box for the (anchor, gt box) pair to be a negative
examples.
Returns:
tuple:
A tuple(predicted_scores, predicted_location, target_label,
target_bbox, bbox_inside_weight) is returned. The predicted_scores
and predicted_location is the predicted result of the RPN.
The target_label and target_bbox is the ground truth,
respectively. The predicted_location is a 2D Tensor with shape
[F, 4], and the shape of target_bbox is same as the shape of
the predicted_location, F is the number of the foreground
anchors. The predicted_scores is a 2D Tensor with shape
[F + B, 1], and the shape of target_label is same as the shape
of the predicted_scores, B is the number of the background
anchors, the F and B is depends on the input of this operator.
Bbox_inside_weight represents whether the predicted loc is fake_fg
or not and the shape is [F, 4].
Examples:
.. code-block:: python
bbox_pred = layers.data(name='bbox_pred', shape=[100, 4],
append_batch_size=False, dtype='float32')
cls_logits = layers.data(name='cls_logits', shape=[100, 1],
append_batch_size=False, dtype='float32')
anchor_box = layers.data(name='anchor_box', shape=[20, 4],
append_batch_size=False, dtype='float32')
gt_boxes = layers.data(name='gt_boxes', shape=[10, 4],
append_batch_size=False, dtype='float32')
loc_pred, score_pred, loc_target, score_target, bbox_inside_weight =
fluid.layers.rpn_target_assign(bbox_pred=bbox_pred,
cls_logits=cls_logits,
anchor_box=anchor_box,
gt_boxes=gt_boxes)
"""
helper = LayerHelper('rpn_target_assign', **locals())
# Assign target label to anchors
loc_index = helper.create_variable_for_type_inference(dtype='int32')
score_index = helper.create_variable_for_type_inference(dtype='int32')
target_label = helper.create_variable_for_type_inference(dtype='int32')
target_bbox = helper.create_variable_for_type_inference(
dtype=anchor_box.dtype)
bbox_inside_weight = helper.create_variable_for_type_inference(
dtype=anchor_box.dtype)
helper.append_op(
type="rpn_target_assign",
inputs={
'Anchor': anchor_box,
'GtBoxes': gt_boxes,
'IsCrowd': is_crowd,
'ImInfo': im_info
},
outputs={
'LocationIndex': loc_index,
'ScoreIndex': score_index,
'TargetLabel': target_label,
'TargetBBox': target_bbox,
'BBoxInsideWeight': bbox_inside_weight
},
attrs={
'rpn_batch_size_per_im': rpn_batch_size_per_im,
'rpn_straddle_thresh': rpn_straddle_thresh,
'rpn_positive_overlap': rpn_positive_overlap,
'rpn_negative_overlap': rpn_negative_overlap,
'rpn_fg_fraction': rpn_fg_fraction,
'use_random': use_random
})
loc_index.stop_gradient = True
score_index.stop_gradient = True
target_label.stop_gradient = True
target_bbox.stop_gradient = True
bbox_inside_weight.stop_gradient = True
cls_logits = nn.reshape(x=cls_logits, shape=(-1, 1))
bbox_pred = nn.reshape(x=bbox_pred, shape=(-1, 4))
predicted_cls_logits = nn.gather(cls_logits, score_index)
predicted_bbox_pred = nn.gather(bbox_pred, loc_index)
return predicted_cls_logits, predicted_bbox_pred, target_label, target_bbox, bbox_inside_weight
def detection_output(loc,
scores,
prior_box,
prior_box_var,
background_label=0,
nms_threshold=0.3,
nms_top_k=400,
keep_top_k=200,
score_threshold=0.01,
nms_eta=1.0):
"""
**Detection Output Layer for Single Shot Multibox Detector (SSD).**
This operation is to get the detection results by performing following
two steps:
1. Decode input bounding box predictions according to the prior boxes.
2. Get the final detection results by applying multi-class non maximum
suppression (NMS).
Please note, this operation doesn't clip the final output bounding boxes
to the image window.
Args:
loc(Variable): A 3-D Tensor with shape [N, M, 4] represents the
predicted locations of M bounding bboxes. N is the batch size,
and each bounding box has four coordinate values and the layout
is [xmin, ymin, xmax, ymax].
scores(Variable): A 3-D Tensor with shape [N, M, C] represents the
predicted confidence predictions. N is the batch size, C is the
class number, M is number of bounding boxes. For each category
there are total M scores which corresponding M bounding boxes.
prior_box(Variable): A 2-D Tensor with shape [M, 4] holds M boxes,
each box is represented as [xmin, ymin, xmax, ymax],
[xmin, ymin] is the left top coordinate of the anchor box,
if the input is image feature map, they are close to the origin
of the coordinate system. [xmax, ymax] is the right bottom
coordinate of the anchor box.
prior_box_var(Variable): A 2-D Tensor with shape [M, 4] holds M group
of variance.
background_label(float): The index of background label,
the background label will be ignored. If set to -1, then all
categories will be considered.
nms_threshold(float): The threshold to be used in NMS.
nms_top_k(int): Maximum number of detections to be kept according
to the confidences aftern the filtering detections based on
score_threshold.
keep_top_k(int): Number of total bboxes to be kept per image after
NMS step. -1 means keeping all bboxes after NMS step.
score_threshold(float): Threshold to filter out bounding boxes with
low confidence score. If not provided, consider all boxes.
nms_eta(float): The parameter for adaptive NMS.
Returns:
Variable:
The detection outputs is a LoDTensor with shape [No, 6].
Each row has six values: [label, confidence, xmin, ymin, xmax, ymax].
`No` is the total number of detections in this mini-batch. For each
instance, the offsets in first dimension are called LoD, the offset
number is N + 1, N is the batch size. The i-th image has
`LoD[i + 1] - LoD[i]` detected results, if it is 0, the i-th image
has no detected results. If all images have not detected results,
LoD will be set to {1}, and output tensor only contains one
value, which is -1.
(After version 1.3, when no boxes detected, the lod is changed
from {0} to {1}.)
Examples:
.. code-block:: python
pb = layers.data(name='prior_box', shape=[10, 4],
append_batch_size=False, dtype='float32')
pbv = layers.data(name='prior_box_var', shape=[10, 4],
append_batch_size=False, dtype='float32')
loc = layers.data(name='target_box', shape=[2, 21, 4],
append_batch_size=False, dtype='float32')
scores = layers.data(name='scores', shape=[2, 21, 10],
append_batch_size=False, dtype='float32')
nmsed_outs = fluid.layers.detection_output(scores=scores,
loc=loc,
prior_box=pb,
prior_box_var=pbv)
"""
helper = LayerHelper("detection_output", **locals())
decoded_box = box_coder(
prior_box=prior_box,
prior_box_var=prior_box_var,
target_box=loc,
code_type='decode_center_size')
scores = nn.softmax(input=scores)
scores = nn.transpose(scores, perm=[0, 2, 1])
scores.stop_gradient = True
nmsed_outs = helper.create_variable_for_type_inference(
dtype=decoded_box.dtype)
helper.append_op(
type="multiclass_nms",
inputs={'Scores': scores,
'BBoxes': decoded_box},
outputs={'Out': nmsed_outs},
attrs={
'background_label': 0,
'nms_threshold': nms_threshold,
'nms_top_k': nms_top_k,
'keep_top_k': keep_top_k,
'score_threshold': score_threshold,
'nms_eta': 1.0
})
nmsed_outs.stop_gradient = True
return nmsed_outs
@templatedoc()
def iou_similarity(x, y, name=None):
"""
${comment}
Args:
x(${x_type}): ${x_comment}
y(${y_type}): ${y_comment}
Returns:
out(${out_type}): ${out_comment}
"""
helper = LayerHelper("iou_similarity", **locals())
if name is None:
out = helper.create_variable_for_type_inference(dtype=x.dtype)
else:
out = helper.create_variable(
name=name, dtype=x.dtype, persistable=False)
helper.append_op(
type="iou_similarity",
inputs={"X": x,
"Y": y},
attrs={},
outputs={"Out": out})
return out
@templatedoc()
def box_coder(prior_box,
prior_box_var,
target_box,
code_type="encode_center_size",
box_normalized=True,
name=None,
axis=0):
"""
**Box Coder Layer**
Encode/Decode the target bounding box with the priorbox information.
The Encoding schema described below:
.. math::
ox = (tx - px) / pw / pxv
oy = (ty - py) / ph / pyv
ow = \log(\abs(tw / pw)) / pwv
oh = \log(\abs(th / ph)) / phv
The Decoding schema described below:
.. math::
ox = (pw * pxv * tx * + px) - tw / 2
oy = (ph * pyv * ty * + py) - th / 2
ow = \exp(pwv * tw) * pw + tw / 2
oh = \exp(phv * th) * ph + th / 2
where `tx`, `ty`, `tw`, `th` denote the target box's center coordinates,
width and height respectively. Similarly, `px`, `py`, `pw`, `ph` denote
the priorbox's (anchor) center coordinates, width and height. `pxv`,
`pyv`, `pwv`, `phv` denote the variance of the priorbox and `ox`, `oy`,
`ow`, `oh` denote the encoded/decoded coordinates, width and height.
During Box Decoding, two modes for broadcast are supported. Say target
box has shape [N, M, 4], and the shape of prior box can be [N, 4] or
[M, 4]. Then prior box will broadcast to target box along the
assigned axis.
Args:
prior_box(Variable): Box list prior_box is a 2-D Tensor with shape
[M, 4] holds M boxes, each box is represented as
[xmin, ymin, xmax, ymax], [xmin, ymin] is the
left top coordinate of the anchor box, if the
input is image feature map, they are close to
the origin of the coordinate system. [xmax, ymax]
is the right bottom coordinate of the anchor box.
prior_box_var(Variable|list): prior_box_var supports two types of input.
One is variable with shape [M, 4] holds M group.
The other one is list consist of 4 elements
shared by all boxes.
target_box(Variable): This input can be a 2-D LoDTensor with shape
[N, 4] when code_type is 'encode_center_size'.
This input also can be a 3-D Tensor with shape
[N, M, 4] when code_type is 'decode_center_size'.
Each box is represented as
[xmin, ymin, xmax, ymax]. This tensor can
contain LoD information to represent a batch
of inputs.
code_type(string): The code type used with the target box. It can be
encode_center_size or decode_center_size
box_normalized(int): Whether treat the priorbox as a noramlized box.
Set true by default.
name(string): The name of box coder.
axis(int): Which axis in PriorBox to broadcast for box decode,
for example, if axis is 0 and TargetBox has shape
[N, M, 4] and PriorBox has shape [M, 4], then PriorBox
will broadcast to [N, M, 4] for decoding. It is only valid
when code type is decode_center_size. Set 0 by default.
Returns:
output_box(Variable): When code_type is 'encode_center_size', the
output tensor of box_coder_op with shape
[N, M, 4] representing the result of N target
boxes encoded with M Prior boxes and variances.
When code_type is 'decode_center_size',
N represents the batch size and M represents
the number of deocded boxes.
Examples:
.. code-block:: python
prior_box = fluid.layers.data(name='prior_box',
shape=[512, 4],
dtype='float32',
append_batch_size=False)
target_box = fluid.layers.data(name='target_box',
shape=[512,81,4],
dtype='float32',
append_batch_size=False)
output = fluid.layers.box_coder(prior_box=prior_box,
prior_box_var=[0.1,0.1,0.2,0.2],
target_box=target_box,
code_type="decode_center_size",
box_normalized=False,
axis=1)
"""
helper = LayerHelper("box_coder", **locals())
if name is None:
output_box = helper.create_variable_for_type_inference(
dtype=prior_box.dtype)
else:
output_box = helper.create_variable(
name=name, dtype=prior_box.dtype, persistable=False)
inputs = {"PriorBox": prior_box, "TargetBox": target_box}
attrs = {
"code_type": code_type,
"box_normalized": box_normalized,
"axis": axis
}
if isinstance(prior_box_var, Variable):
inputs['PriorBoxVar'] = prior_box_var
elif isinstance(prior_box_var, list):
attrs['variance'] = prior_box_var
else:
raise TypeError("Input variance of box_coder must be Variable or lisz")
helper.append_op(
type="box_coder",
inputs=inputs,
attrs=attrs,
outputs={"OutputBox": output_box})
return output_box
@templatedoc()
def polygon_box_transform(input, name=None):
"""
${comment}
Args:
input(${input_type}): ${input_comment}
Returns:
output(${output_type}): ${output_comment}
"""
helper = LayerHelper("polygon_box_transform", **locals())
if name is None:
output = helper.create_variable_for_type_inference(dtype=input.dtype)
else:
output = helper.create_variable(
name=name, dtype=prior_box.input, persistable=False)
helper.append_op(
type="polygon_box_transform",
inputs={"Input": input},
attrs={},
outputs={"Output": output})
return output
@templatedoc(op_type="yolov3_loss")
def yolov3_loss(x,
gtbox,
gtlabel,
anchors,
anchor_mask,
class_num,
ignore_thresh,
downsample_ratio,
name=None):
"""
${comment}
Args:
x (Variable): ${x_comment}
gtbox (Variable): groud truth boxes, should be in shape of [N, B, 4],
in the third dimenstion, x, y, w, h should be stored
and x, y, w, h should be relative value of input image.
N is the batch number and B is the max box number in
an image.
gtlabel (Variable): class id of ground truth boxes, shoud be in shape
of [N, B].
anchors (list|tuple): ${anchors_comment}
anchor_mask (list|tuple): ${anchor_mask_comment}
class_num (int): ${class_num_comment}
ignore_thresh (float): ${ignore_thresh_comment}
downsample_ratio (int): ${downsample_ratio_comment}
name (string): the name of yolov3 loss
Returns:
Variable: A 1-D tensor with shape [1], the value of yolov3 loss
Raises:
TypeError: Input x of yolov3_loss must be Variable
TypeError: Input gtbox of yolov3_loss must be Variable"
TypeError: Input gtlabel of yolov3_loss must be Variable"
TypeError: Attr anchors of yolov3_loss must be list or tuple
TypeError: Attr class_num of yolov3_loss must be an integer
TypeError: Attr ignore_thresh of yolov3_loss must be a float number
Examples:
.. code-block:: python
x = fluid.layers.data(name='x', shape=[255, 13, 13], dtype='float32')
gtbox = fluid.layers.data(name='gtbox', shape=[6, 5], dtype='float32')
gtlabel = fluid.layers.data(name='gtlabel', shape=[6, 1], dtype='int32')
anchors = [10, 13, 16, 30, 33, 23, 30, 61, 62, 45, 59, 119, 116, 90, 156, 198, 373, 326]
anchor_mask = [0, 1, 2]
loss = fluid.layers.yolov3_loss(x=x, gtbox=gtbox, gtlabel=gtlabel, anchors=anchors,
anchor_mask=anchor_mask, class_num=80,
ignore_thresh=0.7, downsample_ratio=32)
"""
helper = LayerHelper('yolov3_loss', **locals())
if not isinstance(x, Variable):
raise TypeError("Input x of yolov3_loss must be Variable")
if not isinstance(gtbox, Variable):
raise TypeError("Input gtbox of yolov3_loss must be Variable")
if not isinstance(gtlabel, Variable):
raise TypeError("Input gtlabel of yolov3_loss must be Variable")
if not isinstance(anchors, list) and not isinstance(anchors, tuple):
raise TypeError("Attr anchors of yolov3_loss must be list or tuple")
if not isinstance(anchor_mask, list) and not isinstance(anchor_mask, tuple):
raise TypeError("Attr anchor_mask of yolov3_loss must be list or tuple")
if not isinstance(class_num, int):
raise TypeError("Attr class_num of yolov3_loss must be an integer")
if not isinstance(ignore_thresh, float):
raise TypeError(
"Attr ignore_thresh of yolov3_loss must be a float number")
if name is None:
loss = helper.create_variable_for_type_inference(dtype=x.dtype)
else:
loss = helper.create_variable(
name=name, dtype=x.dtype, persistable=False)
objectness_mask = helper.create_variable_for_type_inference(dtype='int32')
gt_match_mask = helper.create_variable_for_type_inference(dtype='int32')
attrs = {
"anchors": anchors,
"anchor_mask": anchor_mask,
"class_num": class_num,
"ignore_thresh": ignore_thresh,
"downsample_ratio": downsample_ratio,
}
helper.append_op(
type='yolov3_loss',
inputs={
"X": x,
"GTBox": gtbox,
"GTLabel": gtlabel,
},
outputs={
'Loss': loss,
'ObjectnessMask': objectness_mask,
'GTMatchMask': gt_match_mask
},
attrs=attrs)
return loss
@templatedoc()
def detection_map(detect_res,
label,
class_num,
background_label=0,
overlap_threshold=0.3,
evaluate_difficult=True,
has_state=None,
input_states=None,
out_states=None,
ap_version='integral'):
"""
${comment}
Args:
detect_res: ${detect_res_comment}
label: ${label_comment}
class_num: ${class_num_comment}
background_label: ${background_label_comment}
overlap_threshold: ${overlap_threshold_comment}
evaluate_difficult: ${evaluate_difficult_comment}
has_state: ${has_state_comment}
input_states: If not None, It contains 3 elements:
1. pos_count ${pos_count_comment}.
2. true_pos ${true_pos_comment}.
3. false_pos ${false_pos_comment}.
out_states: If not None, it contains 3 elements.
1. accum_pos_count ${accum_pos_count_comment}.
2. accum_true_pos ${accum_true_pos_comment}.
3. accum_false_pos ${accum_false_pos_comment}.
ap_version: ${ap_type_comment}
Returns:
${map_comment}
Examples:
.. code-block:: python
detect_res = fluid.layers.data(
name='detect_res',
shape=[10, 6],
append_batch_size=False,
dtype='float32')
label = fluid.layers.data(
name='label',
shape=[10, 6],
append_batch_size=False,
dtype='float32')
map_out = fluid.layers.detection_map(detect_res, label, 21)
"""
helper = LayerHelper("detection_map", **locals())
def __create_var(type):
return helper.create_variable_for_type_inference(dtype=type)
map_out = __create_var('float32')
accum_pos_count_out = out_states[0] if out_states else __create_var('int32')
accum_true_pos_out = out_states[1] if out_states else __create_var(
'float32')
accum_false_pos_out = out_states[2] if out_states else __create_var(
'float32')
pos_count = input_states[0] if input_states else None
true_pos = input_states[1] if input_states else None
false_pos = input_states[2] if input_states else None
helper.append_op(
type="detection_map",
inputs={
'Label': label,
'DetectRes': detect_res,
'HasState': has_state,
'PosCount': pos_count,
'TruePos': true_pos,
'FalsePos': false_pos
},
outputs={
'MAP': map_out,
'AccumPosCount': accum_pos_count_out,
'AccumTruePos': accum_true_pos_out,
'AccumFalsePos': accum_false_pos_out
},
attrs={
'overlap_threshold': overlap_threshold,
'evaluate_difficult': evaluate_difficult,
'ap_type': ap_version,
'class_num': class_num,
})
return map_out
def bipartite_match(dist_matrix,
match_type=None,
dist_threshold=None,
name=None):
"""
This operator implements a greedy bipartite matching algorithm, which is
used to obtain the matching with the maximum distance based on the input
distance matrix. For input 2D matrix, the bipartite matching algorithm can
find the matched column for each row (matched means the largest distance),
also can find the matched row for each column. And this operator only
calculate matched indices from column to row. For each instance,
the number of matched indices is the column number of the input distance
matrix.
There are two outputs, matched indices and distance.
A simple description, this algorithm matched the best (maximum distance)
row entity to the column entity and the matched indices are not duplicated
in each row of ColToRowMatchIndices. If the column entity is not matched
any row entity, set -1 in ColToRowMatchIndices.
NOTE: the input DistMat can be LoDTensor (with LoD) or Tensor.
If LoDTensor with LoD, the height of ColToRowMatchIndices is batch size.
If Tensor, the height of ColToRowMatchIndices is 1.
NOTE: This API is a very low level API. It is used by :code:`ssd_loss`
layer. Please consider to use :code:`ssd_loss` instead.
Args:
dist_matrix(Variable): This input is a 2-D LoDTensor with shape
[K, M]. It is pair-wise distance matrix between the entities
represented by each row and each column. For example, assumed one
entity is A with shape [K], another entity is B with shape [M]. The
dist_matrix[i][j] is the distance between A[i] and B[j]. The bigger
the distance is, the better matching the pairs are.
NOTE: This tensor can contain LoD information to represent a batch
of inputs. One instance of this batch can contain different numbers
of entities.
match_type(string|None): The type of matching method, should be
'bipartite' or 'per_prediction'. [default 'bipartite'].
dist_threshold(float|None): If `match_type` is 'per_prediction',
this threshold is to determine the extra matching bboxes based
on the maximum distance, 0.5 by default.
Returns:
tuple: a tuple with two elements is returned. The first is
matched_indices, the second is matched_distance.
The matched_indices is a 2-D Tensor with shape [N, M] in int type.
N is the batch size. If match_indices[i][j] is -1, it
means B[j] does not match any entity in i-th instance.
Otherwise, it means B[j] is matched to row
match_indices[i][j] in i-th instance. The row number of
i-th instance is saved in match_indices[i][j].
The matched_distance is a 2-D Tensor with shape [N, M] in float type
. N is batch size. If match_indices[i][j] is -1,
match_distance[i][j] is also -1.0. Otherwise, assumed
match_distance[i][j] = d, and the row offsets of each instance
are called LoD. Then match_distance[i][j] =
dist_matrix[d+LoD[i]][j].
Examples:
>>> x = fluid.layers.data(name='x', shape=[4], dtype='float32')
>>> y = fluid.layers.data(name='y', shape=[4], dtype='float32')
>>> iou = fluid.layers.iou_similarity(x=x, y=y)
>>> matched_indices, matched_dist = fluid.layers.bipartite_match(iou)
"""
helper = LayerHelper('bipartite_match', **locals())
match_indices = helper.create_variable_for_type_inference(dtype='int32')
match_distance = helper.create_variable_for_type_inference(
dtype=dist_matrix.dtype)
helper.append_op(
type='bipartite_match',
inputs={'DistMat': dist_matrix},
attrs={
'match_type': match_type,
'dist_threshold': dist_threshold,
},
outputs={
'ColToRowMatchIndices': match_indices,
'ColToRowMatchDist': match_distance
})
return match_indices, match_distance
def target_assign(input,
matched_indices,
negative_indices=None,
mismatch_value=None,
name=None):
"""
This operator can be, for given the target bounding boxes or labels,
to assign classification and regression targets to each prediction as well as
weights to prediction. The weights is used to specify which prediction would
not contribute to training loss.
For each instance, the output `out` and`out_weight` are assigned based on
`match_indices` and `negative_indices`.
Assumed that the row offset for each instance in `input` is called lod,
this operator assigns classification/regression targets by performing the
following steps:
1. Assigning all outpts based on `match_indices`:
.. code-block:: text
If id = match_indices[i][j] > 0,
out[i][j][0 : K] = X[lod[i] + id][j % P][0 : K]
out_weight[i][j] = 1.
Otherwise,
out[j][j][0 : K] = {mismatch_value, mismatch_value, ...}
out_weight[i][j] = 0.
2. Assigning out_weight based on `neg_indices` if `neg_indices` is provided:
Assumed that the row offset for each instance in `neg_indices` is called neg_lod,
for i-th instance and each `id` of neg_indices in this instance:
.. code-block:: text
out[i][id][0 : K] = {mismatch_value, mismatch_value, ...}
out_weight[i][id] = 1.0
Args:
inputs (Variable): This input is a 3D LoDTensor with shape [M, P, K].
matched_indices (Variable): Tensor<int>), The input matched indices
is 2D Tenosr<int32> with shape [N, P], If MatchIndices[i][j] is -1,
the j-th entity of column is not matched to any entity of row in
i-th instance.
negative_indices (Variable): The input negative example indices are
an optional input with shape [Neg, 1] and int32 type, where Neg is
the total number of negative example indices.
mismatch_value (float32): Fill this value to the mismatched location.
Returns:
tuple:
A tuple(out, out_weight) is returned. out is a 3D Tensor with
shape [N, P, K], N and P is the same as they are in
`neg_indices`, K is the same as it in input of X. If
`match_indices[i][j]`. out_weight is the weight for output with
the shape of [N, P, 1].
Examples:
.. code-block:: python
matched_indices, matched_dist = fluid.layers.bipartite_match(iou)
gt = layers.data(
name='gt', shape=[1, 1], dtype='int32', lod_level=1)
trg, trg_weight = layers.target_assign(
gt, matched_indices, mismatch_value=0)
"""
helper = LayerHelper('target_assign', **locals())
out = helper.create_variable_for_type_inference(dtype=input.dtype)
out_weight = helper.create_variable_for_type_inference(dtype='float32')
helper.append_op(
type='target_assign',
inputs={
'X': input,
'MatchIndices': matched_indices,
'NegIndices': negative_indices
},
outputs={'Out': out,
'OutWeight': out_weight},
attrs={'mismatch_value': mismatch_value})
return out, out_weight
def ssd_loss(location,
confidence,
gt_box,
gt_label,
prior_box,
prior_box_var=None,
background_label=0,
overlap_threshold=0.5,
neg_pos_ratio=3.0,
neg_overlap=0.5,
loc_loss_weight=1.0,
conf_loss_weight=1.0,
match_type='per_prediction',
mining_type='max_negative',
normalize=True,
sample_size=None):
"""
**Multi-box loss layer for object detection algorithm of SSD**
This layer is to compute dection loss for SSD given the location offset
predictions, confidence predictions, prior boxes and ground-truth boudding
boxes and labels, and the type of hard example mining. The returned loss
is a weighted sum of the localization loss (or regression loss) and
confidence loss (or classification loss) by performing the following steps:
1. Find matched bounding box by bipartite matching algorithm.
1.1 Compute IOU similarity between ground-truth boxes and prior boxes.
1.2 Compute matched boundding box by bipartite matching algorithm.
2. Compute confidence for mining hard examples
2.1. Get the target label based on matched indices.
2.2. Compute confidence loss.
3. Apply hard example mining to get the negative example indices and update
the matched indices.
4. Assign classification and regression targets
4.1. Encoded bbox according to the prior boxes.
4.2. Assign regression targets.
4.3. Assign classification targets.
5. Compute the overall objective loss.
5.1 Compute confidence loss.
5.1 Compute localization loss.
5.3 Compute the overall weighted loss.
Args:
location (Variable): The location predictions are a 3D Tensor with
shape [N, Np, 4], N is the batch size, Np is total number of
predictions for each instance. 4 is the number of coordinate values,
the layout is [xmin, ymin, xmax, ymax].
confidence (Variable): The confidence predictions are a 3D Tensor
with shape [N, Np, C], N and Np are the same as they are in
`location`, C is the class number.
gt_box (Variable): The ground-truth boudding boxes (bboxes) are a 2D
LoDTensor with shape [Ng, 4], Ng is the total number of ground-truth
bboxes of mini-batch input.
gt_label (Variable): The ground-truth labels are a 2D LoDTensor
with shape [Ng, 1].
prior_box (Variable): The prior boxes are a 2D Tensor with shape [Np, 4].
prior_box_var (Variable): The variance of prior boxes are a 2D Tensor
with shape [Np, 4].
background_label (int): The index of background label, 0 by default.
overlap_threshold (float): If match_type is 'per_prediction', use
`overlap_threshold` to determine the extra matching bboxes when
finding matched boxes. 0.5 by default.
neg_pos_ratio (float): The ratio of the negative boxes to the positive
boxes, used only when mining_type is 'max_negative', 3.0 by defalut.
neg_overlap (float): The negative overlap upper bound for the unmatched
predictions. Use only when mining_type is 'max_negative',
0.5 by default.
loc_loss_weight (float): Weight for localization loss, 1.0 by default.
conf_loss_weight (float): Weight for confidence loss, 1.0 by default.
match_type (str): The type of matching method during training, should
be 'bipartite' or 'per_prediction', 'per_prediction' by defalut.
mining_type (str): The hard example mining type, should be 'hard_example'
or 'max_negative', now only support `max_negative`.
normalize (bool): Whether to normalize the SSD loss by the total number
of output locations, True by default.
sample_size (int): The max sample size of negative box, used only when
mining_type is 'hard_example'.
Returns:
The weighted sum of the localization loss and confidence loss, with \
shape [N * Np, 1], N and Np are the same as they are in `location`.
Raises:
ValueError: If mining_type is 'hard_example', now only support mining \
type of `max_negative`.
Examples:
>>> pb = fluid.layers.data(
>>> name='prior_box',
>>> shape=[10, 4],
>>> append_batch_size=False,
>>> dtype='float32')
>>> pbv = fluid.layers.data(
>>> name='prior_box_var',
>>> shape=[10, 4],
>>> append_batch_size=False,
>>> dtype='float32')
>>> loc = fluid.layers.data(name='target_box', shape=[10, 4], dtype='float32')
>>> scores = fluid.layers.data(name='scores', shape=[10, 21], dtype='float32')
>>> gt_box = fluid.layers.data(
>>> name='gt_box', shape=[4], lod_level=1, dtype='float32')
>>> gt_label = fluid.layers.data(
>>> name='gt_label', shape=[1], lod_level=1, dtype='float32')
>>> loss = fluid.layers.ssd_loss(loc, scores, gt_box, gt_label, pb, pbv)
"""
helper = LayerHelper('ssd_loss', **locals())
if mining_type != 'max_negative':
raise ValueError("Only support mining_type == max_negative now.")
num, num_prior, num_class = confidence.shape
conf_shape = nn.shape(confidence)