-
Notifications
You must be signed in to change notification settings - Fork 273
/
object_detection_metrics.py
726 lines (647 loc) · 31.1 KB
/
object_detection_metrics.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
# Copyright 2019 Google LLC
#
# 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
#
# https://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.
"""COCO object detection metrics."""
import logging
import math
from typing import Any, Dict, List, Optional, Tuple, Union
import numpy as np
from tensorflow_model_analysis.metrics import confusion_matrix_metrics
from tensorflow_model_analysis.metrics import metric_types
from tensorflow_model_analysis.metrics import metric_util
from tensorflow_model_analysis.metrics import object_detection_confusion_matrix_metrics
AVERAGE_RECALL_NAME = 'average_recall'
AVERAGE_PRECISION_NAME = 'average_precision'
MEAN_AVERAGE_PRECISION_NAME = 'mean_average_precision'
MEAN_AVERAGE_RECALL_NAME = 'mean_average_recall'
class COCOAveragePrecision(metric_types.Metric):
"""Confusion matrix at thresholds.
It computes the average precision of object detections for a single class and
a single iou_threshold.
"""
def __init__(self,
num_thresholds: Optional[int] = None,
iou_threshold: Optional[float] = None,
class_id: Optional[int] = None,
class_weight: Optional[float] = None,
area_range: Optional[Tuple[float, float]] = None,
max_num_detections: Optional[int] = None,
recalls: Optional[List[float]] = None,
num_recalls: Optional[int] = None,
name: Optional[str] = None,
labels_to_stack: Optional[List[str]] = None,
predictions_to_stack: Optional[List[str]] = None,
num_detections_key: Optional[str] = None,
allow_missing_key: bool = False):
"""Initialize average precision metric.
This metric is only used in object-detection setting. It does not support
sub_key parameters due to the matching algorithm of bounding boxes.
The metric supports using multiple outputs to form the labels/predictions if
the user specifies the label/predcition keys to stack. In this case, the
metric is not expected to work with multi-outputs. The metric only supports
multi outputs if the output of model is already pre-stacked in the expected
format, i.e. ['xmin', 'ymin', 'xmax', 'ymax', 'class_id'] for labels and
['xmin', 'ymin', 'xmax', 'ymax', 'class_id', 'confidence scores'] for
predictions.
Args:
num_thresholds: (Optional) Number of thresholds to use for calculating the
matrices and finding the precision at given recall.
iou_threshold: (Optional) Threholds for a detection and ground truth pair
with specific iou to be considered as a match.
class_id: (Optional) The class id for calculating metrics.
class_weight: (Optional) The weight associated with the object class id.
area_range: (Optional) The area-range for objects to be considered for
metrics.
max_num_detections: (Optional) The maximum number of detections for a
single image.
recalls: (Optional) recalls at which precisions will be calculated.
num_recalls: (Optional) Used for objecth detection, the number of recalls
for calculating average precision, it equally generates points bewteen 0
and 1. (Only one of recalls and num_recalls should be used).
name: (Optional) string name of the metric instance.
labels_to_stack: (Optional) Keys for columns to be stacked as a single
numpy array as the labels. It is searched under the key labels, features
and transformed features. The desired format is [left bounadary, top
boudnary, right boundary, bottom boundary, class id]. e.g. ['xmin',
'ymin', 'xmax', 'ymax', 'class_id']
predictions_to_stack: (Optional) Output names for columns to be stacked as
a single numpy array as the prediction. It should be the model's output
names. The desired format is [left bounadary, top boudnary, right
boundary, bottom boundary, class id, confidence score]. e.g. ['xmin',
'ymin', 'xmax', 'ymax', 'class_id', 'scores']
num_detections_key: (Optional) An output name in which to find the number
of detections to use for evaluation for a given example. It does nothing
if predictions_to_stack is not set. The value for this output should be
a scalar value or a single-value tensor. The stacked predicitions will
be truncated with the specified number of detections.
allow_missing_key: (Optional) If true, the preprocessor will return empty
array instead of raising errors.
"""
if recalls is not None:
recall_thresholds = recalls
elif num_recalls is not None:
recall_thresholds = np.linspace(0.0, 1.0, num_recalls)
else:
# by default set recall_thresholds to [0.0:0.01:1.0].
recall_thresholds = np.linspace(0.0, 1.0, 101)
super().__init__(
metric_util.merge_per_key_computations(self._metric_computations),
num_thresholds=num_thresholds,
iou_threshold=iou_threshold,
class_id=class_id,
class_weight=class_weight,
area_range=area_range,
max_num_detections=max_num_detections,
recall_thresholds=recall_thresholds,
name=name,
labels_to_stack=labels_to_stack,
predictions_to_stack=predictions_to_stack,
num_detections_key=num_detections_key,
allow_missing_key=allow_missing_key)
def _default_name(self) -> str:
return AVERAGE_PRECISION_NAME
def _metric_computations(
self,
num_thresholds: Optional[int] = None,
iou_threshold: Optional[float] = None,
class_id: Optional[int] = None,
class_weight: Optional[float] = None,
max_num_detections: Optional[int] = None,
area_range: Optional[Tuple[float, float]] = None,
recall_thresholds: Optional[List[float]] = None,
name: Optional[str] = None,
model_name: str = '',
output_name: str = '',
example_weighted: bool = False,
labels_to_stack: Optional[List[str]] = None,
predictions_to_stack: Optional[List[str]] = None,
num_detections_key: Optional[str] = None,
allow_missing_key: bool = False,
) -> metric_types.MetricComputations:
"""Returns computations for confusion matrix metric."""
metric_util.validate_object_detection_arguments(
class_id=class_id,
class_weight=class_weight,
area_range=area_range,
max_num_detections=max_num_detections,
labels_to_stack=labels_to_stack,
predictions_to_stack=predictions_to_stack,
output_name=output_name)
key = metric_types.MetricKey(
name=name,
model_name=model_name,
output_name=output_name,
sub_key=None,
example_weighted=example_weighted,
aggregation_type=None)
if recall_thresholds is None:
# If recall thresholds is not defined, initialize it as [0.0]
recall_thresholds = [0.0]
if num_thresholds is None:
num_thresholds = 10000
thresholds = [1.e-12] + [
(i + 1) * 1.0 / (num_thresholds - 1) for i in range(num_thresholds - 2)
] + [1.0 - 1.e-12]
# PrecisionAtRecall is a public function. To hide it from users who do not
# need it, we make the name private with '_'.
precision_at_recall_name = metric_util.generate_private_name_from_arguments(
confusion_matrix_metrics.PRECISION_AT_RECALL_NAME,
recall=recall_thresholds,
num_thresholds=num_thresholds,
iou_threshold=iou_threshold,
class_id=class_id,
class_weight=class_weight,
area_range=area_range,
max_num_detections=max_num_detections,
allow_missing_key=allow_missing_key)
pr = (
object_detection_confusion_matrix_metrics
.ObjectDetectionPrecisionAtRecall(
recall=recall_thresholds,
thresholds=thresholds,
iou_threshold=iou_threshold,
class_id=class_id,
class_weight=class_weight,
area_range=area_range,
max_num_detections=max_num_detections,
name=precision_at_recall_name,
labels_to_stack=labels_to_stack,
predictions_to_stack=predictions_to_stack,
num_detections_key=num_detections_key,
allow_missing_key=allow_missing_key))
computations = pr.computations(
model_names=[model_name], output_names=[output_name])
precisions_key = computations[-1].keys[-1]
def result(
metrics: Dict[metric_types.MetricKey, Any]
) -> Dict[metric_types.MetricKey, Union[float, np.ndarray]]:
value = np.nanmean(metrics[precisions_key])
return {key: value}
derived_computation = metric_types.DerivedMetricComputation(
keys=[key], result=result)
computations.append(derived_computation)
return computations
metric_types.register_metric(COCOAveragePrecision)
class COCOMeanAveragePrecision(metric_types.Metric):
"""Mean average precision for object detections.
It calculates the mean average precision metric for object detections. It
averages COCOAveragePrecision over multiple classes and IoU thresholds.
"""
def __init__(self,
num_thresholds: Optional[int] = None,
iou_thresholds: Optional[List[float]] = None,
class_ids: Optional[List[int]] = None,
class_weights: Optional[List[float]] = None,
area_range: Optional[Tuple[float, float]] = None,
max_num_detections: Optional[int] = None,
recalls: Optional[List[float]] = None,
num_recalls: Optional[int] = None,
name: Optional[str] = None,
labels_to_stack: Optional[List[str]] = None,
predictions_to_stack: Optional[List[str]] = None,
num_detections_key: Optional[str] = None,
allow_missing_key: bool = False):
"""Initializes mean average precision metric.
This metric is only used in object-detection setting. It does not support
sub_key parameters due to the matching algorithm of bounding boxes.
The metric supports using multiple outputs to form the labels/predictions if
the user specifies the label/predcition keys to stack. In this case, the
metric is not expected to work with multi-outputs. The metric only supports
multi outputs if the output of model is already pre-stacked in the expected
format, i.e. ['xmin', 'ymin', 'xmax', 'ymax', 'class_id'] for labels and
['xmin', 'ymin', 'xmax', 'ymax', 'class_id', 'confidence scores'] for
predictions.
Args:
num_thresholds: (Optional) Number of thresholds to use for calculating the
matrices and finding the precision at given recall.
iou_thresholds: (Optional) Threholds for a detection and ground truth pair
with specific iou to be considered as a match.
class_ids: (Optional) The class ids for calculating metrics.
class_weights: (Optional) The weight associated with the object class ids.
If it is provided, it should have the same length as class_ids.
area_range: (Optional) The area-range for objects to be considered for
metrics.
max_num_detections: (Optional) The maximum number of detections for a
single image.
recalls: (Optional) recalls at which precisions will be calculated.
num_recalls: (Optional) Used for objecth detection, the number of recalls
for calculating average precision, it equally generates points bewteen 0
and 1. (Only one of recalls and num_recalls should be used).
name: (Optional) Metric name.
labels_to_stack: (Optional) Keys for columns to be stacked as a single
numpy array as the labels. It is searched under the key labels, features
and transformed features. The desired format is [left bounadary, top
boudnary, right boundary, bottom boundary, class id]. e.g. ['xmin',
'ymin', 'xmax', 'ymax', 'class_id']
predictions_to_stack: (Optional) Output names for columns to be stacked as
a single numpy array as the prediction. It should be the model's output
names. The desired format is [left bounadary, top boudnary, right
boundary, bottom boundary, class id, confidence score]. e.g. ['xmin',
'ymin', 'xmax', 'ymax', 'class_id', 'scores']
num_detections_key: (Optional) An output name in which to find the number
of detections to use for evaluation for a given example. It does nothing
if predictions_to_stack is not set. The value for this output should be
a scalar value or a single-value tensor. The stacked predicitions will
be truncated with the specified number of detections.
allow_missing_key: (Optional) If true, the preprocessor will return empty
array instead of raising errors.
"""
super().__init__(
metric_util.merge_per_key_computations(self._metric_computations),
num_thresholds=num_thresholds,
iou_thresholds=iou_thresholds,
class_ids=class_ids,
class_weights=class_weights,
area_range=area_range,
max_num_detections=max_num_detections,
recalls=recalls,
num_recalls=num_recalls,
name=name,
labels_to_stack=labels_to_stack,
predictions_to_stack=predictions_to_stack,
num_detections_key=num_detections_key,
allow_missing_key=allow_missing_key)
def _default_name(self) -> str:
return MEAN_AVERAGE_PRECISION_NAME
def _metric_computations(self,
num_thresholds: Optional[int] = None,
iou_thresholds: Optional[List[float]] = None,
class_ids: Optional[List[int]] = None,
class_weights: Optional[List[float]] = None,
max_num_detections: Optional[int] = None,
area_range: Optional[Tuple[float, float]] = None,
recalls: Optional[List[float]] = None,
num_recalls: Optional[int] = None,
name: Optional[str] = None,
model_name: str = '',
output_name: str = '',
example_weighted: bool = False,
labels_to_stack: Optional[List[str]] = None,
predictions_to_stack: Optional[List[str]] = None,
num_detections_key: Optional[str] = None,
allow_missing_key: bool = False,
**kwargs) -> metric_types.MetricComputations:
"""Returns computations for confusion matrix metric."""
metric_util.validate_object_detection_arguments(
class_id=class_ids,
class_weight=class_weights,
area_range=area_range,
max_num_detections=max_num_detections,
labels_to_stack=labels_to_stack,
predictions_to_stack=predictions_to_stack,
output_name=output_name)
# set default value according to COCO metrics
if iou_thresholds is None:
iou_thresholds = np.linspace(0.5, 0.95, 10)
if class_weights is None:
class_weights = [1.0] * len(class_ids)
key = metric_types.MetricKey(
name=name,
model_name=model_name,
output_name=output_name,
sub_key=None,
example_weighted=example_weighted,
aggregation_type=None)
computations = []
precisions_keys = []
for iou_threshold in iou_thresholds:
for class_id, class_weight in zip(class_ids, class_weights):
average_precision_name = (
metric_util.generate_private_name_from_arguments(
AVERAGE_PRECISION_NAME,
recall=recalls,
num_recalls=num_recalls,
num_thresholds=num_thresholds,
iou_threshold=iou_threshold,
class_id=class_id,
class_weight=class_weight,
area_range=area_range,
max_num_detections=max_num_detections,
allow_missing_key=allow_missing_key))
ap = COCOAveragePrecision(
num_thresholds=num_thresholds,
iou_threshold=iou_threshold,
class_id=class_id,
class_weight=class_weight,
area_range=area_range,
max_num_detections=max_num_detections,
recalls=recalls,
num_recalls=num_recalls,
name=average_precision_name,
labels_to_stack=labels_to_stack,
predictions_to_stack=predictions_to_stack,
num_detections_key=num_detections_key,
allow_missing_key=allow_missing_key)
computations.extend(
ap.computations(
model_names=[model_name], output_names=[output_name]))
precisions_keys.append(computations[-1].keys[-1])
def result(
metrics: Dict[metric_types.MetricKey, Any]
) -> Dict[metric_types.MetricKey, Union[float, np.ndarray]]:
precisions = [
metrics[precisions_key] for precisions_key in precisions_keys
]
value = np.nanmean(precisions)
return {key: value}
derived_computation = metric_types.DerivedMetricComputation(
keys=[key], result=result)
computations.append(derived_computation)
return computations
metric_types.register_metric(COCOMeanAveragePrecision)
class COCOAverageRecall(metric_types.Metric):
"""Average recall metric for object detection.
It computes the average precision metric for object detections for a single
class. It averages MaxRecall metric over mulitple IoU thresholds.
"""
def __init__(self,
iou_thresholds: Optional[List[float]] = None,
class_id: Optional[int] = None,
class_weight: Optional[float] = None,
area_range: Optional[Tuple[float, float]] = None,
max_num_detections: Optional[int] = None,
name: Optional[str] = None,
labels_to_stack: Optional[List[str]] = None,
predictions_to_stack: Optional[List[str]] = None,
num_detections_key: Optional[str] = None,
allow_missing_key: bool = False):
"""Initializes average recall metric.
This metric is only used in object-detection setting. It does not support
sub_key parameters due to the matching algorithm of bounding boxes.
The metric supports using multiple outputs to form the labels/predictions if
the user specifies the label/predcition keys to stack. In this case, the
metric is not expected to work with multi-outputs. The metric only supports
multi outputs if the output of model is already pre-stacked in the expected
format, i.e. ['xmin', 'ymin', 'xmax', 'ymax', 'class_id'] for labels and
['xmin', 'ymin', 'xmax', 'ymax', 'class_id', 'confidence scores'] for
predictions.
Args:
iou_thresholds: (Optional) Threholds for a detection and ground truth pair
with specific iou to be considered as a match.
class_id: (Optional) The class ids for calculating metrics.
class_weight: (Optional) The weight associated with the object class ids.
If it is provided, it should have the same length as class_ids.
area_range: (Optional) The area-range for objects to be considered for
metrics.
max_num_detections: (Optional) The maximum number of detections for a
single image.
name: (Optional) Metric name.
labels_to_stack: (Optional) Keys for columns to be stacked as a single
numpy array as the labels. It is searched under the key labels, features
and transformed features. The desired format is [left bounadary, top
boudnary, right boundary, bottom boundary, class id]. e.g. ['xmin',
'ymin', 'xmax', 'ymax', 'class_id']
predictions_to_stack: (Optional) Output names for columns to be stacked as
a single numpy array as the prediction. It should be the model's output
names. The desired format is [left bounadary, top boudnary, right
boundary, bottom boundary, class id, confidence score]. e.g. ['xmin',
'ymin', 'xmax', 'ymax', 'class_id', 'scores']
num_detections_key: (Optional) An output name in which to find the number
of detections to use for evaluation for a given example. It does nothing
if predictions_to_stack is not set. The value for this output should be
a scalar value or a single-value tensor. The stacked predicitions will
be truncated with the specified number of detections.
allow_missing_key: (Optional) If true, the preprocessor will return empty
array instead of raising errors.
"""
super().__init__(
metric_util.merge_per_key_computations(self._metric_computations),
iou_thresholds=iou_thresholds,
class_id=class_id,
class_weight=class_weight,
area_range=area_range,
max_num_detections=max_num_detections,
name=name,
labels_to_stack=labels_to_stack,
predictions_to_stack=predictions_to_stack,
num_detections_key=num_detections_key,
allow_missing_key=allow_missing_key)
def _default_name(self) -> str:
return AVERAGE_RECALL_NAME
def _metric_computations(
self,
iou_thresholds: Optional[Union[float, List[float]]] = None,
class_id: Optional[int] = None,
class_weight: Optional[float] = None,
max_num_detections: Optional[int] = None,
area_range: Optional[Tuple[float, float]] = None,
name: Optional[str] = None,
model_name: str = '',
output_name: str = '',
example_weighted: bool = False,
labels_to_stack: Optional[List[str]] = None,
predictions_to_stack: Optional[List[str]] = None,
num_detections_key: Optional[str] = None,
allow_missing_key: bool = False,
) -> metric_types.MetricComputations:
"""Returns computations for confusion matrix metric."""
metric_util.validate_object_detection_arguments(
class_id=class_id,
class_weight=class_weight,
area_range=area_range,
max_num_detections=max_num_detections,
labels_to_stack=labels_to_stack,
predictions_to_stack=predictions_to_stack,
output_name=output_name)
# set default value according to COCO metrics
if iou_thresholds is None:
iou_thresholds = np.linspace(0.5, 0.95, 10)
if class_weight is None:
class_weight = 1.0
key = metric_types.MetricKey(
name=name,
model_name=model_name,
output_name=output_name,
sub_key=None,
example_weighted=example_weighted,
aggregation_type=None)
computations = []
recalls_keys = []
for iou_threshold in iou_thresholds:
max_recall_name = metric_util.generate_private_name_from_arguments(
confusion_matrix_metrics.MAX_RECALL_NAME,
iou_threshold=iou_threshold,
class_id=class_id,
class_weight=class_weight,
area_range=area_range,
max_num_detections=max_num_detections,
allow_missing_key=allow_missing_key)
mr = object_detection_confusion_matrix_metrics.ObjectDetectionMaxRecall(
iou_threshold=iou_threshold,
class_id=class_id,
class_weight=class_weight,
area_range=area_range,
max_num_detections=max_num_detections,
name=max_recall_name,
labels_to_stack=labels_to_stack,
predictions_to_stack=predictions_to_stack,
num_detections_key=num_detections_key,
allow_missing_key=allow_missing_key)
computations.extend(
mr.computations(model_names=[model_name], output_names=[output_name]))
recalls_keys.append(computations[-1].keys[-1])
def result(
metrics: Dict[metric_types.MetricKey, Any]
) -> Dict[metric_types.MetricKey, Union[float, np.ndarray]]:
for recalls_key in recalls_keys:
if math.isnan(metrics[recalls_key]):
logging.warning(
'Recall with metric key %s is NaN, it will be'
' ignored in the following calculation.', recalls_key)
recalls = [metrics[recalls_key] for recalls_key in recalls_keys]
value = np.nanmean(recalls)
return {key: value}
derived_computation = metric_types.DerivedMetricComputation(
keys=[key], result=result)
computations.append(derived_computation)
return computations
metric_types.register_metric(COCOAverageRecall)
class COCOMeanAverageRecall(metric_types.Metric):
"""Mean Average recall metric for object detection.
It computes the mean average precision metric for object detections for a
single class. It averages COCOAverageRecall metric over mulitple classes.
"""
def __init__(self,
iou_thresholds: Optional[List[float]] = None,
class_ids: Optional[List[int]] = None,
class_weights: Optional[List[float]] = None,
area_range: Optional[Tuple[float, float]] = None,
max_num_detections: Optional[int] = None,
name: Optional[str] = None,
labels_to_stack: Optional[List[str]] = None,
predictions_to_stack: Optional[List[str]] = None,
num_detections_key: Optional[str] = None,
allow_missing_key: bool = False):
"""Initializes average recall metric.
This metric is only used in object-detection setting. It does not support
sub_key parameters due to the matching algorithm of bounding boxes.
The metric supports using multiple outputs to form the labels/predictions if
the user specifies the label/predcition keys to stack. In this case, the
metric is not expected to work with multi-outputs. The metric only supports
multi outputs if the output of model is already pre-stacked in the expected
format, i.e. ['xmin', 'ymin', 'xmax', 'ymax', 'class_id'] for labels and
['xmin', 'ymin', 'xmax', 'ymax', 'class_id', 'confidence scores'] for
predictions.
Args:
iou_thresholds: (Optional) Threholds for a detection and ground truth pair
with specific iou to be considered as a match.
class_ids: (Optional) The class ids for calculating metrics.
class_weights: (Optional) The weight associated with the object class ids.
If it is provided, it should have the same length as class_ids.
area_range: (Optional) The area-range for objects to be considered for
metrics.
max_num_detections: (Optional) The maximum number of detections for a
single image.
name: (Optional) Metric name.
labels_to_stack: (Optional) Keys for columns to be stacked as a single
numpy array as the labels. It is searched under the key labels, features
and transformed features. The desired format is [left bounadary, top
boudnary, right boundary, bottom boundary, class id]. e.g. ['xmin',
'ymin', 'xmax', 'ymax', 'class_id']
predictions_to_stack: (Optional) Output names for columns to be stacked as
a single numpy array as the prediction. It should be the model's output
names. The desired format is [left bounadary, top boudnary, right
boundary, bottom boundary, class id, confidence score]. e.g. ['xmin',
'ymin', 'xmax', 'ymax', 'class_id', 'scores']
num_detections_key: (Optional) An output name in which to find the number
of detections to use for evaluation for a given example. It does nothing
if predictions_to_stack is not set. The value for this output should be
a scalar value or a single-value tensor. The stacked predicitions will
be truncated with the specified number of detections.
allow_missing_key: (Optional) If true, the preprocessor will return empty
array instead of raising errors.
"""
super().__init__(
metric_util.merge_per_key_computations(self._metric_computations),
iou_thresholds=iou_thresholds,
class_ids=class_ids,
class_weights=class_weights,
area_range=area_range,
max_num_detections=max_num_detections,
name=name,
labels_to_stack=labels_to_stack,
predictions_to_stack=predictions_to_stack,
num_detections_key=num_detections_key,
allow_missing_key=allow_missing_key)
def _default_name(self) -> str:
return MEAN_AVERAGE_RECALL_NAME
def _metric_computations(
self,
iou_thresholds: Optional[List[float]] = None,
class_ids: Optional[Union[int, List[int]]] = None,
class_weights: Optional[Union[float, List[float]]] = None,
max_num_detections: Optional[int] = None,
area_range: Optional[Tuple[float, float]] = None,
name: Optional[str] = None,
model_name: str = '',
output_name: str = '',
example_weighted: bool = False,
labels_to_stack: Optional[List[str]] = None,
predictions_to_stack: Optional[List[str]] = None,
num_detections_key: Optional[str] = None,
allow_missing_key: bool = False,
) -> metric_types.MetricComputations:
"""Returns computations for confusion matrix metric."""
metric_util.validate_object_detection_arguments(
class_id=class_ids,
class_weight=class_weights,
area_range=area_range,
max_num_detections=max_num_detections,
labels_to_stack=labels_to_stack,
predictions_to_stack=predictions_to_stack,
output_name=output_name)
if class_weights is None:
class_weights = [1.0] * len(class_ids)
key = metric_types.MetricKey(
name=name,
model_name=model_name,
output_name=output_name,
sub_key=None,
example_weighted=example_weighted,
aggregation_type=None)
computations = []
recalls_keys = []
for class_id, class_weight in zip(class_ids, class_weights):
max_recall_name = metric_util.generate_private_name_from_arguments(
AVERAGE_RECALL_NAME,
iou_thresholds=iou_thresholds,
class_id=class_id,
class_weight=class_weight,
area_range=area_range,
max_num_detections=max_num_detections,
allow_missing_key=allow_missing_key)
mr = COCOAverageRecall(
iou_thresholds=iou_thresholds,
class_id=class_id,
class_weight=class_weight,
area_range=area_range,
max_num_detections=max_num_detections,
name=max_recall_name,
labels_to_stack=labels_to_stack,
predictions_to_stack=predictions_to_stack,
num_detections_key=num_detections_key,
allow_missing_key=allow_missing_key)
computations.extend(
mr.computations(model_names=[model_name], output_names=[output_name]))
recalls_keys.append(computations[-1].keys[-1])
def result(
metrics: Dict[metric_types.MetricKey, Any]
) -> Dict[metric_types.MetricKey, Union[float, np.ndarray]]:
recalls = [metrics[recalls_key] for recalls_key in recalls_keys]
value = np.nanmean(recalls)
return {key: value}
derived_computation = metric_types.DerivedMetricComputation(
keys=[key], result=result)
computations.append(derived_computation)
return computations
metric_types.register_metric(COCOMeanAverageRecall)