Register
Login
Resources
Docs Blog Datasets Glossary Case Studies Tutorials & Webinars
Product
Data Engine LLMs Platform Enterprise
Pricing Explore
Connect to our Discord channel
Resources  
Docs Blog Datasets Glossary Case Studies Tutorials & Webinars
Explore Pricing
Product  
Data Engine LLMs Platform Enterprise
Connect to Dagshub Discord channel
Register
Login

Deci-AI
/
super-gradients
connected to https://github.com/Deci-AI/super-gradients.git

Watch 4
Issues 62 Pull Requests 44 Discussions 0

#378 Feature/sg 281 add kd notebook

Merged
Ghost merged 1 commits into Deci-AI:master from deci-ai:feature/SG-281-add_kd_notebook
Conversation 0 Files changed 1 Commits 1
File Compare
super-gradients
/
docs
/
_modules
/
super_gradients
/
training
/
utils
/
detection_utils.html
View left file View right file
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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202 
  1. <!DOCTYPE html>
    2. 

  2. <html class="writer-html5" lang="en" >
    3. 

  3. <head>
    4. 

  4.   <meta charset="utf-8" />
    5. 

  5.   <meta name="viewport" content="width=device-width, initial-scale=1.0" />
    6. 

  6.   <title>super_gradients.training.utils.detection_utils &mdash; SuperGradients 1.0 documentation</title>
    7. 

  7.       <link rel="stylesheet" href="../../../../_static/pygments.css" type="text/css" />
    8. 

  8.       <link rel="stylesheet" href="../../../../_static/css/theme.css" type="text/css" />
    9. 

  9.       <link rel="stylesheet" href="../../../../_static/graphviz.css" type="text/css" />
    10. 

  10.   <!--[if lt IE 9]>
    11. 

  11.     <script src="../../../../_static/js/html5shiv.min.js"></script>
    12. 

  12.   <![endif]-->
    13. 

  13.   
    14. 

  14.         <script data-url_root="../../../../" id="documentation_options" src="../../../../_static/documentation_options.js"></script>
    15. 

  15.         <script src="../../../../_static/jquery.js"></script>
    16. 

  16.         <script src="../../../../_static/underscore.js"></script>
    17. 

  17.         <script src="../../../../_static/doctools.js"></script>
    18. 

  18.     <script src="../../../../_static/js/theme.js"></script>
    19. 

  19.     <link rel="index" title="Index" href="../../../../genindex.html" />
    20. 

  20.     <link rel="search" title="Search" href="../../../../search.html" /> 
    21. 

  21. </head>
    22. 

  22. 

  23. <body class="wy-body-for-nav"> 
    24. 

  24.   <div class="wy-grid-for-nav">
    25. 

  25.     <nav data-toggle="wy-nav-shift" class="wy-nav-side">
    26. 

  26.       <div class="wy-side-scroll">
    27. 

  27.         <div class="wy-side-nav-search" >
    28. 

  28.             <a href="../../../../index.html" class="icon icon-home"> SuperGradients
    29. 

  29.           </a>
    30. 

  30. <div role="search">
    31. 

  31.   <form id="rtd-search-form" class="wy-form" action="../../../../search.html" method="get">
    32. 

  32.     <input type="text" name="q" placeholder="Search docs" />
    33. 

  33.     <input type="hidden" name="check_keywords" value="yes" />
    34. 

  34.     <input type="hidden" name="area" value="default" />
    35. 

  35.   </form>
    36. 

  36. </div>
    37. 

  37.         </div><div class="wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="Navigation menu">
    38. 

  38.               <p class="caption"><span class="caption-text">Welcome To SuperGradients</span></p>
    39. 

  39. <ul>
    40. 

  40. <li class="toctree-l1"><a class="reference internal" href="../../../../welcome.html">Fill our 4-question quick survey! We will raffle free SuperGradients swag between those who will participate -&gt; Fill Survey</a></li>
    41. 

  41. <li class="toctree-l1"><a class="reference internal" href="../../../../welcome.html#supergradients">SuperGradients</a></li>
    42. 

  42. </ul>
    43. 

  43. <p class="caption"><span class="caption-text">Technical Documentation</span></p>
    44. 

  44. <ul>
    45. 

  45. <li class="toctree-l1"><a class="reference internal" href="../../../../super_gradients.common.html">Common package</a></li>
    46. 

  46. <li class="toctree-l1"><a class="reference internal" href="../../../../super_gradients.training.html">Training package</a></li>
    47. 

  47. </ul>
    48. 

  48. <p class="caption"><span class="caption-text">User Guide</span></p>
    49. 

  49. <ul>
    50. 

  50. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html">What is SuperGradients?</a></li>
    51. 

  51. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#introducing-the-supergradients-library">Introducing the SuperGradients library</a></li>
    52. 

  52. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#installation">Installation</a></li>
    53. 

  53. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#integrating-your-training-code-complete-walkthrough">Integrating Your Training Code - Complete Walkthrough</a></li>
    54. 

  54. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#training-parameters">Training Parameters</a></li>
    55. 

  55. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#logs-and-checkpoints">Logs and Checkpoints</a></li>
    56. 

  56. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#dataset-parameters">Dataset Parameters</a></li>
    57. 

  57. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#network-architectures">Network Architectures</a></li>
    58. 

  58. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#pretrained-models">Pretrained Models</a></li>
    59. 

  59. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#how-to-reproduce-our-training-recipes">How To Reproduce Our Training Recipes</a></li>
    60. 

  60. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#professional-tools-integration">Professional Tools Integration</a></li>
    61. 

  61. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#supergradients-faq">SuperGradients FAQ</a></li>
    62. 

  62. </ul>
    63. 

  63. 

  64.         </div>
    65. 

  65.       </div>
    66. 

  66.     </nav>
    67. 

  67. 

  68.     <section data-toggle="wy-nav-shift" class="wy-nav-content-wrap"><nav class="wy-nav-top" aria-label="Mobile navigation menu" >
    69. 

  69.           <i data-toggle="wy-nav-top" class="fa fa-bars"></i>
    70. 

  70.           <a href="../../../../index.html">SuperGradients</a>
    71. 

  71.       </nav>
    72. 

  72. 

  73.       <div class="wy-nav-content">
    74. 

  74.         <div class="rst-content">
    75. 

  75.           <div role="navigation" aria-label="Page navigation">
    76. 

  76.   <ul class="wy-breadcrumbs">
    77. 

  77.       <li><a href="../../../../index.html" class="icon icon-home"></a> &raquo;</li>
    78. 

  78.           <li><a href="../../../index.html">Module code</a> &raquo;</li>
    79. 

  79.       <li>super_gradients.training.utils.detection_utils</li>
    80. 

  80.       <li class="wy-breadcrumbs-aside">
    81. 

  81.       </li>
    82. 

  82.   </ul>
    83. 

  83.   <hr/>
    84. 

  84. </div>
    85. 

  85.           <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
    86. 

  86.            <div itemprop="articleBody">
    87. 

  87.              
    88. 

  88.   <h1>Source code for super_gradients.training.utils.detection_utils</h1><div class="highlight"><pre>
    89. 

  89. <span></span><span class="kn">import</span> <span class="nn">math</span>
    90. 

  90. <span class="kn">import</span> <span class="nn">os</span>
    91. 

  91. <span class="kn">import</span> <span class="nn">pathlib</span>
    92. 

  92. <span class="kn">from</span> <span class="nn">abc</span> <span class="kn">import</span> <span class="n">ABC</span><span class="p">,</span> <span class="n">abstractmethod</span>
    93. 

  93. <span class="kn">from</span> <span class="nn">enum</span> <span class="kn">import</span> <span class="n">Enum</span>
    94. 

  94. <span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">Callable</span><span class="p">,</span> <span class="n">List</span><span class="p">,</span> <span class="n">Union</span><span class="p">,</span> <span class="n">Tuple</span><span class="p">,</span> <span class="n">Optional</span><span class="p">,</span> <span class="n">Dict</span>
    95. 

  95. 

  96. <span class="kn">import</span> <span class="nn">cv2</span>
    97. 

  97. <span class="kn">import</span> <span class="nn">matplotlib.pyplot</span> <span class="k">as</span> <span class="nn">plt</span>
    98. 

  98. 

  99. <span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
    100. 

  100. <span class="kn">import</span> <span class="nn">torch</span>
    101. 

  101. <span class="kn">import</span> <span class="nn">torchvision</span>
    102. 

  102. <span class="kn">from</span> <span class="nn">torch</span> <span class="kn">import</span> <span class="n">nn</span>
    103. 

  103. <span class="kn">from</span> <span class="nn">torch.utils.data._utils.collate</span> <span class="kn">import</span> <span class="n">default_collate</span>
    104. 

  104. <span class="kn">from</span> <span class="nn">omegaconf</span> <span class="kn">import</span> <span class="n">ListConfig</span>
    105. 

  105. 

  106. 

  107. <div class="viewcode-block" id="DetectionTargetsFormat"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.DetectionTargetsFormat">[docs]</a><span class="k">class</span> <span class="nc">DetectionTargetsFormat</span><span class="p">(</span><span class="n">Enum</span><span class="p">):</span>
    108. 

  108.     <span class="sd">&quot;&quot;&quot;</span>
    109. 

  109. <span class="sd">    Enum class for the different detection output formats</span>
    110. 

  110. 

  111. <span class="sd">    When NORMALIZED is not specified- the type refers to unnormalized image coordinates (of the bboxes).</span>
    112. 

  112. 

  113. <span class="sd">    For example:</span>
    114. 

  114. <span class="sd">    LABEL_NORMALIZED_XYXY means [class_idx,x1,y1,x2,y2]</span>
    115. 

  115. <span class="sd">    &quot;&quot;&quot;</span>
    116. 

  116.     <span class="n">LABEL_XYXY</span> <span class="o">=</span> <span class="s2">&quot;LABEL_XYXY&quot;</span>
    117. 

  117.     <span class="n">XYXY_LABEL</span> <span class="o">=</span> <span class="s2">&quot;XYXY_LABEL&quot;</span>
    118. 

  118.     <span class="n">LABEL_NORMALIZED_XYXY</span> <span class="o">=</span> <span class="s2">&quot;LABEL_NORMALIZED_XYXY&quot;</span>
    119. 

  119.     <span class="n">NORMALIZED_XYXY_LABEL</span> <span class="o">=</span> <span class="s2">&quot;NORMALIZED_XYXY_LABEL&quot;</span>
    120. 

  120.     <span class="n">LABEL_CXCYWH</span> <span class="o">=</span> <span class="s2">&quot;LABEL_CXCYWH&quot;</span>
    121. 

  121.     <span class="n">CXCYWH_LABEL</span> <span class="o">=</span> <span class="s2">&quot;CXCYWH_LABEL&quot;</span>
    122. 

  122.     <span class="n">LABEL_NORMALIZED_CXCYWH</span> <span class="o">=</span> <span class="s2">&quot;LABEL_NORMALIZED_CXCYWH&quot;</span>
    123. 

  123.     <span class="n">NORMALIZED_CXCYWH_LABEL</span> <span class="o">=</span> <span class="s2">&quot;NORMALIZED_CXCYWH_LABEL&quot;</span></div>
    124. 

  124. 

  125. 

  126. <div class="viewcode-block" id="get_cls_posx_in_target"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.get_cls_posx_in_target">[docs]</a><span class="k">def</span> <span class="nf">get_cls_posx_in_target</span><span class="p">(</span><span class="n">target_format</span><span class="p">:</span> <span class="n">DetectionTargetsFormat</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="nb">int</span><span class="p">:</span>
    127. 

  127.     <span class="sd">&quot;&quot;&quot;Get the label of a given target</span>
    128. 

  128. <span class="sd">    :param target_format:   Representation of the target (ex: LABEL_XYXY)</span>
    129. 

  129. <span class="sd">    :return:                Position of the class id in a bbox</span>
    130. 

  130. <span class="sd">                                ex: 0 if bbox of format label_xyxy | -1 if bbox of format xyxy_label</span>
    131. 

  131. <span class="sd">    &quot;&quot;&quot;</span>
    132. 

  132.     <span class="n">format_split</span> <span class="o">=</span> <span class="n">target_format</span><span class="o">.</span><span class="n">value</span><span class="o">.</span><span class="n">split</span><span class="p">(</span><span class="s2">&quot;_&quot;</span><span class="p">)</span>
    133. 

  133.     <span class="k">if</span> <span class="n">format_split</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="o">==</span> <span class="s2">&quot;LABEL&quot;</span><span class="p">:</span>
    134. 

  134.         <span class="k">return</span> <span class="mi">0</span>
    135. 

  135.     <span class="k">elif</span> <span class="n">format_split</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span> <span class="o">==</span> <span class="s2">&quot;LABEL&quot;</span><span class="p">:</span>
    136. 

  136.         <span class="k">return</span> <span class="o">-</span><span class="mi">1</span>
    137. 

  137.     <span class="k">else</span><span class="p">:</span>
    138. 

  138.         <span class="k">raise</span> <span class="ne">NotImplementedError</span><span class="p">(</span><span class="sa">f</span><span class="s2">&quot;No implementation to find index of LABEL in </span><span class="si">{</span><span class="n">target_format</span><span class="o">.</span><span class="n">value</span><span class="si">}</span><span class="s2">&quot;</span><span class="p">)</span></div>
    139. 

  139. 

  140. 

  141. <span class="k">def</span> <span class="nf">_set_batch_labels_index</span><span class="p">(</span><span class="n">labels_batch</span><span class="p">):</span>
    142. 

  142.     <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">labels</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="n">labels_batch</span><span class="p">):</span>
    143. 

  143.         <span class="n">labels</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">=</span> <span class="n">i</span>
    144. 

  144.     <span class="k">return</span> <span class="n">labels_batch</span>
    145. 

  145. 

  146. 

  147. <div class="viewcode-block" id="convert_xywh_bbox_to_xyxy"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.convert_xywh_bbox_to_xyxy">[docs]</a><span class="k">def</span> <span class="nf">convert_xywh_bbox_to_xyxy</span><span class="p">(</span><span class="n">input_bbox</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span>
    148. 

  148.     <span class="sd">&quot;&quot;&quot;</span>
    149. 

  149. <span class="sd">    Converts bounding box format from [x, y, w, h] to [x1, y1, x2, y2]</span>
    150. 

  150. <span class="sd">        :param input_bbox:  input bbox either 2-dimensional (for all boxes of a single image) or 3-dimensional (for</span>
    151. 

  151. <span class="sd">                            boxes of a batch of images)</span>
    152. 

  152. <span class="sd">        :return:            Converted bbox in same dimensions as the original</span>
    153. 

  153. <span class="sd">    &quot;&quot;&quot;</span>
    154. 

  154.     <span class="n">need_squeeze</span> <span class="o">=</span> <span class="kc">False</span>
    155. 

  155.     <span class="c1"># the input is always processed as a batch. in case it not a batch, it is unsqueezed, process and than squeeze back.</span>
    156. 

  156.     <span class="k">if</span> <span class="n">input_bbox</span><span class="o">.</span><span class="n">dim</span><span class="p">()</span> <span class="o">&lt;</span> <span class="mi">3</span><span class="p">:</span>
    157. 

  157.         <span class="n">need_squeeze</span> <span class="o">=</span> <span class="kc">True</span>
    158. 

  158.         <span class="n">input_bbox</span> <span class="o">=</span> <span class="n">input_bbox</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span>
    159. 

  159. 

  160.     <span class="n">converted_bbox</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros_like</span><span class="p">(</span><span class="n">input_bbox</span><span class="p">)</span> <span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">input_bbox</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">)</span> <span class="k">else</span> <span class="n">np</span><span class="o">.</span><span class="n">zeros_like</span><span class="p">(</span><span class="n">input_bbox</span><span class="p">)</span>
    161. 

  161.     <span class="n">converted_bbox</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">=</span> <span class="n">input_bbox</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">-</span> <span class="n">input_bbox</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">2</span><span class="p">]</span> <span class="o">/</span> <span class="mi">2</span>
    162. 

  162.     <span class="n">converted_bbox</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">1</span><span class="p">]</span> <span class="o">=</span> <span class="n">input_bbox</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">1</span><span class="p">]</span> <span class="o">-</span> <span class="n">input_bbox</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">3</span><span class="p">]</span> <span class="o">/</span> <span class="mi">2</span>
    163. 

  163.     <span class="n">converted_bbox</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">2</span><span class="p">]</span> <span class="o">=</span> <span class="n">input_bbox</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">+</span> <span class="n">input_bbox</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">2</span><span class="p">]</span> <span class="o">/</span> <span class="mi">2</span>
    164. 

  164.     <span class="n">converted_bbox</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">3</span><span class="p">]</span> <span class="o">=</span> <span class="n">input_bbox</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">1</span><span class="p">]</span> <span class="o">+</span> <span class="n">input_bbox</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">3</span><span class="p">]</span> <span class="o">/</span> <span class="mi">2</span>
    165. 

  165. 

  166.     <span class="c1"># squeeze back if needed</span>
    167. 

  167.     <span class="k">if</span> <span class="n">need_squeeze</span><span class="p">:</span>
    168. 

  168.         <span class="n">converted_bbox</span> <span class="o">=</span> <span class="n">converted_bbox</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>
    169. 

  169. 

  170.     <span class="k">return</span> <span class="n">converted_bbox</span></div>
    171. 

  171. 

  172. 

  173. <span class="k">def</span> <span class="nf">_iou</span><span class="p">(</span><span class="n">CIoU</span><span class="p">:</span> <span class="nb">bool</span><span class="p">,</span> <span class="n">DIoU</span><span class="p">:</span> <span class="nb">bool</span><span class="p">,</span> <span class="n">GIoU</span><span class="p">:</span> <span class="nb">bool</span><span class="p">,</span> <span class="n">b1_x1</span><span class="p">,</span> <span class="n">b1_x2</span><span class="p">,</span> <span class="n">b1_y1</span><span class="p">,</span> <span class="n">b1_y2</span><span class="p">,</span> <span class="n">b2_x1</span><span class="p">,</span> <span class="n">b2_x2</span><span class="p">,</span> <span class="n">b2_y1</span><span class="p">,</span> <span class="n">b2_y2</span><span class="p">,</span> <span class="n">eps</span><span class="p">):</span>
    174. 

  174.     <span class="sd">&quot;&quot;&quot;</span>
    175. 

  175. <span class="sd">    Internal function for the use of calculate_bbox_iou_matrix and calculate_bbox_iou_elementwise functions</span>
    176. 

  176. <span class="sd">    DO NOT CALL THIS FUNCTIONS DIRECTLY - use one of the functions mentioned above</span>
    177. 

  177. <span class="sd">    &quot;&quot;&quot;</span>
    178. 

  178.     <span class="c1"># Intersection area</span>
    179. 

  179.     <span class="n">intersection_area</span> <span class="o">=</span> <span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">min</span><span class="p">(</span><span class="n">b1_x2</span><span class="p">,</span> <span class="n">b2_x2</span><span class="p">)</span> <span class="o">-</span> <span class="n">torch</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="n">b1_x1</span><span class="p">,</span> <span class="n">b2_x1</span><span class="p">))</span><span class="o">.</span><span class="n">clamp</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span> <span class="o">*</span> \
    180. 

  180.                         <span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">min</span><span class="p">(</span><span class="n">b1_y2</span><span class="p">,</span> <span class="n">b2_y2</span><span class="p">)</span> <span class="o">-</span> <span class="n">torch</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="n">b1_y1</span><span class="p">,</span> <span class="n">b2_y1</span><span class="p">))</span><span class="o">.</span><span class="n">clamp</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span>
    181. 

  181.     <span class="c1"># Union Area</span>
    182. 

  182.     <span class="n">w1</span><span class="p">,</span> <span class="n">h1</span> <span class="o">=</span> <span class="n">b1_x2</span> <span class="o">-</span> <span class="n">b1_x1</span><span class="p">,</span> <span class="n">b1_y2</span> <span class="o">-</span> <span class="n">b1_y1</span>
    183. 

  183.     <span class="n">w2</span><span class="p">,</span> <span class="n">h2</span> <span class="o">=</span> <span class="n">b2_x2</span> <span class="o">-</span> <span class="n">b2_x1</span><span class="p">,</span> <span class="n">b2_y2</span> <span class="o">-</span> <span class="n">b2_y1</span>
    184. 

  184.     <span class="n">union_area</span> <span class="o">=</span> <span class="n">w1</span> <span class="o">*</span> <span class="n">h1</span> <span class="o">+</span> <span class="n">w2</span> <span class="o">*</span> <span class="n">h2</span> <span class="o">-</span> <span class="n">intersection_area</span> <span class="o">+</span> <span class="n">eps</span>
    185. 

  185.     <span class="n">iou</span> <span class="o">=</span> <span class="n">intersection_area</span> <span class="o">/</span> <span class="n">union_area</span>  <span class="c1"># iou</span>
    186. 

  186.     <span class="k">if</span> <span class="n">GIoU</span> <span class="ow">or</span> <span class="n">DIoU</span> <span class="ow">or</span> <span class="n">CIoU</span><span class="p">:</span>
    187. 

  187.         <span class="n">cw</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="n">b1_x2</span><span class="p">,</span> <span class="n">b2_x2</span><span class="p">)</span> <span class="o">-</span> <span class="n">torch</span><span class="o">.</span><span class="n">min</span><span class="p">(</span><span class="n">b1_x1</span><span class="p">,</span> <span class="n">b2_x1</span><span class="p">)</span>  <span class="c1"># convex (smallest enclosing box) width</span>
    188. 

  188.         <span class="n">ch</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="n">b1_y2</span><span class="p">,</span> <span class="n">b2_y2</span><span class="p">)</span> <span class="o">-</span> <span class="n">torch</span><span class="o">.</span><span class="n">min</span><span class="p">(</span><span class="n">b1_y1</span><span class="p">,</span> <span class="n">b2_y1</span><span class="p">)</span>  <span class="c1"># convex height</span>
    189. 

  189.         <span class="c1"># Generalized IoU https://arxiv.org/pdf/1902.09630.pdf</span>
    190. 

  190.         <span class="k">if</span> <span class="n">GIoU</span><span class="p">:</span>
    191. 

  191.             <span class="n">c_area</span> <span class="o">=</span> <span class="n">cw</span> <span class="o">*</span> <span class="n">ch</span> <span class="o">+</span> <span class="n">eps</span>  <span class="c1"># convex area</span>
    192. 

  192.             <span class="n">iou</span> <span class="o">-=</span> <span class="p">(</span><span class="n">c_area</span> <span class="o">-</span> <span class="n">union_area</span><span class="p">)</span> <span class="o">/</span> <span class="n">c_area</span>  <span class="c1"># GIoU</span>
    193. 

  193.         <span class="c1"># Distance or Complete IoU https://arxiv.org/abs/1911.08287v1</span>
    194. 

  194.         <span class="k">if</span> <span class="n">DIoU</span> <span class="ow">or</span> <span class="n">CIoU</span><span class="p">:</span>
    195. 

  195.             <span class="c1"># convex diagonal squared</span>
    196. 

  196.             <span class="n">c2</span> <span class="o">=</span> <span class="n">cw</span> <span class="o">**</span> <span class="mi">2</span> <span class="o">+</span> <span class="n">ch</span> <span class="o">**</span> <span class="mi">2</span> <span class="o">+</span> <span class="n">eps</span>
    197. 

  197.             <span class="c1"># centerpoint distance squared</span>
    198. 

  198.             <span class="n">rho2</span> <span class="o">=</span> <span class="p">((</span><span class="n">b2_x1</span> <span class="o">+</span> <span class="n">b2_x2</span> <span class="o">-</span> <span class="n">b1_x1</span> <span class="o">-</span> <span class="n">b1_x2</span><span class="p">)</span> <span class="o">**</span> <span class="mi">2</span> <span class="o">+</span> <span class="p">(</span><span class="n">b2_y1</span> <span class="o">+</span> <span class="n">b2_y2</span> <span class="o">-</span> <span class="n">b1_y1</span> <span class="o">-</span> <span class="n">b1_y2</span><span class="p">)</span> <span class="o">**</span> <span class="mi">2</span><span class="p">)</span> <span class="o">/</span> <span class="mi">4</span>
    199. 

  199.             <span class="k">if</span> <span class="n">DIoU</span><span class="p">:</span>
    200. 

  200.                 <span class="n">iou</span> <span class="o">-=</span> <span class="n">rho2</span> <span class="o">/</span> <span class="n">c2</span>  <span class="c1"># DIoU</span>
    201. 

  201.             <span class="k">elif</span> <span class="n">CIoU</span><span class="p">:</span>  <span class="c1"># https://github.com/Zzh-tju/DIoU-SSD-pytorch/blob/master/utils/box/box_utils.py#L47</span>
    202. 

  202.                 <span class="n">v</span> <span class="o">=</span> <span class="p">(</span><span class="mi">4</span> <span class="o">/</span> <span class="n">math</span><span class="o">.</span><span class="n">pi</span> <span class="o">**</span> <span class="mi">2</span><span class="p">)</span> <span class="o">*</span> <span class="n">torch</span><span class="o">.</span><span class="n">pow</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">atan</span><span class="p">(</span><span class="n">w2</span> <span class="o">/</span> <span class="n">h2</span><span class="p">)</span> <span class="o">-</span> <span class="n">torch</span><span class="o">.</span><span class="n">atan</span><span class="p">(</span><span class="n">w1</span> <span class="o">/</span> <span class="n">h1</span><span class="p">),</span> <span class="mi">2</span><span class="p">)</span>
    203. 

  203.                 <span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">no_grad</span><span class="p">():</span>
    204. 

  204.                     <span class="n">alpha</span> <span class="o">=</span> <span class="n">v</span> <span class="o">/</span> <span class="p">((</span><span class="mi">1</span> <span class="o">+</span> <span class="n">eps</span><span class="p">)</span> <span class="o">-</span> <span class="n">iou</span> <span class="o">+</span> <span class="n">v</span><span class="p">)</span>
    205. 

  205.                 <span class="n">iou</span> <span class="o">-=</span> <span class="p">(</span><span class="n">rho2</span> <span class="o">/</span> <span class="n">c2</span> <span class="o">+</span> <span class="n">v</span> <span class="o">*</span> <span class="n">alpha</span><span class="p">)</span>  <span class="c1"># CIoU</span>
    206. 

  206.     <span class="k">return</span> <span class="n">iou</span>
    207. 

  207. 

  208. 

  209. <div class="viewcode-block" id="calculate_bbox_iou_matrix"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.calculate_bbox_iou_matrix">[docs]</a><span class="k">def</span> <span class="nf">calculate_bbox_iou_matrix</span><span class="p">(</span><span class="n">box1</span><span class="p">,</span> <span class="n">box2</span><span class="p">,</span> <span class="n">x1y1x2y2</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="n">GIoU</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">False</span><span class="p">,</span> <span class="n">DIoU</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span> <span class="n">CIoU</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span> <span class="n">eps</span><span class="o">=</span><span class="mf">1e-9</span><span class="p">):</span>
    210. 

  210.     <span class="sd">&quot;&quot;&quot;</span>
    211. 

  211. <span class="sd">    calculate iou matrix containing the iou of every couple iuo(i,j) where i is in box1 and j is in box2</span>
    212. 

  212. <span class="sd">        :param box1: a 2D tensor of boxes (shape N x 4)</span>
    213. 

  213. <span class="sd">        :param box2: a 2D tensor of boxes (shape M x 4)</span>
    214. 

  214. <span class="sd">        :param x1y1x2y2: boxes format is x1y1x2y2 (True) or xywh where xy is the center (False)</span>
    215. 

  215. <span class="sd">        :return: a 2D iou matrix (shape NxM)</span>
    216. 

  216. <span class="sd">    &quot;&quot;&quot;</span>
    217. 

  217.     <span class="k">if</span> <span class="n">box1</span><span class="o">.</span><span class="n">dim</span><span class="p">()</span> <span class="o">&gt;</span> <span class="mi">1</span><span class="p">:</span>
    218. 

  218.         <span class="n">box1</span> <span class="o">=</span> <span class="n">box1</span><span class="o">.</span><span class="n">T</span>
    219. 

  219. 

  220.     <span class="c1"># Get the coordinates of bounding boxes</span>
    221. 

  221.     <span class="k">if</span> <span class="n">x1y1x2y2</span><span class="p">:</span>  <span class="c1"># x1, y1, x2, y2 = box1</span>
    222. 

  222.         <span class="n">b1_x1</span><span class="p">,</span> <span class="n">b1_y1</span><span class="p">,</span> <span class="n">b1_x2</span><span class="p">,</span> <span class="n">b1_y2</span> <span class="o">=</span> <span class="n">box1</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">box1</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span> <span class="n">box1</span><span class="p">[</span><span class="mi">2</span><span class="p">],</span> <span class="n">box1</span><span class="p">[</span><span class="mi">3</span><span class="p">]</span>
    223. 

  223.         <span class="n">b2_x1</span><span class="p">,</span> <span class="n">b2_y1</span><span class="p">,</span> <span class="n">b2_x2</span><span class="p">,</span> <span class="n">b2_y2</span> <span class="o">=</span> <span class="n">box2</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">],</span> <span class="n">box2</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">],</span> <span class="n">box2</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">],</span> <span class="n">box2</span><span class="p">[:,</span> <span class="mi">3</span><span class="p">]</span>
    224. 

  224.     <span class="k">else</span><span class="p">:</span>  <span class="c1"># x, y, w, h = box1</span>
    225. 

  225.         <span class="n">b1_x1</span><span class="p">,</span> <span class="n">b1_x2</span> <span class="o">=</span> <span class="n">box1</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="o">-</span> <span class="n">box1</span><span class="p">[</span><span class="mi">2</span><span class="p">]</span> <span class="o">/</span> <span class="mi">2</span><span class="p">,</span> <span class="n">box1</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="o">+</span> <span class="n">box1</span><span class="p">[</span><span class="mi">2</span><span class="p">]</span> <span class="o">/</span> <span class="mi">2</span>
    226. 

  226.         <span class="n">b1_y1</span><span class="p">,</span> <span class="n">b1_y2</span> <span class="o">=</span> <span class="n">box1</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span> <span class="o">-</span> <span class="n">box1</span><span class="p">[</span><span class="mi">3</span><span class="p">]</span> <span class="o">/</span> <span class="mi">2</span><span class="p">,</span> <span class="n">box1</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span> <span class="o">+</span> <span class="n">box1</span><span class="p">[</span><span class="mi">3</span><span class="p">]</span> <span class="o">/</span> <span class="mi">2</span>
    227. 

  227.         <span class="n">b2_x1</span><span class="p">,</span> <span class="n">b2_x2</span> <span class="o">=</span> <span class="n">box2</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">-</span> <span class="n">box2</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">]</span> <span class="o">/</span> <span class="mi">2</span><span class="p">,</span> <span class="n">box2</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">+</span> <span class="n">box2</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">]</span> <span class="o">/</span> <span class="mi">2</span>
    228. 

  228.         <span class="n">b2_y1</span><span class="p">,</span> <span class="n">b2_y2</span> <span class="o">=</span> <span class="n">box2</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">]</span> <span class="o">-</span> <span class="n">box2</span><span class="p">[:,</span> <span class="mi">3</span><span class="p">]</span> <span class="o">/</span> <span class="mi">2</span><span class="p">,</span> <span class="n">box2</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">]</span> <span class="o">+</span> <span class="n">box2</span><span class="p">[:,</span> <span class="mi">3</span><span class="p">]</span> <span class="o">/</span> <span class="mi">2</span>
    229. 

  229. 

  230.     <span class="n">b1_x1</span><span class="p">,</span> <span class="n">b1_y1</span><span class="p">,</span> <span class="n">b1_x2</span><span class="p">,</span> <span class="n">b1_y2</span> <span class="o">=</span> <span class="n">b1_x1</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="mi">1</span><span class="p">),</span> <span class="n">b1_y1</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="mi">1</span><span class="p">),</span> <span class="n">b1_x2</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="mi">1</span><span class="p">),</span> <span class="n">b1_y2</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="mi">1</span><span class="p">)</span>
    231. 

  231. 

  232.     <span class="k">return</span> <span class="n">_iou</span><span class="p">(</span><span class="n">CIoU</span><span class="p">,</span> <span class="n">DIoU</span><span class="p">,</span> <span class="n">GIoU</span><span class="p">,</span> <span class="n">b1_x1</span><span class="p">,</span> <span class="n">b1_x2</span><span class="p">,</span> <span class="n">b1_y1</span><span class="p">,</span> <span class="n">b1_y2</span><span class="p">,</span> <span class="n">b2_x1</span><span class="p">,</span> <span class="n">b2_x2</span><span class="p">,</span> <span class="n">b2_y1</span><span class="p">,</span> <span class="n">b2_y2</span><span class="p">,</span> <span class="n">eps</span><span class="p">)</span></div>
    233. 

  233. 

  234. 

  235. <div class="viewcode-block" id="calc_bbox_iou_matrix"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.calc_bbox_iou_matrix">[docs]</a><span class="k">def</span> <span class="nf">calc_bbox_iou_matrix</span><span class="p">(</span><span class="n">pred</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span>
    236. 

  236.     <span class="sd">&quot;&quot;&quot;</span>
    237. 

  237. <span class="sd">    calculate iou for every pair of boxes in the boxes vector</span>
    238. 

  238. <span class="sd">    :param pred: a 3-dimensional tensor containing all boxes for a batch of images [N, num_boxes, 4], where</span>
    239. 

  239. <span class="sd">                 each box format is [x1,y1,x2,y2]</span>
    240. 

  240. <span class="sd">    :return: a 3-dimensional matrix where M_i_j_k is the iou of box j and box k of the i&#39;th image in the batch</span>
    241. 

  241. <span class="sd">    &quot;&quot;&quot;</span>
    242. 

  242.     <span class="n">box</span> <span class="o">=</span> <span class="n">pred</span><span class="p">[:,</span> <span class="p">:,</span> <span class="p">:</span><span class="mi">4</span><span class="p">]</span>  <span class="c1">#</span>
    243. 

  243.     <span class="n">b1_x1</span><span class="p">,</span> <span class="n">b1_y1</span> <span class="o">=</span> <span class="n">box</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">0</span><span class="p">]</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="mi">1</span><span class="p">),</span> <span class="n">box</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">1</span><span class="p">]</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="mi">1</span><span class="p">)</span>
    244. 

  244.     <span class="n">b1_x2</span><span class="p">,</span> <span class="n">b1_y2</span> <span class="o">=</span> <span class="n">box</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">2</span><span class="p">]</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="mi">1</span><span class="p">),</span> <span class="n">box</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">3</span><span class="p">]</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="mi">1</span><span class="p">)</span>
    245. 

  245. 

  246.     <span class="n">b2_x1</span> <span class="o">=</span> <span class="n">b1_x1</span><span class="o">.</span><span class="n">transpose</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
    247. 

  247.     <span class="n">b2_x2</span> <span class="o">=</span> <span class="n">b1_x2</span><span class="o">.</span><span class="n">transpose</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
    248. 

  248.     <span class="n">b2_y1</span> <span class="o">=</span> <span class="n">b1_y1</span><span class="o">.</span><span class="n">transpose</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
    249. 

  249.     <span class="n">b2_y2</span> <span class="o">=</span> <span class="n">b1_y2</span><span class="o">.</span><span class="n">transpose</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
    250. 

  250.     <span class="n">intersection_area</span> <span class="o">=</span> <span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">min</span><span class="p">(</span><span class="n">b1_x2</span><span class="p">,</span> <span class="n">b2_x2</span><span class="p">)</span> <span class="o">-</span> <span class="n">torch</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="n">b1_x1</span><span class="p">,</span> <span class="n">b2_x1</span><span class="p">))</span><span class="o">.</span><span class="n">clamp</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span> <span class="o">*</span> \
    251. 

  251.                         <span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">min</span><span class="p">(</span><span class="n">b1_y2</span><span class="p">,</span> <span class="n">b2_y2</span><span class="p">)</span> <span class="o">-</span> <span class="n">torch</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="n">b1_y1</span><span class="p">,</span> <span class="n">b2_y1</span><span class="p">))</span><span class="o">.</span><span class="n">clamp</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span>
    252. 

  252.     <span class="c1"># Union Area</span>
    253. 

  253.     <span class="n">w1</span><span class="p">,</span> <span class="n">h1</span> <span class="o">=</span> <span class="n">b1_x2</span> <span class="o">-</span> <span class="n">b1_x1</span><span class="p">,</span> <span class="n">b1_y2</span> <span class="o">-</span> <span class="n">b1_y1</span>
    254. 

  254.     <span class="n">w2</span><span class="p">,</span> <span class="n">h2</span> <span class="o">=</span> <span class="n">b2_x2</span> <span class="o">-</span> <span class="n">b2_x1</span><span class="p">,</span> <span class="n">b2_y2</span> <span class="o">-</span> <span class="n">b2_y1</span>
    255. 

  255.     <span class="n">union_area</span> <span class="o">=</span> <span class="p">(</span><span class="n">w1</span> <span class="o">*</span> <span class="n">h1</span> <span class="o">+</span> <span class="mf">1e-16</span><span class="p">)</span> <span class="o">+</span> <span class="n">w2</span> <span class="o">*</span> <span class="n">h2</span> <span class="o">-</span> <span class="n">intersection_area</span>
    256. 

  256.     <span class="n">ious</span> <span class="o">=</span> <span class="n">intersection_area</span> <span class="o">/</span> <span class="n">union_area</span>
    257. 

  257.     <span class="k">return</span> <span class="n">ious</span></div>
    258. 

  258. 

  259. 

  260. <div class="viewcode-block" id="change_bbox_bounds_for_image_size"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.change_bbox_bounds_for_image_size">[docs]</a><span class="k">def</span> <span class="nf">change_bbox_bounds_for_image_size</span><span class="p">(</span><span class="n">boxes</span><span class="p">,</span> <span class="n">img_shape</span><span class="p">):</span>
    261. 

  261.     <span class="c1"># CLIP BOUNDING XYXY BOUNDING BOXES TO IMAGE SHAPE (HEIGHT, WIDTH)</span>
    262. 

  262.     <span class="n">boxes</span><span class="p">[:,</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">]]</span> <span class="o">=</span> <span class="n">boxes</span><span class="p">[:,</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">]]</span><span class="o">.</span><span class="n">clamp</span><span class="p">(</span><span class="nb">min</span><span class="o">=</span><span class="mi">0</span><span class="p">,</span> <span class="nb">max</span><span class="o">=</span><span class="n">img_shape</span><span class="p">[</span><span class="mi">1</span><span class="p">])</span>
    263. 

  263.     <span class="n">boxes</span><span class="p">[:,</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">]]</span> <span class="o">=</span> <span class="n">boxes</span><span class="p">[:,</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">]]</span><span class="o">.</span><span class="n">clamp</span><span class="p">(</span><span class="nb">min</span><span class="o">=</span><span class="mi">0</span><span class="p">,</span> <span class="nb">max</span><span class="o">=</span><span class="n">img_shape</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span>
    264. 

  264.     <span class="k">return</span> <span class="n">boxes</span></div>
    265. 

  265. 

  266. 

  267. <div class="viewcode-block" id="DetectionPostPredictionCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback">[docs]</a><span class="k">class</span> <span class="nc">DetectionPostPredictionCallback</span><span class="p">(</span><span class="n">ABC</span><span class="p">,</span> <span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
    268. 

  268. 

  269.     <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="kc">None</span><span class="p">:</span>
    270. 

  270.         <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
    271. 

  271. 

  272. <div class="viewcode-block" id="DetectionPostPredictionCallback.forward"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback.forward">[docs]</a>    <span class="nd">@abstractmethod</span>
    273. 

  273.     <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">,</span> <span class="n">device</span><span class="p">:</span> <span class="nb">str</span><span class="p">):</span>
    274. 

  274.         <span class="sd">&quot;&quot;&quot;</span>
    275. 

  275. 

  276. <span class="sd">        :param x:       the output of your model</span>
    277. 

  277. <span class="sd">        :param device:  the device to move all output tensors into</span>
    278. 

  278. <span class="sd">        :return:        a list with length batch_size, each item in the list is a detections</span>
    279. 

  279. <span class="sd">                        with shape: nx6 (x1, y1, x2, y2, confidence, class) where x and y are in range [0,1]</span>
    280. 

  280. <span class="sd">        &quot;&quot;&quot;</span>
    281. 

  281.         <span class="k">raise</span> <span class="ne">NotImplementedError</span></div></div>
    282. 

  282. 

  283. 

  284. <div class="viewcode-block" id="IouThreshold"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.IouThreshold">[docs]</a><span class="k">class</span> <span class="nc">IouThreshold</span><span class="p">(</span><span class="nb">tuple</span><span class="p">,</span> <span class="n">Enum</span><span class="p">):</span>
    285. 

  285.     <span class="n">MAP_05</span> <span class="o">=</span> <span class="p">(</span><span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.5</span><span class="p">)</span>
    286. 

  286.     <span class="n">MAP_05_TO_095</span> <span class="o">=</span> <span class="p">(</span><span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.95</span><span class="p">)</span>
    287. 

  287. 

  288. <div class="viewcode-block" id="IouThreshold.is_range"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.IouThreshold.is_range">[docs]</a>    <span class="k">def</span> <span class="nf">is_range</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
    289. 

  289.         <span class="k">return</span> <span class="bp">self</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="o">!=</span> <span class="bp">self</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span></div>
    290. 

  290. 

  291. <div class="viewcode-block" id="IouThreshold.to_tensor"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.IouThreshold.to_tensor">[docs]</a>    <span class="k">def</span> <span class="nf">to_tensor</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
    292. 

  292.         <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">is_range</span><span class="p">():</span>
    293. 

  293.             <span class="n">n_iou_thresh</span> <span class="o">=</span> <span class="nb">int</span><span class="p">(</span><span class="nb">round</span><span class="p">((</span><span class="bp">self</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span> <span class="o">-</span> <span class="bp">self</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span> <span class="o">/</span> <span class="mf">0.05</span><span class="p">))</span> <span class="o">+</span> <span class="mi">1</span>
    294. 

  294.             <span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">linspace</span><span class="p">(</span><span class="bp">self</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="bp">self</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span> <span class="n">n_iou_thresh</span><span class="p">)</span>
    295. 

  295.         <span class="k">else</span><span class="p">:</span>
    296. 

  296.             <span class="n">n_iou_thresh</span> <span class="o">=</span> <span class="mi">1</span>
    297. 

  297.             <span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">([</span><span class="bp">self</span><span class="p">[</span><span class="mi">0</span><span class="p">]])</span></div></div>
    298. 

  298. 

  299. 

  300. <div class="viewcode-block" id="box_iou"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.box_iou">[docs]</a><span class="k">def</span> <span class="nf">box_iou</span><span class="p">(</span><span class="n">box1</span><span class="p">,</span> <span class="n">box2</span><span class="p">):</span>
    301. 

  301.     <span class="c1"># https://github.com/pytorch/vision/blob/master/torchvision/ops/boxes.py</span>
    302. 

  302.     <span class="sd">&quot;&quot;&quot;</span>
    303. 

  303. <span class="sd">    Return intersection-over-union (Jaccard index) of boxes.</span>
    304. 

  304. <span class="sd">    Both sets of boxes are expected to be in (x1, y1, x2, y2) format.</span>
    305. 

  305. <span class="sd">    Arguments:</span>
    306. 

  306. <span class="sd">        box1 (Tensor[N, 4])</span>
    307. 

  307. <span class="sd">        box2 (Tensor[M, 4])</span>
    308. 

  308. <span class="sd">    Returns:</span>
    309. 

  309. <span class="sd">        iou (Tensor[N, M]): the NxM matrix containing the pairwise</span>
    310. 

  310. <span class="sd">            IoU values for every element in boxes1 and boxes2</span>
    311. 

  311. <span class="sd">    &quot;&quot;&quot;</span>
    312. 

  312. 

  313.     <span class="k">def</span> <span class="nf">box_area</span><span class="p">(</span><span class="n">box</span><span class="p">):</span>
    314. 

  314.         <span class="c1"># box = 4xn</span>
    315. 

  315.         <span class="k">return</span> <span class="p">(</span><span class="n">box</span><span class="p">[</span><span class="mi">2</span><span class="p">]</span> <span class="o">-</span> <span class="n">box</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span> <span class="o">*</span> <span class="p">(</span><span class="n">box</span><span class="p">[</span><span class="mi">3</span><span class="p">]</span> <span class="o">-</span> <span class="n">box</span><span class="p">[</span><span class="mi">1</span><span class="p">])</span>
    316. 

  316. 

  317.     <span class="n">area1</span> <span class="o">=</span> <span class="n">box_area</span><span class="p">(</span><span class="n">box1</span><span class="o">.</span><span class="n">T</span><span class="p">)</span>
    318. 

  318.     <span class="n">area2</span> <span class="o">=</span> <span class="n">box_area</span><span class="p">(</span><span class="n">box2</span><span class="o">.</span><span class="n">T</span><span class="p">)</span>
    319. 

  319. 

  320.     <span class="c1"># inter(N,M) = (rb(N,M,2) - lt(N,M,2)).clamp(0).prod(2)</span>
    321. 

  321.     <span class="n">inter</span> <span class="o">=</span> <span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">min</span><span class="p">(</span><span class="n">box1</span><span class="p">[:,</span> <span class="kc">None</span><span class="p">,</span> <span class="mi">2</span><span class="p">:],</span> <span class="n">box2</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">:])</span> <span class="o">-</span> <span class="n">torch</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="n">box1</span><span class="p">[:,</span> <span class="kc">None</span><span class="p">,</span> <span class="p">:</span><span class="mi">2</span><span class="p">],</span> <span class="n">box2</span><span class="p">[:,</span> <span class="p">:</span><span class="mi">2</span><span class="p">]))</span><span class="o">.</span><span class="n">clamp</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span><span class="o">.</span><span class="n">prod</span><span class="p">(</span><span class="mi">2</span><span class="p">)</span>
    322. 

  322.     <span class="k">return</span> <span class="n">inter</span> <span class="o">/</span> <span class="p">(</span><span class="n">area1</span><span class="p">[:,</span> <span class="kc">None</span><span class="p">]</span> <span class="o">+</span> <span class="n">area2</span> <span class="o">-</span> <span class="n">inter</span><span class="p">)</span>  <span class="c1"># iou = inter / (area1 + area2 - inter)</span></div>
    323. 

  323. 

  324. 

  325. <div class="viewcode-block" id="non_max_suppression"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.non_max_suppression">[docs]</a><span class="k">def</span> <span class="nf">non_max_suppression</span><span class="p">(</span><span class="n">prediction</span><span class="p">,</span> <span class="n">conf_thres</span><span class="o">=</span><span class="mf">0.1</span><span class="p">,</span> <span class="n">iou_thres</span><span class="o">=</span><span class="mf">0.6</span><span class="p">,</span>
    326. 

  326.                         <span class="n">multi_label_per_box</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">True</span><span class="p">,</span> <span class="n">with_confidence</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">False</span><span class="p">):</span>
    327. 

  327.     <span class="sd">&quot;&quot;&quot;</span>
    328. 

  328. <span class="sd">    Performs Non-Maximum Suppression (NMS) on inference results</span>
    329. 

  329. <span class="sd">        :param prediction: raw model prediction</span>
    330. 

  330. <span class="sd">        :param conf_thres: below the confidence threshold - prediction are discarded</span>
    331. 

  331. <span class="sd">        :param iou_thres: IoU threshold for the nms algorithm</span>
    332. 

  332. <span class="sd">        :param multi_label_per_box: whether to use re-use each box with all possible labels</span>
    333. 

  333. <span class="sd">                                    (instead of the maximum confidence all confidences above threshold</span>
    334. 

  334. <span class="sd">                                    will be sent to NMS); by default is set to True</span>
    335. 

  335. <span class="sd">        :param with_confidence: whether to multiply objectness score with class score.</span>
    336. 

  336. <span class="sd">                                usually valid for Yolo models only.</span>
    337. 

  337. <span class="sd">        :return:  (x1, y1, x2, y2, object_conf, class_conf, class)</span>
    338. 

  338. <span class="sd">    Returns:</span>
    339. 

  339. <span class="sd">         detections with shape: nx6 (x1, y1, x2, y2, conf, cls)</span>
    340. 

  340. <span class="sd">    &quot;&quot;&quot;</span>
    341. 

  341.     <span class="n">candidates_above_thres</span> <span class="o">=</span> <span class="n">prediction</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="mi">4</span><span class="p">]</span> <span class="o">&gt;</span> <span class="n">conf_thres</span>  <span class="c1"># filter by confidence</span>
    342. 

  342.     <span class="n">output</span> <span class="o">=</span> <span class="p">[</span><span class="kc">None</span><span class="p">]</span> <span class="o">*</span> <span class="n">prediction</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>
    343. 

  343. 

  344.     <span class="k">for</span> <span class="n">image_idx</span><span class="p">,</span> <span class="n">pred</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="n">prediction</span><span class="p">):</span>
    345. 

  345. 

  346.         <span class="n">pred</span> <span class="o">=</span> <span class="n">pred</span><span class="p">[</span><span class="n">candidates_above_thres</span><span class="p">[</span><span class="n">image_idx</span><span class="p">]]</span>  <span class="c1"># confident</span>
    347. 

  347. 

  348.         <span class="k">if</span> <span class="ow">not</span> <span class="n">pred</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">]:</span>  <span class="c1"># If none remain process next image</span>
    349. 

  349.             <span class="k">continue</span>
    350. 

  350. 

  351.         <span class="k">if</span> <span class="n">with_confidence</span><span class="p">:</span>
    352. 

  352.             <span class="n">pred</span><span class="p">[:,</span> <span class="mi">5</span><span class="p">:]</span> <span class="o">*=</span> <span class="n">pred</span><span class="p">[:,</span> <span class="mi">4</span><span class="p">:</span><span class="mi">5</span><span class="p">]</span>  <span class="c1"># multiply objectness score with class score</span>
    353. 

  353. 

  354.         <span class="n">box</span> <span class="o">=</span> <span class="n">convert_xywh_bbox_to_xyxy</span><span class="p">(</span><span class="n">pred</span><span class="p">[:,</span> <span class="p">:</span><span class="mi">4</span><span class="p">])</span>  <span class="c1"># xywh to xyxy</span>
    355. 

  355. 

  356.         <span class="c1"># Detections matrix nx6 (xyxy, conf, cls)</span>
    357. 

  357.         <span class="k">if</span> <span class="n">multi_label_per_box</span><span class="p">:</span>  <span class="c1"># try for all good confidence classes</span>
    358. 

  358.             <span class="n">i</span><span class="p">,</span> <span class="n">j</span> <span class="o">=</span> <span class="p">(</span><span class="n">pred</span><span class="p">[:,</span> <span class="mi">5</span><span class="p">:]</span> <span class="o">&gt;</span> <span class="n">conf_thres</span><span class="p">)</span><span class="o">.</span><span class="n">nonzero</span><span class="p">(</span><span class="n">as_tuple</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span><span class="o">.</span><span class="n">T</span>
    359. 

  359.             <span class="n">pred</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">((</span><span class="n">box</span><span class="p">[</span><span class="n">i</span><span class="p">],</span> <span class="n">pred</span><span class="p">[</span><span class="n">i</span><span class="p">,</span> <span class="n">j</span> <span class="o">+</span> <span class="mi">5</span><span class="p">,</span> <span class="kc">None</span><span class="p">],</span> <span class="n">j</span><span class="p">[:,</span> <span class="kc">None</span><span class="p">]</span><span class="o">.</span><span class="n">float</span><span class="p">()),</span> <span class="mi">1</span><span class="p">)</span>
    360. 

  360. 

  361.         <span class="k">else</span><span class="p">:</span>  <span class="c1"># best class only</span>
    362. 

  362.             <span class="n">conf</span><span class="p">,</span> <span class="n">j</span> <span class="o">=</span> <span class="n">pred</span><span class="p">[:,</span> <span class="mi">5</span><span class="p">:]</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="n">keepdim</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
    363. 

  363.             <span class="n">pred</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">((</span><span class="n">box</span><span class="p">,</span> <span class="n">conf</span><span class="p">,</span> <span class="n">j</span><span class="o">.</span><span class="n">float</span><span class="p">()),</span> <span class="mi">1</span><span class="p">)[</span><span class="n">conf</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">)</span> <span class="o">&gt;</span> <span class="n">conf_thres</span><span class="p">]</span>
    364. 

  364. 

  365.         <span class="k">if</span> <span class="ow">not</span> <span class="n">pred</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">]:</span>  <span class="c1"># If none remain process next image</span>
    366. 

  366.             <span class="k">continue</span>
    367. 

  367. 

  368.         <span class="c1"># Apply torch batched NMS algorithm</span>
    369. 

  369.         <span class="n">boxes</span><span class="p">,</span> <span class="n">scores</span><span class="p">,</span> <span class="n">cls_idx</span> <span class="o">=</span> <span class="n">pred</span><span class="p">[:,</span> <span class="p">:</span><span class="mi">4</span><span class="p">],</span> <span class="n">pred</span><span class="p">[:,</span> <span class="mi">4</span><span class="p">],</span> <span class="n">pred</span><span class="p">[:,</span> <span class="mi">5</span><span class="p">]</span>
    370. 

  370.         <span class="n">idx_to_keep</span> <span class="o">=</span> <span class="n">torchvision</span><span class="o">.</span><span class="n">ops</span><span class="o">.</span><span class="n">boxes</span><span class="o">.</span><span class="n">batched_nms</span><span class="p">(</span><span class="n">boxes</span><span class="p">,</span> <span class="n">scores</span><span class="p">,</span> <span class="n">cls_idx</span><span class="p">,</span> <span class="n">iou_thres</span><span class="p">)</span>
    371. 

  371.         <span class="n">output</span><span class="p">[</span><span class="n">image_idx</span><span class="p">]</span> <span class="o">=</span> <span class="n">pred</span><span class="p">[</span><span class="n">idx_to_keep</span><span class="p">]</span>
    372. 

  372. 

  373.     <span class="k">return</span> <span class="n">output</span></div>
    374. 

  374. 

  375. 

  376. <div class="viewcode-block" id="matrix_non_max_suppression"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.matrix_non_max_suppression">[docs]</a><span class="k">def</span> <span class="nf">matrix_non_max_suppression</span><span class="p">(</span><span class="n">pred</span><span class="p">,</span> <span class="n">conf_thres</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.1</span><span class="p">,</span> <span class="n">kernel</span><span class="p">:</span> <span class="nb">str</span> <span class="o">=</span> <span class="s1">&#39;gaussian&#39;</span><span class="p">,</span>
    377. 

  377.                                <span class="n">sigma</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">3.0</span><span class="p">,</span> <span class="n">max_num_of_detections</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">500</span><span class="p">):</span>
    378. 

  378.     <span class="sd">&quot;&quot;&quot;Performs Matrix Non-Maximum Suppression (NMS) on inference results</span>
    379. 

  379. <span class="sd">        https://arxiv.org/pdf/1912.04488.pdf</span>
    380. 

  380. <span class="sd">        :param pred: raw model prediction (in test mode) - a Tensor of shape [batch, num_predictions, 85]</span>
    381. 

  381. <span class="sd">        where each item format is (x, y, w, h, object_conf, class_conf, ... 80 classes score ...)</span>
    382. 

  382. <span class="sd">        :param conf_thres: below the confidence threshold - prediction are discarded</span>
    383. 

  383. <span class="sd">        :param kernel: type of kernel to use [&#39;gaussian&#39;, &#39;linear&#39;]</span>
    384. 

  384. <span class="sd">        :param sigma: sigma for the gussian kernel</span>
    385. 

  385. <span class="sd">        :param max_num_of_detections: maximum number of boxes to output</span>
    386. 

  386. <span class="sd">        :return:  list of (x1, y1, x2, y2, object_conf, class_conf, class)</span>
    387. 

  387. 

  388. <span class="sd">    Returns:</span>
    389. 

  389. <span class="sd">         detections list with shape: (x1, y1, x2, y2, conf, cls)</span>
    390. 

  390. <span class="sd">    &quot;&quot;&quot;</span>
    391. 

  391.     <span class="c1"># MULTIPLY CONF BY CLASS CONF TO GET COMBINED CONFIDENCE</span>
    392. 

  392.     <span class="n">class_conf</span><span class="p">,</span> <span class="n">class_pred</span> <span class="o">=</span> <span class="n">pred</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">5</span><span class="p">:]</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="mi">2</span><span class="p">)</span>
    393. 

  393.     <span class="n">pred</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">4</span><span class="p">]</span> <span class="o">*=</span> <span class="n">class_conf</span>
    394. 

  394. 

  395.     <span class="c1"># BOX (CENTER X, CENTER Y, WIDTH, HEIGHT) TO (X1, Y1, X2, Y2)</span>
    396. 

  396.     <span class="n">pred</span><span class="p">[:,</span> <span class="p">:,</span> <span class="p">:</span><span class="mi">4</span><span class="p">]</span> <span class="o">=</span> <span class="n">convert_xywh_bbox_to_xyxy</span><span class="p">(</span><span class="n">pred</span><span class="p">[:,</span> <span class="p">:,</span> <span class="p">:</span><span class="mi">4</span><span class="p">])</span>
    397. 

  397. 

  398.     <span class="c1"># DETECTIONS ORDERED AS (x1y1x2y2, obj_conf, class_conf, class_pred)</span>
    399. 

  399.     <span class="n">pred</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">((</span><span class="n">pred</span><span class="p">[:,</span> <span class="p">:,</span> <span class="p">:</span><span class="mi">5</span><span class="p">],</span> <span class="n">class_pred</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="mi">2</span><span class="p">)),</span> <span class="mi">2</span><span class="p">)</span>
    400. 

  400. 

  401.     <span class="c1"># SORT DETECTIONS BY DECREASING CONFIDENCE SCORES</span>
    402. 

  402.     <span class="n">sort_ind</span> <span class="o">=</span> <span class="p">(</span><span class="o">-</span><span class="n">pred</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">4</span><span class="p">])</span><span class="o">.</span><span class="n">argsort</span><span class="p">()</span>
    403. 

  403.     <span class="n">pred</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">stack</span><span class="p">([</span><span class="n">pred</span><span class="p">[</span><span class="n">i</span><span class="p">,</span> <span class="n">sort_ind</span><span class="p">[</span><span class="n">i</span><span class="p">]]</span> <span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">pred</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">])])[:,</span> <span class="mi">0</span><span class="p">:</span><span class="n">max_num_of_detections</span><span class="p">]</span>
    404. 

  404. 

  405.     <span class="n">ious</span> <span class="o">=</span> <span class="n">calc_bbox_iou_matrix</span><span class="p">(</span><span class="n">pred</span><span class="p">)</span>
    406. 

  406. 

  407.     <span class="n">ious</span> <span class="o">=</span> <span class="n">ious</span><span class="o">.</span><span class="n">triu</span><span class="p">(</span><span class="mi">1</span><span class="p">)</span>
    408. 

  408. 

  409.     <span class="c1"># CREATE A LABELS MASK, WE WANT ONLY BOXES WITH THE SAME LABEL TO AFFECT EACH OTHER</span>
    410. 

  410.     <span class="n">labels</span> <span class="o">=</span> <span class="n">pred</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">5</span><span class="p">:]</span>
    411. 

  411.     <span class="n">labeles_matrix</span> <span class="o">=</span> <span class="p">(</span><span class="n">labels</span> <span class="o">==</span> <span class="n">labels</span><span class="o">.</span><span class="n">transpose</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">))</span><span class="o">.</span><span class="n">float</span><span class="p">()</span><span class="o">.</span><span class="n">triu</span><span class="p">(</span><span class="mi">1</span><span class="p">)</span>
    412. 

  412. 

  413.     <span class="n">ious</span> <span class="o">*=</span> <span class="n">labeles_matrix</span>
    414. 

  414.     <span class="n">ious_cmax</span><span class="p">,</span> <span class="n">_</span> <span class="o">=</span> <span class="n">ious</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="mi">1</span><span class="p">)</span>
    415. 

  415.     <span class="n">ious_cmax</span> <span class="o">=</span> <span class="n">ious_cmax</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="mi">2</span><span class="p">)</span><span class="o">.</span><span class="n">repeat</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">max_num_of_detections</span><span class="p">)</span>
    416. 

  416. 

  417.     <span class="k">if</span> <span class="n">kernel</span> <span class="o">==</span> <span class="s1">&#39;gaussian&#39;</span><span class="p">:</span>
    418. 

  418.         <span class="n">decay_matrix</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">exp</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span> <span class="o">*</span> <span class="n">sigma</span> <span class="o">*</span> <span class="p">(</span><span class="n">ious</span> <span class="o">**</span> <span class="mi">2</span><span class="p">))</span>
    419. 

  419.         <span class="n">compensate_matrix</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">exp</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span> <span class="o">*</span> <span class="n">sigma</span> <span class="o">*</span> <span class="p">(</span><span class="n">ious_cmax</span> <span class="o">**</span> <span class="mi">2</span><span class="p">))</span>
    420. 

  420.         <span class="n">decay</span><span class="p">,</span> <span class="n">_</span> <span class="o">=</span> <span class="p">(</span><span class="n">decay_matrix</span> <span class="o">/</span> <span class="n">compensate_matrix</span><span class="p">)</span><span class="o">.</span><span class="n">min</span><span class="p">(</span><span class="n">dim</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
    421. 

  421.     <span class="k">else</span><span class="p">:</span>
    422. 

  422.         <span class="n">decay</span> <span class="o">=</span> <span class="p">(</span><span class="mi">1</span> <span class="o">-</span> <span class="n">ious</span><span class="p">)</span> <span class="o">/</span> <span class="p">(</span><span class="mi">1</span> <span class="o">-</span> <span class="n">ious_cmax</span><span class="p">)</span>
    423. 

  423.         <span class="n">decay</span><span class="p">,</span> <span class="n">_</span> <span class="o">=</span> <span class="n">decay</span><span class="o">.</span><span class="n">min</span><span class="p">(</span><span class="n">dim</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
    424. 

  424. 

  425.     <span class="n">pred</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">4</span><span class="p">]</span> <span class="o">*=</span> <span class="n">decay</span>
    426. 

  426. 

  427.     <span class="n">output</span> <span class="o">=</span> <span class="p">[</span><span class="n">pred</span><span class="p">[</span><span class="n">i</span><span class="p">,</span> <span class="n">pred</span><span class="p">[</span><span class="n">i</span><span class="p">,</span> <span class="p">:,</span> <span class="mi">4</span><span class="p">]</span> <span class="o">&gt;</span> <span class="n">conf_thres</span><span class="p">]</span> <span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">pred</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">])]</span>
    428. 

  428. 

  429.     <span class="k">return</span> <span class="n">output</span></div>
    430. 

  430. 

  431. 

  432. <div class="viewcode-block" id="NMS_Type"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.NMS_Type">[docs]</a><span class="k">class</span> <span class="nc">NMS_Type</span><span class="p">(</span><span class="nb">str</span><span class="p">,</span> <span class="n">Enum</span><span class="p">):</span>
    433. 

  433.     <span class="sd">&quot;&quot;&quot;</span>
    434. 

  434. <span class="sd">    Type of non max suppression algorithm that can be used for post processing detection</span>
    435. 

  435. <span class="sd">    &quot;&quot;&quot;</span>
    436. 

  436.     <span class="n">ITERATIVE</span> <span class="o">=</span> <span class="s1">&#39;iterative&#39;</span>
    437. 

  437.     <span class="n">MATRIX</span> <span class="o">=</span> <span class="s1">&#39;matrix&#39;</span></div>
    438. 

  438. 

  439. 

  440. <div class="viewcode-block" id="undo_image_preprocessing"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.undo_image_preprocessing">[docs]</a><span class="k">def</span> <span class="nf">undo_image_preprocessing</span><span class="p">(</span><span class="n">im_tensor</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="n">np</span><span class="o">.</span><span class="n">ndarray</span><span class="p">:</span>
    441. 

  441.     <span class="sd">&quot;&quot;&quot;</span>
    442. 

  442. <span class="sd">    :param im_tensor: images in a batch after preprocessing for inference, RGB, (B, C, H, W)</span>
    443. 

  443. <span class="sd">    :return:          images in a batch in cv2 format, BGR, (B, H, W, C)</span>
    444. 

  444. <span class="sd">    &quot;&quot;&quot;</span>
    445. 

  445.     <span class="n">im_np</span> <span class="o">=</span> <span class="n">im_tensor</span><span class="o">.</span><span class="n">cpu</span><span class="p">()</span><span class="o">.</span><span class="n">numpy</span><span class="p">()</span>
    446. 

  446.     <span class="n">im_np</span> <span class="o">=</span> <span class="n">im_np</span><span class="p">[:,</span> <span class="p">::</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="p">:,</span> <span class="p">:]</span><span class="o">.</span><span class="n">transpose</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
    447. 

  447.     <span class="n">im_np</span> <span class="o">*=</span> <span class="mf">255.</span>
    448. 

  448.     <span class="k">return</span> <span class="n">np</span><span class="o">.</span><span class="n">ascontiguousarray</span><span class="p">(</span><span class="n">im_np</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">np</span><span class="o">.</span><span class="n">uint8</span><span class="p">)</span></div>
    449. 

  449. 

  450. 

  451. <div class="viewcode-block" id="DetectionVisualization"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.DetectionVisualization">[docs]</a><span class="k">class</span> <span class="nc">DetectionVisualization</span><span class="p">:</span>
    452. 

  452.     <span class="nd">@staticmethod</span>
    453. 

  453.     <span class="k">def</span> <span class="nf">_generate_color_mapping</span><span class="p">(</span><span class="n">num_classes</span><span class="p">:</span> <span class="nb">int</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="n">List</span><span class="p">[</span><span class="n">Tuple</span><span class="p">[</span><span class="nb">int</span><span class="p">]]:</span>
    454. 

  454.         <span class="sd">&quot;&quot;&quot;</span>
    455. 

  455. <span class="sd">        Generate a unique BGR color for each class</span>
    456. 

  456. <span class="sd">        &quot;&quot;&quot;</span>
    457. 

  457.         <span class="n">cmap</span> <span class="o">=</span> <span class="n">plt</span><span class="o">.</span><span class="n">cm</span><span class="o">.</span><span class="n">get_cmap</span><span class="p">(</span><span class="s1">&#39;gist_rainbow&#39;</span><span class="p">,</span> <span class="n">num_classes</span><span class="p">)</span>
    458. 

  458.         <span class="n">colors</span> <span class="o">=</span> <span class="p">[</span><span class="n">cmap</span><span class="p">(</span><span class="n">i</span><span class="p">,</span> <span class="nb">bytes</span><span class="o">=</span><span class="kc">True</span><span class="p">)[:</span><span class="mi">3</span><span class="p">][::</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span> <span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">num_classes</span><span class="p">)]</span>
    459. 

  459.         <span class="k">return</span> <span class="p">[</span><span class="nb">tuple</span><span class="p">(</span><span class="nb">int</span><span class="p">(</span><span class="n">v</span><span class="p">)</span> <span class="k">for</span> <span class="n">v</span> <span class="ow">in</span> <span class="n">c</span><span class="p">)</span> <span class="k">for</span> <span class="n">c</span> <span class="ow">in</span> <span class="n">colors</span><span class="p">]</span>
    460. 

  460. 

  461.     <span class="nd">@staticmethod</span>
    462. 

  462.     <span class="k">def</span> <span class="nf">_draw_box_title</span><span class="p">(</span><span class="n">color_mapping</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="n">Tuple</span><span class="p">[</span><span class="nb">int</span><span class="p">]],</span> <span class="n">class_names</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="nb">str</span><span class="p">],</span> <span class="n">box_thickness</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
    463. 

  463.                         <span class="n">image_np</span><span class="p">:</span> <span class="n">np</span><span class="o">.</span><span class="n">ndarray</span><span class="p">,</span> <span class="n">x1</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">y1</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">x2</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">y2</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">class_id</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
    464. 

  464.                         <span class="n">pred_conf</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span> <span class="n">is_target</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">False</span><span class="p">):</span>
    465. 

  465.         <span class="n">color</span> <span class="o">=</span> <span class="n">color_mapping</span><span class="p">[</span><span class="n">class_id</span><span class="p">]</span>
    466. 

  466.         <span class="n">class_name</span> <span class="o">=</span> <span class="n">class_names</span><span class="p">[</span><span class="n">class_id</span><span class="p">]</span>
    467. 

  467. 

  468.         <span class="c1"># Draw the box</span>
    469. 

  469.         <span class="n">image_np</span> <span class="o">=</span> <span class="n">cv2</span><span class="o">.</span><span class="n">rectangle</span><span class="p">(</span><span class="n">image_np</span><span class="p">,</span> <span class="p">(</span><span class="n">x1</span><span class="p">,</span> <span class="n">y1</span><span class="p">),</span> <span class="p">(</span><span class="n">x2</span><span class="p">,</span> <span class="n">y2</span><span class="p">),</span> <span class="n">color</span><span class="p">,</span> <span class="n">box_thickness</span><span class="p">)</span>
    470. 

  470. 

  471.         <span class="c1"># Caption with class name and confidence if given</span>
    472. 

  472.         <span class="n">text_color</span> <span class="o">=</span> <span class="p">(</span><span class="mi">255</span><span class="p">,</span> <span class="mi">255</span><span class="p">,</span> <span class="mi">255</span><span class="p">)</span>  <span class="c1"># white</span>
    473. 

  473. 

  474.         <span class="k">if</span> <span class="n">is_target</span><span class="p">:</span>
    475. 

  475.             <span class="n">title</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">&#39;[GT] </span><span class="si">{</span><span class="n">class_name</span><span class="si">}</span><span class="s1">&#39;</span>
    476. 

  476.         <span class="k">if</span> <span class="ow">not</span> <span class="n">is_target</span><span class="p">:</span>
    477. 

  477.             <span class="n">title</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">&#39;[Pred] </span><span class="si">{</span><span class="n">class_name</span><span class="si">}</span><span class="s1">  </span><span class="si">{</span><span class="nb">str</span><span class="p">(</span><span class="nb">round</span><span class="p">(</span><span class="n">pred_conf</span><span class="p">,</span> <span class="mi">2</span><span class="p">))</span> <span class="k">if</span> <span class="n">pred_conf</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span> <span class="k">else</span> <span class="s2">&quot;&quot;</span><span class="si">}</span><span class="s1">&#39;</span>
    478. 

  478. 

  479.         <span class="n">image_np</span> <span class="o">=</span> <span class="n">cv2</span><span class="o">.</span><span class="n">rectangle</span><span class="p">(</span><span class="n">image_np</span><span class="p">,</span> <span class="p">(</span><span class="n">x1</span><span class="p">,</span> <span class="n">y1</span> <span class="o">-</span> <span class="mi">15</span><span class="p">),</span> <span class="p">(</span><span class="n">x1</span> <span class="o">+</span> <span class="nb">len</span><span class="p">(</span><span class="n">title</span><span class="p">)</span> <span class="o">*</span> <span class="mi">10</span><span class="p">,</span> <span class="n">y1</span><span class="p">),</span> <span class="n">color</span><span class="p">,</span> <span class="n">cv2</span><span class="o">.</span><span class="n">FILLED</span><span class="p">)</span>
    480. 

  480.         <span class="n">image_np</span> <span class="o">=</span> <span class="n">cv2</span><span class="o">.</span><span class="n">putText</span><span class="p">(</span><span class="n">image_np</span><span class="p">,</span> <span class="n">title</span><span class="p">,</span> <span class="p">(</span><span class="n">x1</span><span class="p">,</span> <span class="n">y1</span> <span class="o">-</span> <span class="n">box_thickness</span><span class="p">),</span> <span class="mi">2</span><span class="p">,</span> <span class="mf">.5</span><span class="p">,</span> <span class="n">text_color</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">lineType</span><span class="o">=</span><span class="n">cv2</span><span class="o">.</span><span class="n">LINE_AA</span><span class="p">)</span>
    481. 

  481. 

  482.         <span class="k">return</span> <span class="n">image_np</span>
    483. 

  483. 

  484.     <span class="nd">@staticmethod</span>
    485. 

  485.     <span class="k">def</span> <span class="nf">_visualize_image</span><span class="p">(</span><span class="n">image_np</span><span class="p">:</span> <span class="n">np</span><span class="o">.</span><span class="n">ndarray</span><span class="p">,</span> <span class="n">pred_boxes</span><span class="p">:</span> <span class="n">np</span><span class="o">.</span><span class="n">ndarray</span><span class="p">,</span> <span class="n">target_boxes</span><span class="p">:</span> <span class="n">np</span><span class="o">.</span><span class="n">ndarray</span><span class="p">,</span>
    486. 

  486.                          <span class="n">class_names</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="nb">str</span><span class="p">],</span> <span class="n">box_thickness</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">gt_alpha</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span> <span class="n">image_scale</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span>
    487. 

  487.                          <span class="n">checkpoint_dir</span><span class="p">:</span> <span class="nb">str</span><span class="p">,</span> <span class="n">image_name</span><span class="p">:</span> <span class="nb">str</span><span class="p">):</span>
    488. 

  488.         <span class="n">image_np</span> <span class="o">=</span> <span class="n">cv2</span><span class="o">.</span><span class="n">resize</span><span class="p">(</span><span class="n">image_np</span><span class="p">,</span> <span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">0</span><span class="p">),</span> <span class="n">fx</span><span class="o">=</span><span class="n">image_scale</span><span class="p">,</span> <span class="n">fy</span><span class="o">=</span><span class="n">image_scale</span><span class="p">,</span> <span class="n">interpolation</span><span class="o">=</span><span class="n">cv2</span><span class="o">.</span><span class="n">INTER_NEAREST</span><span class="p">)</span>
    489. 

  489.         <span class="n">color_mapping</span> <span class="o">=</span> <span class="n">DetectionVisualization</span><span class="o">.</span><span class="n">_generate_color_mapping</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">class_names</span><span class="p">))</span>
    490. 

  490. 

  491.         <span class="c1"># Draw predictions</span>
    492. 

  492.         <span class="n">pred_boxes</span><span class="p">[:,</span> <span class="p">:</span><span class="mi">4</span><span class="p">]</span> <span class="o">*=</span> <span class="n">image_scale</span>
    493. 

  493.         <span class="k">for</span> <span class="n">box</span> <span class="ow">in</span> <span class="n">pred_boxes</span><span class="p">:</span>
    494. 

  494.             <span class="n">image_np</span> <span class="o">=</span> <span class="n">DetectionVisualization</span><span class="o">.</span><span class="n">_draw_box_title</span><span class="p">(</span><span class="n">color_mapping</span><span class="p">,</span> <span class="n">class_names</span><span class="p">,</span> <span class="n">box_thickness</span><span class="p">,</span>
    495. 

  495.                                                               <span class="n">image_np</span><span class="p">,</span> <span class="o">*</span><span class="n">box</span><span class="p">[:</span><span class="mi">4</span><span class="p">]</span><span class="o">.</span><span class="n">astype</span><span class="p">(</span><span class="nb">int</span><span class="p">),</span>
    496. 

  496.                                                               <span class="n">class_id</span><span class="o">=</span><span class="nb">int</span><span class="p">(</span><span class="n">box</span><span class="p">[</span><span class="mi">5</span><span class="p">]),</span> <span class="n">pred_conf</span><span class="o">=</span><span class="n">box</span><span class="p">[</span><span class="mi">4</span><span class="p">])</span>
    497. 

  497. 

  498.         <span class="c1"># Draw ground truths</span>
    499. 

  499.         <span class="n">target_boxes_image</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">zeros_like</span><span class="p">(</span><span class="n">image_np</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">uint8</span><span class="p">)</span>
    500. 

  500.         <span class="k">for</span> <span class="n">box</span> <span class="ow">in</span> <span class="n">target_boxes</span><span class="p">:</span>
    501. 

  501.             <span class="n">target_boxes_image</span> <span class="o">=</span> <span class="n">DetectionVisualization</span><span class="o">.</span><span class="n">_draw_box_title</span><span class="p">(</span><span class="n">color_mapping</span><span class="p">,</span> <span class="n">class_names</span><span class="p">,</span> <span class="n">box_thickness</span><span class="p">,</span>
    502. 

  502.                                                                         <span class="n">target_boxes_image</span><span class="p">,</span> <span class="o">*</span><span class="n">box</span><span class="p">[</span><span class="mi">2</span><span class="p">:],</span>
    503. 

  503.                                                                         <span class="n">class_id</span><span class="o">=</span><span class="n">box</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span> <span class="n">is_target</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
    504. 

  504. 

  505.         <span class="c1"># Transparent overlay of ground truth boxes</span>
    506. 

  506.         <span class="n">mask</span> <span class="o">=</span> <span class="n">target_boxes_image</span><span class="o">.</span><span class="n">astype</span><span class="p">(</span><span class="nb">bool</span><span class="p">)</span>
    507. 

  507.         <span class="n">image_np</span><span class="p">[</span><span class="n">mask</span><span class="p">]</span> <span class="o">=</span> <span class="n">cv2</span><span class="o">.</span><span class="n">addWeighted</span><span class="p">(</span><span class="n">image_np</span><span class="p">,</span> <span class="mi">1</span> <span class="o">-</span> <span class="n">gt_alpha</span><span class="p">,</span> <span class="n">target_boxes_image</span><span class="p">,</span> <span class="n">gt_alpha</span><span class="p">,</span> <span class="mi">0</span><span class="p">)[</span><span class="n">mask</span><span class="p">]</span>
    508. 

  508. 

  509.         <span class="k">if</span> <span class="n">checkpoint_dir</span> <span class="ow">is</span> <span class="kc">None</span><span class="p">:</span>
    510. 

  510.             <span class="k">return</span> <span class="n">image_np</span>
    511. 

  511.         <span class="k">else</span><span class="p">:</span>
    512. 

  512.             <span class="n">pathlib</span><span class="o">.</span><span class="n">Path</span><span class="p">(</span><span class="n">checkpoint_dir</span><span class="p">)</span><span class="o">.</span><span class="n">mkdir</span><span class="p">(</span><span class="n">parents</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="n">exist_ok</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
    513. 

  513.             <span class="n">cv2</span><span class="o">.</span><span class="n">imwrite</span><span class="p">(</span><span class="n">os</span><span class="o">.</span><span class="n">path</span><span class="o">.</span><span class="n">join</span><span class="p">(</span><span class="n">checkpoint_dir</span><span class="p">,</span> <span class="nb">str</span><span class="p">(</span><span class="n">image_name</span><span class="p">)</span> <span class="o">+</span> <span class="s1">&#39;.jpg&#39;</span><span class="p">),</span> <span class="n">image_np</span><span class="p">)</span>
    514. 

  514. 

  515.     <span class="nd">@staticmethod</span>
    516. 

  516.     <span class="k">def</span> <span class="nf">_scaled_ccwh_to_xyxy</span><span class="p">(</span><span class="n">target_boxes</span><span class="p">:</span> <span class="n">np</span><span class="o">.</span><span class="n">ndarray</span><span class="p">,</span> <span class="n">h</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">w</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">image_scale</span><span class="p">:</span> <span class="nb">float</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="n">np</span><span class="o">.</span><span class="n">ndarray</span><span class="p">:</span>
    517. 

  517.         <span class="sd">&quot;&quot;&quot;</span>
    518. 

  518. <span class="sd">        Modifies target_boxes inplace</span>
    519. 

  519. <span class="sd">        :param target_boxes:    (c1, c2, w, h) boxes in [0, 1] range</span>
    520. 

  520. <span class="sd">        :param h:               image height</span>
    521. 

  521. <span class="sd">        :param w:               image width</span>
    522. 

  522. <span class="sd">        :param image_scale:     desired scale for the boxes w.r.t. w and h</span>
    523. 

  523. <span class="sd">        :return:                targets in (x1, y1, x2, y2) format</span>
    524. 

  524. <span class="sd">                                in range [0, w * self.image_scale] [0, h * self.image_scale]</span>
    525. 

  525. <span class="sd">        &quot;&quot;&quot;</span>
    526. 

  526.         <span class="c1"># unscale</span>
    527. 

  527.         <span class="n">target_boxes</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">:]</span> <span class="o">*=</span> <span class="n">np</span><span class="o">.</span><span class="n">array</span><span class="p">([[</span><span class="n">w</span><span class="p">,</span> <span class="n">h</span><span class="p">,</span> <span class="n">w</span><span class="p">,</span> <span class="n">h</span><span class="p">]])</span>
    528. 

  528. 

  529.         <span class="c1"># x1 = c1 - w // 2; y1 = c2 - h // 2</span>
    530. 

  530.         <span class="n">target_boxes</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">]</span> <span class="o">-=</span> <span class="n">target_boxes</span><span class="p">[:,</span> <span class="mi">4</span><span class="p">]</span> <span class="o">//</span> <span class="mi">2</span>
    531. 

  531.         <span class="n">target_boxes</span><span class="p">[:,</span> <span class="mi">3</span><span class="p">]</span> <span class="o">-=</span> <span class="n">target_boxes</span><span class="p">[:,</span> <span class="mi">5</span><span class="p">]</span> <span class="o">//</span> <span class="mi">2</span>
    532. 

  532.         <span class="c1"># x2 = w + x1; y2 = h + y1</span>
    533. 

  533.         <span class="n">target_boxes</span><span class="p">[:,</span> <span class="mi">4</span><span class="p">]</span> <span class="o">+=</span> <span class="n">target_boxes</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">]</span>
    534. 

  534.         <span class="n">target_boxes</span><span class="p">[:,</span> <span class="mi">5</span><span class="p">]</span> <span class="o">+=</span> <span class="n">target_boxes</span><span class="p">[:,</span> <span class="mi">3</span><span class="p">]</span>
    535. 

  535. 

  536.         <span class="n">target_boxes</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">:]</span> <span class="o">*=</span> <span class="n">image_scale</span>
    537. 

  537.         <span class="n">target_boxes</span> <span class="o">=</span> <span class="n">target_boxes</span><span class="o">.</span><span class="n">astype</span><span class="p">(</span><span class="nb">int</span><span class="p">)</span>
    538. 

  538.         <span class="k">return</span> <span class="n">target_boxes</span>
    539. 

  539. 

  540. <div class="viewcode-block" id="DetectionVisualization.visualize_batch"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.DetectionVisualization.visualize_batch">[docs]</a>    <span class="nd">@staticmethod</span>
    541. 

  541.     <span class="k">def</span> <span class="nf">visualize_batch</span><span class="p">(</span><span class="n">image_tensor</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">pred_boxes</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">],</span> <span class="n">target_boxes</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    542. 

  542.                         <span class="n">batch_name</span><span class="p">:</span> <span class="n">Union</span><span class="p">[</span><span class="nb">int</span><span class="p">,</span> <span class="nb">str</span><span class="p">],</span> <span class="n">class_names</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="nb">str</span><span class="p">],</span> <span class="n">checkpoint_dir</span><span class="p">:</span> <span class="nb">str</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
    543. 

  543.                         <span class="n">undo_preprocessing_func</span><span class="p">:</span> <span class="n">Callable</span><span class="p">[[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">],</span> <span class="n">np</span><span class="o">.</span><span class="n">ndarray</span><span class="p">]</span> <span class="o">=</span> <span class="n">undo_image_preprocessing</span><span class="p">,</span>
    544. 

  544.                         <span class="n">box_thickness</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">2</span><span class="p">,</span> <span class="n">image_scale</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">1.</span><span class="p">,</span> <span class="n">gt_alpha</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">.4</span><span class="p">):</span>
    545. 

  545.         <span class="sd">&quot;&quot;&quot;</span>
    546. 

  546. <span class="sd">        A helper function to visualize detections predicted by a network:</span>
    547. 

  547. <span class="sd">        saves images into a given path with a name that is {batch_name}_{imade_idx_in_the_batch}.jpg, one batch per call.</span>
    548. 

  548. <span class="sd">        Colors are generated on the fly: uniformly sampled from color wheel to support all given classes.</span>
    549. 

  549. 

  550. <span class="sd">        Adjustable:</span>
    551. 

  551. <span class="sd">            * Ground truth box transparency;</span>
    552. 

  552. <span class="sd">            * Box width;</span>
    553. 

  553. <span class="sd">            * Image size (larger or smaller than what&#39;s provided)</span>
    554. 

  554. 

  555. <span class="sd">        :param image_tensor:            rgb images, (B, H, W, 3)</span>
    556. 

  556. <span class="sd">        :param pred_boxes:              boxes after NMS for each image in a batch, each (Num_boxes, 6),</span>
    557. 

  557. <span class="sd">                                        values on dim 1 are: x1, y1, x2, y2, confidence, class</span>
    558. 

  558. <span class="sd">        :param target_boxes:            (Num_targets, 6), values on dim 1 are: image id in a batch, class, x y w h</span>
    559. 

  559. <span class="sd">                                        (coordinates scaled to [0, 1])</span>
    560. 

  560. <span class="sd">        :param batch_name:              id of the current batch to use for image naming</span>
    561. 

  561. 

  562. <span class="sd">        :param class_names:             names of all classes, each on its own index</span>
    563. 

  563. <span class="sd">        :param checkpoint_dir:          a path where images with boxes will be saved. if None, the result images will</span>
    564. 

  564. <span class="sd">                                        be returns as a list of numpy image arrays</span>
    565. 

  565. 

  566. <span class="sd">        :param undo_preprocessing_func: a function to convert preprocessed images tensor into a batch of cv2-like images</span>
    567. 

  567. <span class="sd">        :param box_thickness:           box line thickness in px</span>
    568. 

  568. <span class="sd">        :param image_scale:             scale of an image w.r.t. given image size,</span>
    569. 

  569. <span class="sd">                                        e.g. incoming images are (320x320), use scale = 2. to preview in (640x640)</span>
    570. 

  570. <span class="sd">        :param gt_alpha:                a value in [0., 1.] transparency on ground truth boxes,</span>
    571. 

  571. <span class="sd">                                        0 for invisible, 1 for fully opaque</span>
    572. 

  572. <span class="sd">        &quot;&quot;&quot;</span>
    573. 

  573.         <span class="n">image_np</span> <span class="o">=</span> <span class="n">undo_preprocessing_func</span><span class="p">(</span><span class="n">image_tensor</span><span class="o">.</span><span class="n">detach</span><span class="p">())</span>
    574. 

  574.         <span class="n">targets</span> <span class="o">=</span> <span class="n">DetectionVisualization</span><span class="o">.</span><span class="n">_scaled_ccwh_to_xyxy</span><span class="p">(</span><span class="n">target_boxes</span><span class="o">.</span><span class="n">detach</span><span class="p">()</span><span class="o">.</span><span class="n">cpu</span><span class="p">()</span><span class="o">.</span><span class="n">numpy</span><span class="p">(),</span> <span class="o">*</span><span class="n">image_np</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">1</span><span class="p">:</span><span class="mi">3</span><span class="p">],</span>
    575. 

  575.                                                               <span class="n">image_scale</span><span class="p">)</span>
    576. 

  576. 

  577.         <span class="n">out_images</span> <span class="o">=</span> <span class="p">[]</span>
    578. 

  578.         <span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">image_np</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">]):</span>
    579. 

  579.             <span class="n">preds</span> <span class="o">=</span> <span class="n">pred_boxes</span><span class="p">[</span><span class="n">i</span><span class="p">]</span><span class="o">.</span><span class="n">detach</span><span class="p">()</span><span class="o">.</span><span class="n">cpu</span><span class="p">()</span><span class="o">.</span><span class="n">numpy</span><span class="p">()</span> <span class="k">if</span> <span class="n">pred_boxes</span><span class="p">[</span><span class="n">i</span><span class="p">]</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span> <span class="k">else</span> <span class="n">np</span><span class="o">.</span><span class="n">empty</span><span class="p">((</span><span class="mi">0</span><span class="p">,</span> <span class="mi">6</span><span class="p">))</span>
    580. 

  580.             <span class="n">targets_cur</span> <span class="o">=</span> <span class="n">targets</span><span class="p">[</span><span class="n">targets</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">==</span> <span class="n">i</span><span class="p">]</span>
    581. 

  581. 

  582.             <span class="n">image_name</span> <span class="o">=</span> <span class="s1">&#39;_&#39;</span><span class="o">.</span><span class="n">join</span><span class="p">([</span><span class="nb">str</span><span class="p">(</span><span class="n">batch_name</span><span class="p">),</span> <span class="nb">str</span><span class="p">(</span><span class="n">i</span><span class="p">)])</span>
    583. 

  583.             <span class="n">res_image</span> <span class="o">=</span> <span class="n">DetectionVisualization</span><span class="o">.</span><span class="n">_visualize_image</span><span class="p">(</span><span class="n">image_np</span><span class="p">[</span><span class="n">i</span><span class="p">],</span> <span class="n">preds</span><span class="p">,</span> <span class="n">targets_cur</span><span class="p">,</span> <span class="n">class_names</span><span class="p">,</span> <span class="n">box_thickness</span><span class="p">,</span> <span class="n">gt_alpha</span><span class="p">,</span> <span class="n">image_scale</span><span class="p">,</span>
    584. 

  584.                                                                 <span class="n">checkpoint_dir</span><span class="p">,</span> <span class="n">image_name</span><span class="p">)</span>
    585. 

  585.             <span class="k">if</span> <span class="n">res_image</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
    586. 

  586.                 <span class="n">out_images</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">res_image</span><span class="p">)</span>
    587. 

  587. 

  588.         <span class="k">return</span> <span class="n">out_images</span></div></div>
    589. 

  589. 

  590. 

  591. <div class="viewcode-block" id="Anchors"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.Anchors">[docs]</a><span class="k">class</span> <span class="nc">Anchors</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
    592. 

  592.     <span class="sd">&quot;&quot;&quot;</span>
    593. 

  593. <span class="sd">    A wrapper function to hold the anchors used by detection models such as Yolo</span>
    594. 

  594. <span class="sd">    &quot;&quot;&quot;</span>
    595. 

  595. 

  596.     <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">anchors_list</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="n">List</span><span class="p">],</span> <span class="n">strides</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="nb">int</span><span class="p">]):</span>
    597. 

  597.         <span class="sd">&quot;&quot;&quot;</span>
    598. 

  598. <span class="sd">        :param anchors_list: of the shape [[w1,h1,w2,h2,w3,h3], [w4,h4,w5,h5,w6,h6] .... where each sublist holds</span>
    599. 

  599. <span class="sd">            the width and height of the anchors of a specific detection layer.</span>
    600. 

  600. <span class="sd">            i.e. for a model with 3 detection layers, each containing 5 anchors the format will be a of 3 sublists of 10 numbers each</span>
    601. 

  601. <span class="sd">            The width and height are in pixels (not relative to image size)</span>
    602. 

  602. <span class="sd">        :param strides: a list containing the stride of the layers from which the detection heads are fed.</span>
    603. 

  603. <span class="sd">            i.e. if the firs detection head is connected to the backbone after the input dimensions were reduces by 8, the first number will be 8</span>
    604. 

  604. <span class="sd">        &quot;&quot;&quot;</span>
    605. 

  605.         <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
    606. 

  606. 

  607.         <span class="bp">self</span><span class="o">.</span><span class="n">__anchors_list</span> <span class="o">=</span> <span class="n">anchors_list</span>
    608. 

  608.         <span class="bp">self</span><span class="o">.</span><span class="n">__strides</span> <span class="o">=</span> <span class="n">strides</span>
    609. 

  609. 

  610.         <span class="bp">self</span><span class="o">.</span><span class="n">_check_all_lists</span><span class="p">(</span><span class="n">anchors_list</span><span class="p">)</span>
    611. 

  611.         <span class="bp">self</span><span class="o">.</span><span class="n">_check_all_len_equal_and_even</span><span class="p">(</span><span class="n">anchors_list</span><span class="p">)</span>
    612. 

  612. 

  613.         <span class="bp">self</span><span class="o">.</span><span class="n">_stride</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">(</span><span class="n">strides</span><span class="p">)</span><span class="o">.</span><span class="n">float</span><span class="p">(),</span> <span class="n">requires_grad</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span>
    614. 

  614.         <span class="n">anchors</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">(</span><span class="n">anchors_list</span><span class="p">)</span><span class="o">.</span><span class="n">float</span><span class="p">()</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">anchors_list</span><span class="p">),</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">)</span>
    615. 

  615.         <span class="bp">self</span><span class="o">.</span><span class="n">_anchors</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">(</span><span class="n">anchors</span> <span class="o">/</span> <span class="bp">self</span><span class="o">.</span><span class="n">_stride</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span> <span class="n">requires_grad</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span>
    616. 

  616.         <span class="bp">self</span><span class="o">.</span><span class="n">_anchor_grid</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">(</span><span class="n">anchors</span><span class="o">.</span><span class="n">clone</span><span class="p">()</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">anchors_list</span><span class="p">),</span> <span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">),</span> <span class="n">requires_grad</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span>
    617. 

  617. 

  618.     <span class="nd">@staticmethod</span>
    619. 

  619.     <span class="k">def</span> <span class="nf">_check_all_lists</span><span class="p">(</span><span class="n">anchors</span><span class="p">:</span> <span class="nb">list</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="nb">bool</span><span class="p">:</span>
    620. 

  620.         <span class="k">for</span> <span class="n">a</span> <span class="ow">in</span> <span class="n">anchors</span><span class="p">:</span>
    621. 

  621.             <span class="k">if</span> <span class="ow">not</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="p">(</span><span class="nb">list</span><span class="p">,</span> <span class="n">ListConfig</span><span class="p">)):</span>
    622. 

  622.                 <span class="k">raise</span> <span class="ne">RuntimeError</span><span class="p">(</span><span class="s1">&#39;All objects of anchors_list must be lists&#39;</span><span class="p">)</span>
    623. 

  623. 

  624.     <span class="nd">@staticmethod</span>
    625. 

  625.     <span class="k">def</span> <span class="nf">_check_all_len_equal_and_even</span><span class="p">(</span><span class="n">anchors</span><span class="p">:</span> <span class="nb">list</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="nb">bool</span><span class="p">:</span>
    626. 

  626.         <span class="n">len_of_first</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="n">anchors</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span>
    627. 

  627.         <span class="k">for</span> <span class="n">a</span> <span class="ow">in</span> <span class="n">anchors</span><span class="p">:</span>
    628. 

  628.             <span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="n">a</span><span class="p">)</span> <span class="o">%</span> <span class="mi">2</span> <span class="o">==</span> <span class="mi">1</span> <span class="ow">or</span> <span class="nb">len</span><span class="p">(</span><span class="n">a</span><span class="p">)</span> <span class="o">!=</span> <span class="n">len_of_first</span><span class="p">:</span>
    629. 

  629.                 <span class="k">raise</span> <span class="ne">RuntimeError</span><span class="p">(</span><span class="s1">&#39;All objects of anchors_list must be of the same even length&#39;</span><span class="p">)</span>
    630. 

  630. 

  631.     <span class="nd">@property</span>
    632. 

  632.     <span class="k">def</span> <span class="nf">stride</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">:</span>
    633. 

  633.         <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">_stride</span>
    634. 

  634. 

  635.     <span class="nd">@property</span>
    636. 

  636.     <span class="k">def</span> <span class="nf">anchors</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">:</span>
    637. 

  637.         <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">_anchors</span>
    638. 

  638. 

  639.     <span class="nd">@property</span>
    640. 

  640.     <span class="k">def</span> <span class="nf">anchor_grid</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">:</span>
    641. 

  641.         <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">_anchor_grid</span>
    642. 

  642. 

  643.     <span class="nd">@property</span>
    644. 

  644.     <span class="k">def</span> <span class="nf">detection_layers_num</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="nb">int</span><span class="p">:</span>
    645. 

  645.         <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">_anchors</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>
    646. 

  646. 

  647.     <span class="nd">@property</span>
    648. 

  648.     <span class="k">def</span> <span class="nf">num_anchors</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="nb">int</span><span class="p">:</span>
    649. 

  649.         <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">_anchors</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span>
    650. 

  650. 

  651.     <span class="k">def</span> <span class="fm">__repr__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
    652. 

  652.         <span class="k">return</span> <span class="sa">f</span><span class="s2">&quot;anchors_list: </span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">__anchors_list</span><span class="si">}</span><span class="s2"> strides: </span><span class="si">{</span><span class="bp">self</span><span class="o">.</span><span class="n">__strides</span><span class="si">}</span><span class="s2">&quot;</span></div>
    653. 

  653. 

  654. 

  655. <div class="viewcode-block" id="xyxy2cxcywh"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.xyxy2cxcywh">[docs]</a><span class="k">def</span> <span class="nf">xyxy2cxcywh</span><span class="p">(</span><span class="n">bboxes</span><span class="p">):</span>
    656. 

  656.     <span class="sd">&quot;&quot;&quot;</span>
    657. 

  657. <span class="sd">    Transforms bboxes from xyxy format to centerized xy wh format</span>
    658. 

  658. <span class="sd">    :param bboxes: array, shaped (nboxes, 4)</span>
    659. 

  659. <span class="sd">    :return: modified bboxes</span>
    660. 

  660. <span class="sd">    &quot;&quot;&quot;</span>
    661. 

  661.     <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">]</span> <span class="o">=</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">]</span> <span class="o">-</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span>
    662. 

  662.     <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">3</span><span class="p">]</span> <span class="o">=</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">3</span><span class="p">]</span> <span class="o">-</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">]</span>
    663. 

  663.     <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">=</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">+</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">]</span> <span class="o">*</span> <span class="mf">0.5</span>
    664. 

  664.     <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">]</span> <span class="o">=</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">]</span> <span class="o">+</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">3</span><span class="p">]</span> <span class="o">*</span> <span class="mf">0.5</span>
    665. 

  665.     <span class="k">return</span> <span class="n">bboxes</span></div>
    666. 

  666. 

  667. 

  668. <div class="viewcode-block" id="cxcywh2xyxy"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.cxcywh2xyxy">[docs]</a><span class="k">def</span> <span class="nf">cxcywh2xyxy</span><span class="p">(</span><span class="n">bboxes</span><span class="p">):</span>
    669. 

  669.     <span class="sd">&quot;&quot;&quot;</span>
    670. 

  670. <span class="sd">    Transforms bboxes from centerized xy wh format to xyxy format</span>
    671. 

  671. <span class="sd">    :param bboxes: array, shaped (nboxes, 4)</span>
    672. 

  672. <span class="sd">    :return: modified bboxes</span>
    673. 

  673. <span class="sd">    &quot;&quot;&quot;</span>
    674. 

  674.     <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">]</span> <span class="o">=</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">]</span> <span class="o">-</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">3</span><span class="p">]</span> <span class="o">*</span> <span class="mf">0.5</span>
    675. 

  675.     <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">=</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">-</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">]</span> <span class="o">*</span> <span class="mf">0.5</span>
    676. 

  676.     <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">3</span><span class="p">]</span> <span class="o">=</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">3</span><span class="p">]</span> <span class="o">+</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">]</span>
    677. 

  677.     <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">]</span> <span class="o">=</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">]</span> <span class="o">+</span> <span class="n">bboxes</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span>
    678. 

  678.     <span class="k">return</span> <span class="n">bboxes</span></div>
    679. 

  679. 

  680. 

  681. <div class="viewcode-block" id="get_mosaic_coordinate"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.get_mosaic_coordinate">[docs]</a><span class="k">def</span> <span class="nf">get_mosaic_coordinate</span><span class="p">(</span><span class="n">mosaic_index</span><span class="p">,</span> <span class="n">xc</span><span class="p">,</span> <span class="n">yc</span><span class="p">,</span> <span class="n">w</span><span class="p">,</span> <span class="n">h</span><span class="p">,</span> <span class="n">input_h</span><span class="p">,</span> <span class="n">input_w</span><span class="p">):</span>
    682. 

  682.     <span class="sd">&quot;&quot;&quot;</span>
    683. 

  683. <span class="sd">    Returns the mosaic coordinates of final mosaic image according to mosaic image index.</span>
    684. 

  684. 

  685. <span class="sd">    :param mosaic_index: (int) mosaic image index</span>
    686. 

  686. <span class="sd">    :param xc: (int) center x coordinate of the entire mosaic grid.</span>
    687. 

  687. <span class="sd">    :param yc: (int) center y coordinate of the entire mosaic grid.</span>
    688. 

  688. <span class="sd">    :param w: (int) width of bbox</span>
    689. 

  689. <span class="sd">    :param h: (int) height of bbox</span>
    690. 

  690. <span class="sd">    :param input_h: (int) image input height (should be 1/2 of the final mosaic output image height).</span>
    691. 

  691. <span class="sd">    :param input_w: (int) image input width (should be 1/2 of the final mosaic output image width).</span>
    692. 

  692. <span class="sd">    :return: (x1, y1, x2, y2), (x1s, y1s, x2s, y2s) where (x1, y1, x2, y2) are the coordinates in the final mosaic</span>
    693. 

  693. <span class="sd">        output image, and (x1s, y1s, x2s, y2s) are the coordinates in the placed image.</span>
    694. 

  694. <span class="sd">    &quot;&quot;&quot;</span>
    695. 

  695.     <span class="c1"># index0 to top left part of image</span>
    696. 

  696.     <span class="k">if</span> <span class="n">mosaic_index</span> <span class="o">==</span> <span class="mi">0</span><span class="p">:</span>
    697. 

  697.         <span class="n">x1</span><span class="p">,</span> <span class="n">y1</span><span class="p">,</span> <span class="n">x2</span><span class="p">,</span> <span class="n">y2</span> <span class="o">=</span> <span class="nb">max</span><span class="p">(</span><span class="n">xc</span> <span class="o">-</span> <span class="n">w</span><span class="p">,</span> <span class="mi">0</span><span class="p">),</span> <span class="nb">max</span><span class="p">(</span><span class="n">yc</span> <span class="o">-</span> <span class="n">h</span><span class="p">,</span> <span class="mi">0</span><span class="p">),</span> <span class="n">xc</span><span class="p">,</span> <span class="n">yc</span>
    698. 

  698.         <span class="n">small_coord</span> <span class="o">=</span> <span class="n">w</span> <span class="o">-</span> <span class="p">(</span><span class="n">x2</span> <span class="o">-</span> <span class="n">x1</span><span class="p">),</span> <span class="n">h</span> <span class="o">-</span> <span class="p">(</span><span class="n">y2</span> <span class="o">-</span> <span class="n">y1</span><span class="p">),</span> <span class="n">w</span><span class="p">,</span> <span class="n">h</span>
    699. 

  699.     <span class="c1"># index1 to top right part of image</span>
    700. 

  700.     <span class="k">elif</span> <span class="n">mosaic_index</span> <span class="o">==</span> <span class="mi">1</span><span class="p">:</span>
    701. 

  701.         <span class="n">x1</span><span class="p">,</span> <span class="n">y1</span><span class="p">,</span> <span class="n">x2</span><span class="p">,</span> <span class="n">y2</span> <span class="o">=</span> <span class="n">xc</span><span class="p">,</span> <span class="nb">max</span><span class="p">(</span><span class="n">yc</span> <span class="o">-</span> <span class="n">h</span><span class="p">,</span> <span class="mi">0</span><span class="p">),</span> <span class="nb">min</span><span class="p">(</span><span class="n">xc</span> <span class="o">+</span> <span class="n">w</span><span class="p">,</span> <span class="n">input_w</span> <span class="o">*</span> <span class="mi">2</span><span class="p">),</span> <span class="n">yc</span>
    702. 

  702.         <span class="n">small_coord</span> <span class="o">=</span> <span class="mi">0</span><span class="p">,</span> <span class="n">h</span> <span class="o">-</span> <span class="p">(</span><span class="n">y2</span> <span class="o">-</span> <span class="n">y1</span><span class="p">),</span> <span class="nb">min</span><span class="p">(</span><span class="n">w</span><span class="p">,</span> <span class="n">x2</span> <span class="o">-</span> <span class="n">x1</span><span class="p">),</span> <span class="n">h</span>
    703. 

  703.     <span class="c1"># index2 to bottom left part of image</span>
    704. 

  704.     <span class="k">elif</span> <span class="n">mosaic_index</span> <span class="o">==</span> <span class="mi">2</span><span class="p">:</span>
    705. 

  705.         <span class="n">x1</span><span class="p">,</span> <span class="n">y1</span><span class="p">,</span> <span class="n">x2</span><span class="p">,</span> <span class="n">y2</span> <span class="o">=</span> <span class="nb">max</span><span class="p">(</span><span class="n">xc</span> <span class="o">-</span> <span class="n">w</span><span class="p">,</span> <span class="mi">0</span><span class="p">),</span> <span class="n">yc</span><span class="p">,</span> <span class="n">xc</span><span class="p">,</span> <span class="nb">min</span><span class="p">(</span><span class="n">input_h</span> <span class="o">*</span> <span class="mi">2</span><span class="p">,</span> <span class="n">yc</span> <span class="o">+</span> <span class="n">h</span><span class="p">)</span>
    706. 

  706.         <span class="n">small_coord</span> <span class="o">=</span> <span class="n">w</span> <span class="o">-</span> <span class="p">(</span><span class="n">x2</span> <span class="o">-</span> <span class="n">x1</span><span class="p">),</span> <span class="mi">0</span><span class="p">,</span> <span class="n">w</span><span class="p">,</span> <span class="nb">min</span><span class="p">(</span><span class="n">y2</span> <span class="o">-</span> <span class="n">y1</span><span class="p">,</span> <span class="n">h</span><span class="p">)</span>
    707. 

  707.     <span class="c1"># index2 to bottom right part of image</span>
    708. 

  708.     <span class="k">elif</span> <span class="n">mosaic_index</span> <span class="o">==</span> <span class="mi">3</span><span class="p">:</span>
    709. 

  709.         <span class="n">x1</span><span class="p">,</span> <span class="n">y1</span><span class="p">,</span> <span class="n">x2</span><span class="p">,</span> <span class="n">y2</span> <span class="o">=</span> <span class="n">xc</span><span class="p">,</span> <span class="n">yc</span><span class="p">,</span> <span class="nb">min</span><span class="p">(</span><span class="n">xc</span> <span class="o">+</span> <span class="n">w</span><span class="p">,</span> <span class="n">input_w</span> <span class="o">*</span> <span class="mi">2</span><span class="p">),</span> <span class="nb">min</span><span class="p">(</span><span class="n">input_h</span> <span class="o">*</span> <span class="mi">2</span><span class="p">,</span> <span class="n">yc</span> <span class="o">+</span> <span class="n">h</span><span class="p">)</span>  <span class="c1"># noqa</span>
    710. 

  710.         <span class="n">small_coord</span> <span class="o">=</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="nb">min</span><span class="p">(</span><span class="n">w</span><span class="p">,</span> <span class="n">x2</span> <span class="o">-</span> <span class="n">x1</span><span class="p">),</span> <span class="nb">min</span><span class="p">(</span><span class="n">y2</span> <span class="o">-</span> <span class="n">y1</span><span class="p">,</span> <span class="n">h</span><span class="p">)</span>
    711. 

  711.     <span class="k">return</span> <span class="p">(</span><span class="n">x1</span><span class="p">,</span> <span class="n">y1</span><span class="p">,</span> <span class="n">x2</span><span class="p">,</span> <span class="n">y2</span><span class="p">),</span> <span class="n">small_coord</span></div>
    712. 

  712. 

  713. 

  714. <div class="viewcode-block" id="adjust_box_anns"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.adjust_box_anns">[docs]</a><span class="k">def</span> <span class="nf">adjust_box_anns</span><span class="p">(</span><span class="n">bbox</span><span class="p">,</span> <span class="n">scale_ratio</span><span class="p">,</span> <span class="n">padw</span><span class="p">,</span> <span class="n">padh</span><span class="p">,</span> <span class="n">w_max</span><span class="p">,</span> <span class="n">h_max</span><span class="p">):</span>
    715. 

  715.     <span class="sd">&quot;&quot;&quot;</span>
    716. 

  716. <span class="sd">    Adjusts the bbox annotations of rescaled, padded image.</span>
    717. 

  717. 

  718. <span class="sd">    :param bbox: (np.array) bbox to modify.</span>
    719. 

  719. <span class="sd">    :param scale_ratio: (float) scale ratio between rescale output image and original one.</span>
    720. 

  720. <span class="sd">    :param padw: (int) width padding size.</span>
    721. 

  721. <span class="sd">    :param padh: (int) height padding size.</span>
    722. 

  722. <span class="sd">    :param w_max: (int) width border.</span>
    723. 

  723. <span class="sd">    :param h_max: (int) height border</span>
    724. 

  724. <span class="sd">    :return: modified bbox (np.array)</span>
    725. 

  725. <span class="sd">    &quot;&quot;&quot;</span>
    726. 

  726.     <span class="n">bbox</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">::</span><span class="mi">2</span><span class="p">]</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">clip</span><span class="p">(</span><span class="n">bbox</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">::</span><span class="mi">2</span><span class="p">]</span> <span class="o">*</span> <span class="n">scale_ratio</span> <span class="o">+</span> <span class="n">padw</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="n">w_max</span><span class="p">)</span>
    727. 

  727.     <span class="n">bbox</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">::</span><span class="mi">2</span><span class="p">]</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">clip</span><span class="p">(</span><span class="n">bbox</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">::</span><span class="mi">2</span><span class="p">]</span> <span class="o">*</span> <span class="n">scale_ratio</span> <span class="o">+</span> <span class="n">padh</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="n">h_max</span><span class="p">)</span>
    728. 

  728.     <span class="k">return</span> <span class="n">bbox</span></div>
    729. 

  729. 

  730. 

  731. <div class="viewcode-block" id="DetectionCollateFN"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.DetectionCollateFN">[docs]</a><span class="k">class</span> <span class="nc">DetectionCollateFN</span><span class="p">:</span>
    732. 

  732.     <span class="sd">&quot;&quot;&quot;</span>
    733. 

  733. <span class="sd">    Collate function for Yolox training</span>
    734. 

  734. <span class="sd">    &quot;&quot;&quot;</span>
    735. 

  735.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">data</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="n">Tuple</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">]:</span>
    736. 

  736.         <span class="n">batch</span> <span class="o">=</span> <span class="n">default_collate</span><span class="p">(</span><span class="n">data</span><span class="p">)</span>
    737. 

  737.         <span class="n">ims</span><span class="p">,</span> <span class="n">targets</span> <span class="o">=</span> <span class="n">batch</span><span class="p">[</span><span class="mi">0</span><span class="p">:</span><span class="mi">2</span><span class="p">]</span>
    738. 

  738.         <span class="k">return</span> <span class="n">ims</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">_format_targets</span><span class="p">(</span><span class="n">targets</span><span class="p">)</span>
    739. 

  739. 

  740.     <span class="k">def</span> <span class="nf">_format_targets</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">targets</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">:</span>
    741. 

  741.         <span class="n">nlabel</span> <span class="o">=</span> <span class="p">(</span><span class="n">targets</span><span class="o">.</span><span class="n">sum</span><span class="p">(</span><span class="n">dim</span><span class="o">=</span><span class="mi">2</span><span class="p">)</span> <span class="o">&gt;</span> <span class="mi">0</span><span class="p">)</span><span class="o">.</span><span class="n">sum</span><span class="p">(</span><span class="n">dim</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>  <span class="c1"># number of label per image</span>
    742. 

  742.         <span class="n">targets_merged</span> <span class="o">=</span> <span class="p">[]</span>
    743. 

  743.         <span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">targets</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">]):</span>
    744. 

  744.             <span class="n">targets_im</span> <span class="o">=</span> <span class="n">targets</span><span class="p">[</span><span class="n">i</span><span class="p">,</span> <span class="p">:</span><span class="n">nlabel</span><span class="p">[</span><span class="n">i</span><span class="p">]]</span>
    745. 

  745.             <span class="n">batch_column</span> <span class="o">=</span> <span class="n">targets</span><span class="o">.</span><span class="n">new_ones</span><span class="p">((</span><span class="n">targets_im</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="mi">1</span><span class="p">))</span> <span class="o">*</span> <span class="n">i</span>
    746. 

  746.             <span class="n">targets_merged</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">((</span><span class="n">batch_column</span><span class="p">,</span> <span class="n">targets_im</span><span class="p">),</span> <span class="mi">1</span><span class="p">))</span>
    747. 

  747.         <span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">(</span><span class="n">targets_merged</span><span class="p">,</span> <span class="mi">0</span><span class="p">)</span></div>
    748. 

  748. 

  749. 

  750. <div class="viewcode-block" id="CrowdDetectionCollateFN"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.CrowdDetectionCollateFN">[docs]</a><span class="k">class</span> <span class="nc">CrowdDetectionCollateFN</span><span class="p">(</span><span class="n">DetectionCollateFN</span><span class="p">):</span>
    751. 

  751.     <span class="sd">&quot;&quot;&quot;</span>
    752. 

  752. <span class="sd">    Collate function for Yolox training with additional_batch_items that includes crowd targets</span>
    753. 

  753. <span class="sd">    &quot;&quot;&quot;</span>
    754. 

  754.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">data</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="n">Tuple</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">Dict</span><span class="p">[</span><span class="nb">str</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">]]:</span>
    755. 

  755.         <span class="n">batch</span> <span class="o">=</span> <span class="n">default_collate</span><span class="p">(</span><span class="n">data</span><span class="p">)</span>
    756. 

  756.         <span class="n">ims</span><span class="p">,</span> <span class="n">targets</span><span class="p">,</span> <span class="n">crowd_targets</span> <span class="o">=</span> <span class="n">batch</span><span class="p">[</span><span class="mi">0</span><span class="p">:</span><span class="mi">3</span><span class="p">]</span>
    757. 

  757.         <span class="k">return</span> <span class="n">ims</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">_format_targets</span><span class="p">(</span><span class="n">targets</span><span class="p">),</span> <span class="p">{</span><span class="s2">&quot;crowd_targets&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">_format_targets</span><span class="p">(</span><span class="n">crowd_targets</span><span class="p">)}</span></div>
    758. 

  758. 

  759. 

  760. <div class="viewcode-block" id="compute_box_area"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.compute_box_area">[docs]</a><span class="k">def</span> <span class="nf">compute_box_area</span><span class="p">(</span><span class="n">box</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">:</span>
    761. 

  761.     <span class="sd">&quot;&quot;&quot;Compute the area of one or many boxes.</span>
    762. 

  762. <span class="sd">         :param box: One or many boxes, shape = (4, ?), each box in format (x1, y1, x2, y2)</span>
    763. 

  763. <span class="sd">    Returns:</span>
    764. 

  764. <span class="sd">        Area of every box, shape = (1, ?)</span>
    765. 

  765. <span class="sd">     &quot;&quot;&quot;</span>
    766. 

  766.     <span class="c1"># box = 4xn</span>
    767. 

  767.     <span class="k">return</span> <span class="p">(</span><span class="n">box</span><span class="p">[</span><span class="mi">2</span><span class="p">]</span> <span class="o">-</span> <span class="n">box</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span> <span class="o">*</span> <span class="p">(</span><span class="n">box</span><span class="p">[</span><span class="mi">3</span><span class="p">]</span> <span class="o">-</span> <span class="n">box</span><span class="p">[</span><span class="mi">1</span><span class="p">])</span></div>
    768. 

  768. 

  769. 

  770. <div class="viewcode-block" id="crowd_ioa"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.crowd_ioa">[docs]</a><span class="k">def</span> <span class="nf">crowd_ioa</span><span class="p">(</span><span class="n">det_box</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">crowd_box</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">:</span>
    771. 

  771.     <span class="sd">&quot;&quot;&quot;</span>
    772. 

  772. <span class="sd">    Return intersection-over-detection_area of boxes, used for crowd ground truths.</span>
    773. 

  773. <span class="sd">    Both sets of boxes are expected to be in (x1, y1, x2, y2) format.</span>
    774. 

  774. <span class="sd">    Arguments:</span>
    775. 

  775. <span class="sd">        det_box (Tensor[N, 4])</span>
    776. 

  776. <span class="sd">        crowd_box (Tensor[M, 4])</span>
    777. 

  777. <span class="sd">    Returns:</span>
    778. 

  778. <span class="sd">        crowd_ioa (Tensor[N, M]): the NxM matrix containing the pairwise</span>
    779. 

  779. <span class="sd">            IoA values for every element in det_box and crowd_box</span>
    780. 

  780. <span class="sd">    &quot;&quot;&quot;</span>
    781. 

  781.     <span class="n">det_area</span> <span class="o">=</span> <span class="n">compute_box_area</span><span class="p">(</span><span class="n">det_box</span><span class="o">.</span><span class="n">T</span><span class="p">)</span>
    782. 

  782. 

  783.     <span class="c1"># inter(N,M) = (rb(N,M,2) - lt(N,M,2)).clamp(0).prod(2)</span>
    784. 

  784.     <span class="n">inter</span> <span class="o">=</span> <span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">min</span><span class="p">(</span><span class="n">det_box</span><span class="p">[:,</span> <span class="kc">None</span><span class="p">,</span> <span class="mi">2</span><span class="p">:],</span> <span class="n">crowd_box</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">:])</span> <span class="o">-</span> <span class="n">torch</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="n">det_box</span><span class="p">[:,</span> <span class="kc">None</span><span class="p">,</span> <span class="p">:</span><span class="mi">2</span><span class="p">],</span> <span class="n">crowd_box</span><span class="p">[:,</span> <span class="p">:</span><span class="mi">2</span><span class="p">]))</span> \
    785. 

  785.         <span class="o">.</span><span class="n">clamp</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span><span class="o">.</span><span class="n">prod</span><span class="p">(</span><span class="mi">2</span><span class="p">)</span>
    786. 

  786.     <span class="k">return</span> <span class="n">inter</span> <span class="o">/</span> <span class="n">det_area</span><span class="p">[:,</span> <span class="kc">None</span><span class="p">]</span>  <span class="c1"># crowd_ioa = inter / det_area</span></div>
    787. 

  787. 

  788. 

  789. <div class="viewcode-block" id="compute_detection_matching"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.compute_detection_matching">[docs]</a><span class="k">def</span> <span class="nf">compute_detection_matching</span><span class="p">(</span>
    790. 

  790.     <span class="n">output</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    791. 

  791.     <span class="n">targets</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    792. 

  792.     <span class="n">height</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
    793. 

  793.     <span class="n">width</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
    794. 

  794.     <span class="n">iou_thresholds</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    795. 

  795.     <span class="n">denormalize_targets</span><span class="p">:</span> <span class="nb">bool</span><span class="p">,</span>
    796. 

  796.     <span class="n">device</span><span class="p">:</span> <span class="nb">str</span><span class="p">,</span>
    797. 

  797.     <span class="n">crowd_targets</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
    798. 

  798.     <span class="n">top_k</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">100</span><span class="p">,</span>
    799. 

  799.     <span class="n">return_on_cpu</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">True</span><span class="p">,</span>
    800. 

  800. <span class="p">)</span> <span class="o">-&gt;</span> <span class="n">List</span><span class="p">[</span><span class="n">Tuple</span><span class="p">]:</span>
    801. 

  801.     <span class="sd">&quot;&quot;&quot;</span>
    802. 

  802. <span class="sd">    Match predictions (NMS output) and the targets (ground truth) with respect to IoU and confidence score.</span>
    803. 

  803. <span class="sd">    :param output:          list (of length batch_size) of Tensors of shape (num_predictions, 6)</span>
    804. 

  804. <span class="sd">                            format:     (x1, y1, x2, y2, confidence, class_label) where x1,y1,x2,y2 are according to image size</span>
    805. 

  805. <span class="sd">    :param targets:         targets for all images of shape (total_num_targets, 6)</span>
    806. 

  806. <span class="sd">                            format:     (index, x, y, w, h, label) where x,y,w,h are in range [0,1]</span>
    807. 

  807. <span class="sd">    :param height:          dimensions of the image</span>
    808. 

  808. <span class="sd">    :param width:           dimensions of the image</span>
    809. 

  809. <span class="sd">    :param iou_thresholds:  Threshold to compute the mAP</span>
    810. 

  810. <span class="sd">    :param device:          Device</span>
    811. 

  811. <span class="sd">    :param crowd_targets:   crowd targets for all images of shape (total_num_crowd_targets, 6)</span>
    812. 

  812. <span class="sd">                            format:     (index, x, y, w, h, label) where x,y,w,h are in range [0,1]</span>
    813. 

  813. <span class="sd">    :param top_k:           Number of predictions to keep per class, ordered by confidence score</span>
    814. 

  814. <span class="sd">    :param denormalize_targets: If True, denormalize the targets and crowd_targets</span>
    815. 

  815. <span class="sd">    :param return_on_cpu:   If True, the output will be returned on &quot;CPU&quot;, otherwise it will be returned on &quot;device&quot;</span>
    816. 

  816. 

  817. <span class="sd">    :return:                list of the following tensors, for every image:</span>
    818. 

  818. <span class="sd">        :preds_matched:     Tensor of shape (num_img_predictions, n_iou_thresholds)</span>
    819. 

  819. <span class="sd">                            True when prediction (i) is matched with a target with respect to the (j)th IoU threshold</span>
    820. 

  820. <span class="sd">        :preds_to_ignore:   Tensor of shape (num_img_predictions, n_iou_thresholds)</span>
    821. 

  821. <span class="sd">                            True when prediction (i) is matched with a crowd target with respect to the (j)th IoU threshold</span>
    822. 

  822. <span class="sd">        :preds_scores:      Tensor of shape (num_img_predictions), confidence score for every prediction</span>
    823. 

  823. <span class="sd">        :preds_cls:         Tensor of shape (num_img_predictions), predicted class for every prediction</span>
    824. 

  824. <span class="sd">        :targets_cls:       Tensor of shape (num_img_targets), ground truth class for every target</span>
    825. 

  825. <span class="sd">    &quot;&quot;&quot;</span>
    826. 

  826.     <span class="n">output</span> <span class="o">=</span> <span class="nb">map</span><span class="p">(</span><span class="k">lambda</span> <span class="n">tensor</span><span class="p">:</span> <span class="kc">None</span> <span class="k">if</span> <span class="n">tensor</span> <span class="ow">is</span> <span class="kc">None</span> <span class="k">else</span> <span class="n">tensor</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">),</span> <span class="n">output</span><span class="p">)</span>
    827. 

  827.     <span class="n">targets</span><span class="p">,</span> <span class="n">iou_thresholds</span> <span class="o">=</span> <span class="n">targets</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">),</span> <span class="n">iou_thresholds</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">)</span>
    828. 

  828. 

  829.     <span class="c1"># If crowd_targets is not provided, we patch it with an empty tensor</span>
    830. 

  830.     <span class="n">crowd_targets</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span><span class="n">size</span><span class="o">=</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">6</span><span class="p">),</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span> <span class="k">if</span> <span class="n">crowd_targets</span> <span class="ow">is</span> <span class="kc">None</span> <span class="k">else</span> <span class="n">crowd_targets</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">)</span>
    831. 

  831. 

  832.     <span class="n">batch_metrics</span> <span class="o">=</span> <span class="p">[]</span>
    833. 

  833.     <span class="k">for</span> <span class="n">img_i</span><span class="p">,</span> <span class="n">img_preds</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="n">output</span><span class="p">):</span>
    834. 

  834.         <span class="c1"># If img_preds is None (not prediction for this image), we patch it with an empty tensor</span>
    835. 

  835.         <span class="n">img_preds</span> <span class="o">=</span> <span class="n">img_preds</span> <span class="k">if</span> <span class="n">img_preds</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span> <span class="k">else</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span><span class="n">size</span><span class="o">=</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">6</span><span class="p">),</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
    836. 

  836.         <span class="n">img_targets</span> <span class="o">=</span> <span class="n">targets</span><span class="p">[</span><span class="n">targets</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">==</span> <span class="n">img_i</span><span class="p">,</span> <span class="mi">1</span><span class="p">:]</span>
    837. 

  837.         <span class="n">img_crowd_targets</span> <span class="o">=</span> <span class="n">crowd_targets</span><span class="p">[</span><span class="n">crowd_targets</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">==</span> <span class="n">img_i</span><span class="p">,</span> <span class="mi">1</span><span class="p">:]</span>
    838. 

  838. 

  839.         <span class="n">img_matching_tensors</span> <span class="o">=</span> <span class="n">compute_img_detection_matching</span><span class="p">(</span>
    840. 

  840.             <span class="n">preds</span><span class="o">=</span><span class="n">img_preds</span><span class="p">,</span>
    841. 

  841.             <span class="n">targets</span><span class="o">=</span><span class="n">img_targets</span><span class="p">,</span>
    842. 

  842.             <span class="n">crowd_targets</span><span class="o">=</span><span class="n">img_crowd_targets</span><span class="p">,</span>
    843. 

  843.             <span class="n">denormalize_targets</span><span class="o">=</span><span class="n">denormalize_targets</span><span class="p">,</span>
    844. 

  844.             <span class="n">height</span><span class="o">=</span><span class="n">height</span><span class="p">,</span>
    845. 

  845.             <span class="n">width</span><span class="o">=</span><span class="n">width</span><span class="p">,</span>
    846. 

  846.             <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">,</span>
    847. 

  847.             <span class="n">iou_thresholds</span><span class="o">=</span><span class="n">iou_thresholds</span><span class="p">,</span>
    848. 

  848.             <span class="n">top_k</span><span class="o">=</span><span class="n">top_k</span><span class="p">,</span>
    849. 

  849.             <span class="n">return_on_cpu</span><span class="o">=</span><span class="n">return_on_cpu</span>
    850. 

  850.         <span class="p">)</span>
    851. 

  851.         <span class="n">batch_metrics</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">img_matching_tensors</span><span class="p">)</span>
    852. 

  852. 

  853.     <span class="k">return</span> <span class="n">batch_metrics</span></div>
    854. 

  854. 

  855. 

  856. <div class="viewcode-block" id="compute_img_detection_matching"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.compute_img_detection_matching">[docs]</a><span class="k">def</span> <span class="nf">compute_img_detection_matching</span><span class="p">(</span>
    857. 

  857.         <span class="n">preds</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    858. 

  858.         <span class="n">targets</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    859. 

  859.         <span class="n">crowd_targets</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    860. 

  860.         <span class="n">height</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
    861. 

  861.         <span class="n">width</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
    862. 

  862.         <span class="n">iou_thresholds</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    863. 

  863.         <span class="n">device</span><span class="p">:</span> <span class="nb">str</span><span class="p">,</span>
    864. 

  864.         <span class="n">denormalize_targets</span><span class="p">:</span> <span class="nb">bool</span><span class="p">,</span>
    865. 

  865.         <span class="n">top_k</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">100</span><span class="p">,</span>
    866. 

  866.         <span class="n">return_on_cpu</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">True</span>
    867. 

  867. <span class="p">)</span> <span class="o">-&gt;</span> <span class="n">Tuple</span><span class="p">:</span>
    868. 

  868.     <span class="sd">&quot;&quot;&quot;</span>
    869. 

  869. <span class="sd">    Match predictions (NMS output) and the targets (ground truth) with respect to IoU and confidence score</span>
    870. 

  870. <span class="sd">    for a given image.</span>
    871. 

  871. <span class="sd">    :param preds:           Tensor of shape (num_img_predictions, 6)</span>
    872. 

  872. <span class="sd">                            format:     (x1, y1, x2, y2, confidence, class_label) where x1,y1,x2,y2 are according to image size</span>
    873. 

  873. <span class="sd">    :param targets:         targets for this image of shape (num_img_targets, 6)</span>
    874. 

  874. <span class="sd">                            format:     (index, x, y, w, h, label) where x,y,w,h are in range [0,1]</span>
    875. 

  875. <span class="sd">    :param height:          dimensions of the image</span>
    876. 

  876. <span class="sd">    :param width:           dimensions of the image</span>
    877. 

  877. <span class="sd">    :param iou_thresholds:  Threshold to compute the mAP</span>
    878. 

  878. <span class="sd">    :param device:</span>
    879. 

  879. <span class="sd">    :param crowd_targets:   crowd targets for all images of shape (total_num_crowd_targets, 6)</span>
    880. 

  880. <span class="sd">                            format:     (index, x, y, w, h, label) where x,y,w,h are in range [0,1]</span>
    881. 

  881. <span class="sd">    :param top_k:           Number of predictions to keep per class, ordered by confidence score</span>
    882. 

  882. <span class="sd">    :param device:          Device</span>
    883. 

  883. <span class="sd">    :param denormalize_targets: If True, denormalize the targets and crowd_targets</span>
    884. 

  884. <span class="sd">    :param return_on_cpu:   If True, the output will be returned on &quot;CPU&quot;, otherwise it will be returned on &quot;device&quot;</span>
    885. 

  885. 

  886. <span class="sd">    :return:</span>
    887. 

  887. <span class="sd">        :preds_matched:     Tensor of shape (num_img_predictions, n_iou_thresholds)</span>
    888. 

  888. <span class="sd">                                True when prediction (i) is matched with a target with respect to the (j)th IoU threshold</span>
    889. 

  889. <span class="sd">        :preds_to_ignore:   Tensor of shape (num_img_predictions, n_iou_thresholds)</span>
    890. 

  890. <span class="sd">                                True when prediction (i) is matched with a crowd target with respect to the (j)th IoU threshold</span>
    891. 

  891. <span class="sd">        :preds_scores:      Tensor of shape (num_img_predictions), confidence score for every prediction</span>
    892. 

  892. <span class="sd">        :preds_cls:         Tensor of shape (num_img_predictions), predicted class for every prediction</span>
    893. 

  893. <span class="sd">        :targets_cls:       Tensor of shape (num_img_targets), ground truth class for every target</span>
    894. 

  894. <span class="sd">    &quot;&quot;&quot;</span>
    895. 

  895.     <span class="n">num_iou_thresholds</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="n">iou_thresholds</span><span class="p">)</span>
    896. 

  896. 

  897.     <span class="k">if</span> <span class="n">preds</span> <span class="ow">is</span> <span class="kc">None</span> <span class="ow">or</span> <span class="nb">len</span><span class="p">(</span><span class="n">preds</span><span class="p">)</span> <span class="o">==</span> <span class="mi">0</span><span class="p">:</span>
    898. 

  898.         <span class="k">if</span> <span class="n">return_on_cpu</span><span class="p">:</span>
    899. 

  899.             <span class="n">device</span> <span class="o">=</span> <span class="s2">&quot;cpu&quot;</span>
    900. 

  900.         <span class="n">preds_matched</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">((</span><span class="mi">0</span><span class="p">,</span> <span class="n">num_iou_thresholds</span><span class="p">),</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">bool</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
    901. 

  901.         <span class="n">preds_to_ignore</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">((</span><span class="mi">0</span><span class="p">,</span> <span class="n">num_iou_thresholds</span><span class="p">),</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">bool</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
    902. 

  902.         <span class="n">preds_scores</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">([],</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">float32</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
    903. 

  903.         <span class="n">preds_cls</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">([],</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">float32</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
    904. 

  904.         <span class="n">targets_cls</span> <span class="o">=</span> <span class="n">targets</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
    905. 

  905.         <span class="k">return</span> <span class="n">preds_matched</span><span class="p">,</span> <span class="n">preds_to_ignore</span><span class="p">,</span> <span class="n">preds_scores</span><span class="p">,</span> <span class="n">preds_cls</span><span class="p">,</span> <span class="n">targets_cls</span>
    906. 

  906. 

  907.     <span class="n">preds_matched</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">preds</span><span class="p">),</span> <span class="n">num_iou_thresholds</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">bool</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
    908. 

  908.     <span class="n">targets_matched</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">targets</span><span class="p">),</span> <span class="n">num_iou_thresholds</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">bool</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
    909. 

  909.     <span class="n">preds_to_ignore</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">preds</span><span class="p">),</span> <span class="n">num_iou_thresholds</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">bool</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
    910. 

  910. 

  911.     <span class="n">preds_cls</span><span class="p">,</span> <span class="n">preds_box</span><span class="p">,</span> <span class="n">preds_scores</span> <span class="o">=</span> <span class="n">preds</span><span class="p">[:,</span> <span class="o">-</span><span class="mi">1</span><span class="p">],</span> <span class="n">preds</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">:</span><span class="mi">4</span><span class="p">],</span> <span class="n">preds</span><span class="p">[:,</span> <span class="mi">4</span><span class="p">]</span>
    912. 

  912.     <span class="n">targets_cls</span><span class="p">,</span> <span class="n">targets_box</span> <span class="o">=</span> <span class="n">targets</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">],</span> <span class="n">targets</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">:</span><span class="mi">5</span><span class="p">]</span>
    913. 

  913.     <span class="n">crowd_targets_cls</span><span class="p">,</span> <span class="n">crowd_target_box</span> <span class="o">=</span> <span class="n">crowd_targets</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">],</span> <span class="n">crowd_targets</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">:</span><span class="mi">5</span><span class="p">]</span>
    914. 

  914. 

  915.     <span class="c1"># Ignore all but the predictions that were top_k for their class</span>
    916. 

  916.     <span class="n">preds_idx_to_use</span> <span class="o">=</span> <span class="n">get_top_k_idx_per_cls</span><span class="p">(</span><span class="n">preds_scores</span><span class="p">,</span> <span class="n">preds_cls</span><span class="p">,</span> <span class="n">top_k</span><span class="p">)</span>
    917. 

  917.     <span class="n">preds_to_ignore</span><span class="p">[:,</span> <span class="p">:]</span> <span class="o">=</span> <span class="kc">True</span>
    918. 

  918.     <span class="n">preds_to_ignore</span><span class="p">[</span><span class="n">preds_idx_to_use</span><span class="p">]</span> <span class="o">=</span> <span class="kc">False</span>
    919. 

  919. 

  920.     <span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="n">targets</span><span class="p">)</span> <span class="o">&gt;</span> <span class="mi">0</span> <span class="ow">or</span> <span class="nb">len</span><span class="p">(</span><span class="n">crowd_targets</span><span class="p">)</span> <span class="o">&gt;</span> <span class="mi">0</span><span class="p">:</span>
    921. 

  921. 

  922.         <span class="c1"># CHANGE bboxes TO FIT THE IMAGE SIZE</span>
    923. 

  923.         <span class="n">change_bbox_bounds_for_image_size</span><span class="p">(</span><span class="n">preds</span><span class="p">,</span> <span class="p">(</span><span class="n">height</span><span class="p">,</span> <span class="n">width</span><span class="p">))</span>
    924. 

  924. 

  925.         <span class="c1"># if target_format == &quot;xywh&quot;:</span>
    926. 

  926.         <span class="n">targets_box</span> <span class="o">=</span> <span class="n">convert_xywh_bbox_to_xyxy</span><span class="p">(</span><span class="n">targets_box</span><span class="p">)</span>  <span class="c1"># cxcywh2xyxy</span>
    927. 

  927.         <span class="n">crowd_target_box</span> <span class="o">=</span> <span class="n">convert_xywh_bbox_to_xyxy</span><span class="p">(</span><span class="n">crowd_target_box</span><span class="p">)</span>  <span class="c1"># convert_xywh_bbox_to_xyxy</span>
    928. 

  928. 

  929.         <span class="k">if</span> <span class="n">denormalize_targets</span><span class="p">:</span>
    930. 

  930.             <span class="n">targets_box</span><span class="p">[:,</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">]]</span> <span class="o">*=</span> <span class="n">width</span>
    931. 

  931.             <span class="n">targets_box</span><span class="p">[:,</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">]]</span> <span class="o">*=</span> <span class="n">height</span>
    932. 

  932.             <span class="n">crowd_target_box</span><span class="p">[:,</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">]]</span> <span class="o">*=</span> <span class="n">width</span>
    933. 

  933.             <span class="n">crowd_target_box</span><span class="p">[:,</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">]]</span> <span class="o">*=</span> <span class="n">height</span>
    934. 

  934. 

  935.     <span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="n">targets</span><span class="p">)</span> <span class="o">&gt;</span> <span class="mi">0</span><span class="p">:</span>
    936. 

  936. 

  937.         <span class="c1"># shape = (n_preds x n_targets)</span>
    938. 

  938.         <span class="n">iou</span> <span class="o">=</span> <span class="n">box_iou</span><span class="p">(</span><span class="n">preds_box</span><span class="p">[</span><span class="n">preds_idx_to_use</span><span class="p">],</span> <span class="n">targets_box</span><span class="p">)</span>
    939. 

  939. 

  940.         <span class="c1"># Fill IoU values at index (i, j) with 0 when the prediction (i) and target(j) are of different class</span>
    941. 

  941.         <span class="c1"># Filling with 0 is equivalent to ignore these values since with want IoU &gt; iou_threshold &gt; 0</span>
    942. 

  942.         <span class="n">cls_mismatch</span> <span class="o">=</span> <span class="p">(</span><span class="n">preds_cls</span><span class="p">[</span><span class="n">preds_idx_to_use</span><span class="p">]</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span> <span class="o">!=</span> <span class="n">targets_cls</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">))</span>
    943. 

  943.         <span class="n">iou</span><span class="p">[</span><span class="n">cls_mismatch</span><span class="p">]</span> <span class="o">=</span> <span class="mi">0</span>
    944. 

  944. 

  945.         <span class="c1"># The matching priority is first detection confidence and then IoU value.</span>
    946. 

  946.         <span class="c1"># The detection is already sorted by confidence in NMS, so here for each prediction we order the targets by iou.</span>
    947. 

  947.         <span class="n">sorted_iou</span><span class="p">,</span> <span class="n">target_sorted</span> <span class="o">=</span> <span class="n">iou</span><span class="o">.</span><span class="n">sort</span><span class="p">(</span><span class="n">descending</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="n">stable</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
    948. 

  948. 

  949.         <span class="c1"># Only iterate over IoU values higher than min threshold to speed up the process</span>
    950. 

  950.         <span class="k">for</span> <span class="n">pred_selected_i</span><span class="p">,</span> <span class="n">target_sorted_i</span> <span class="ow">in</span> <span class="p">(</span><span class="n">sorted_iou</span> <span class="o">&gt;</span> <span class="n">iou_thresholds</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span><span class="o">.</span><span class="n">nonzero</span><span class="p">(</span><span class="n">as_tuple</span><span class="o">=</span><span class="kc">False</span><span class="p">):</span>
    951. 

  951. 

  952.             <span class="c1"># pred_selected_i and target_sorted_i are relative to filters/sorting, so we extract their absolute indexes</span>
    953. 

  953.             <span class="n">pred_i</span> <span class="o">=</span> <span class="n">preds_idx_to_use</span><span class="p">[</span><span class="n">pred_selected_i</span><span class="p">]</span>
    954. 

  954.             <span class="n">target_i</span> <span class="o">=</span> <span class="n">target_sorted</span><span class="p">[</span><span class="n">pred_selected_i</span><span class="p">,</span> <span class="n">target_sorted_i</span><span class="p">]</span>
    955. 

  955. 

  956.             <span class="c1"># Vector[j], True when IoU(pred_i, target_i) is above the (j)th threshold</span>
    957. 

  957.             <span class="n">is_iou_above_threshold</span> <span class="o">=</span> <span class="n">sorted_iou</span><span class="p">[</span><span class="n">pred_selected_i</span><span class="p">,</span> <span class="n">target_sorted_i</span><span class="p">]</span> <span class="o">&gt;</span> <span class="n">iou_thresholds</span>
    958. 

  958. 

  959.             <span class="c1"># Vector[j], True when both pred_i and target_i are not matched yet for the (j)th threshold</span>
    960. 

  960.             <span class="n">are_candidates_free</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">logical_and</span><span class="p">(</span><span class="o">~</span><span class="n">preds_matched</span><span class="p">[</span><span class="n">pred_i</span><span class="p">,</span> <span class="p">:],</span> <span class="o">~</span><span class="n">targets_matched</span><span class="p">[</span><span class="n">target_i</span><span class="p">,</span> <span class="p">:])</span>
    961. 

  961. 

  962.             <span class="c1"># Vector[j], True when (pred_i, target_i) can be matched for the (j)th threshold</span>
    963. 

  963.             <span class="n">are_candidates_good</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">logical_and</span><span class="p">(</span><span class="n">is_iou_above_threshold</span><span class="p">,</span> <span class="n">are_candidates_free</span><span class="p">)</span>
    964. 

  964. 

  965.             <span class="c1"># For every threshold (j) where target_i and pred_i can be matched together ( are_candidates_good[j]==True )</span>
    966. 

  966.             <span class="c1"># fill the matching placeholders with True</span>
    967. 

  967.             <span class="n">targets_matched</span><span class="p">[</span><span class="n">target_i</span><span class="p">,</span> <span class="n">are_candidates_good</span><span class="p">]</span> <span class="o">=</span> <span class="kc">True</span>
    968. 

  968.             <span class="n">preds_matched</span><span class="p">[</span><span class="n">pred_i</span><span class="p">,</span> <span class="n">are_candidates_good</span><span class="p">]</span> <span class="o">=</span> <span class="kc">True</span>
    969. 

  969. 

  970.             <span class="c1"># When all the targets are matched with a prediction for every IoU Threshold, stop.</span>
    971. 

  971.             <span class="k">if</span> <span class="n">targets_matched</span><span class="o">.</span><span class="n">all</span><span class="p">():</span>
    972. 

  972.                 <span class="k">break</span>
    973. 

  973. 

  974.     <span class="c1"># Crowd targets can be matched with many predictions.</span>
    975. 

  975.     <span class="c1"># Therefore, for every prediction we just need to check if it has IoA large enough with any crowd target.</span>
    976. 

  976.     <span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="n">crowd_targets</span><span class="p">)</span> <span class="o">&gt;</span> <span class="mi">0</span><span class="p">:</span>
    977. 

  977. 

  978.         <span class="c1"># shape = (n_preds_to_use x n_crowd_targets)</span>
    979. 

  979.         <span class="n">ioa</span> <span class="o">=</span> <span class="n">crowd_ioa</span><span class="p">(</span><span class="n">preds_box</span><span class="p">[</span><span class="n">preds_idx_to_use</span><span class="p">],</span> <span class="n">crowd_target_box</span><span class="p">)</span>
    980. 

  980. 

  981.         <span class="c1"># Fill IoA values at index (i, j) with 0 when the prediction (i) and target(j) are of different class</span>
    982. 

  982.         <span class="c1"># Filling with 0 is equivalent to ignore these values since with want IoA &gt; threshold &gt; 0</span>
    983. 

  983.         <span class="n">cls_mismatch</span> <span class="o">=</span> <span class="p">(</span><span class="n">preds_cls</span><span class="p">[</span><span class="n">preds_idx_to_use</span><span class="p">]</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span> <span class="o">!=</span> <span class="n">crowd_targets_cls</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">))</span>
    984. 

  984.         <span class="n">ioa</span><span class="p">[</span><span class="n">cls_mismatch</span><span class="p">]</span> <span class="o">=</span> <span class="mi">0</span>
    985. 

  985. 

  986.         <span class="c1"># For each prediction, we keep it&#39;s highest score with any crowd target (of same class)</span>
    987. 

  987.         <span class="c1"># shape = (n_preds_to_use)</span>
    988. 

  988.         <span class="n">best_ioa</span><span class="p">,</span> <span class="n">_</span> <span class="o">=</span> <span class="n">ioa</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="mi">1</span><span class="p">)</span>
    989. 

  989. 

  990.         <span class="c1"># If a prediction has IoA higher than threshold (with any target of same class), then there is a match</span>
    991. 

  991.         <span class="c1"># shape = (n_preds_to_use x iou_thresholds)</span>
    992. 

  992.         <span class="n">is_matching_with_crowd</span> <span class="o">=</span> <span class="p">(</span><span class="n">best_ioa</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span> <span class="o">&gt;</span> <span class="n">iou_thresholds</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">))</span>
    993. 

  993. 

  994.         <span class="n">preds_to_ignore</span><span class="p">[</span><span class="n">preds_idx_to_use</span><span class="p">]</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">logical_or</span><span class="p">(</span><span class="n">preds_to_ignore</span><span class="p">[</span><span class="n">preds_idx_to_use</span><span class="p">],</span> <span class="n">is_matching_with_crowd</span><span class="p">)</span>
    995. 

  995. 

  996.     <span class="k">if</span> <span class="n">return_on_cpu</span><span class="p">:</span>
    997. 

  997.         <span class="n">preds_matched</span> <span class="o">=</span> <span class="n">preds_matched</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="s2">&quot;cpu&quot;</span><span class="p">)</span>
    998. 

  998.         <span class="n">preds_to_ignore</span> <span class="o">=</span> <span class="n">preds_to_ignore</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="s2">&quot;cpu&quot;</span><span class="p">)</span>
    999. 

  999.         <span class="n">preds_scores</span> <span class="o">=</span> <span class="n">preds_scores</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="s2">&quot;cpu&quot;</span><span class="p">)</span>
    1000. 

  1000.         <span class="n">preds_cls</span> <span class="o">=</span> <span class="n">preds_cls</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="s2">&quot;cpu&quot;</span><span class="p">)</span>
    1001. 

  1001.         <span class="n">targets_cls</span> <span class="o">=</span> <span class="n">targets_cls</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="s2">&quot;cpu&quot;</span><span class="p">)</span>
    1002. 

  1002. 

  1003.     <span class="k">return</span> <span class="n">preds_matched</span><span class="p">,</span> <span class="n">preds_to_ignore</span><span class="p">,</span> <span class="n">preds_scores</span><span class="p">,</span> <span class="n">preds_cls</span><span class="p">,</span> <span class="n">targets_cls</span></div>
    1004. 

  1004. 

  1005. 

  1006. <div class="viewcode-block" id="get_top_k_idx_per_cls"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.get_top_k_idx_per_cls">[docs]</a><span class="k">def</span> <span class="nf">get_top_k_idx_per_cls</span><span class="p">(</span><span class="n">preds_scores</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">preds_cls</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">top_k</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span>
    1007. 

  1007.     <span class="sd">&quot;&quot;&quot;Get the indexes of all the top k predictions for every class</span>
    1008. 

  1008. 

  1009. <span class="sd">    :param preds_scores:   The confidence scores, vector of shape (n_pred)</span>
    1010. 

  1010. <span class="sd">    :param preds_cls:      The predicted class, vector of shape (n_pred)</span>
    1011. 

  1011. <span class="sd">    :param top_k:          Number of predictions to keep per class, ordered by confidence score</span>
    1012. 

  1012. 

  1013. <span class="sd">    :return top_k_idx:     Indexes of the top k predictions. length &lt;= (k * n_unique_class)</span>
    1014. 

  1014. <span class="sd">    &quot;&quot;&quot;</span>
    1015. 

  1015.     <span class="n">n_unique_cls</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="n">preds_cls</span><span class="p">)</span>
    1016. 

  1016.     <span class="n">mask</span> <span class="o">=</span> <span class="p">(</span><span class="n">preds_cls</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span> <span class="o">==</span> <span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="n">n_unique_cls</span> <span class="o">+</span> <span class="mi">1</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">preds_scores</span><span class="o">.</span><span class="n">device</span><span class="p">)</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">))</span>
    1017. 

  1017.     <span class="n">preds_scores_per_cls</span> <span class="o">=</span> <span class="n">preds_scores</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span> <span class="o">*</span> <span class="n">mask</span>
    1018. 

  1018. 

  1019.     <span class="n">sorted_scores_per_cls</span><span class="p">,</span> <span class="n">sorting_idx</span> <span class="o">=</span> <span class="n">preds_scores_per_cls</span><span class="o">.</span><span class="n">sort</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="n">descending</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
    1020. 

  1020.     <span class="n">idx_with_satisfying_scores</span> <span class="o">=</span> <span class="n">sorted_scores_per_cls</span><span class="p">[:</span><span class="n">top_k</span><span class="p">,</span> <span class="p">:]</span><span class="o">.</span><span class="n">nonzero</span><span class="p">(</span><span class="n">as_tuple</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span>
    1021. 

  1021.     <span class="n">top_k_idx</span> <span class="o">=</span> <span class="n">sorting_idx</span><span class="p">[</span><span class="n">idx_with_satisfying_scores</span><span class="o">.</span><span class="n">split</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="n">dim</span><span class="o">=</span><span class="mi">1</span><span class="p">)]</span>
    1022. 

  1022.     <span class="k">return</span> <span class="n">top_k_idx</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">)</span></div>
    1023. 

  1023. 

  1024. 

  1025. <div class="viewcode-block" id="compute_detection_metrics"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.compute_detection_metrics">[docs]</a><span class="k">def</span> <span class="nf">compute_detection_metrics</span><span class="p">(</span>
    1026. 

  1026.     <span class="n">preds_matched</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    1027. 

  1027.     <span class="n">preds_to_ignore</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    1028. 

  1028.     <span class="n">preds_scores</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    1029. 

  1029.     <span class="n">preds_cls</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    1030. 

  1030.     <span class="n">targets_cls</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    1031. 

  1031.     <span class="n">device</span><span class="p">:</span> <span class="nb">str</span><span class="p">,</span>
    1032. 

  1032.     <span class="n">recall_thresholds</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
    1033. 

  1033.     <span class="n">score_threshold</span><span class="p">:</span> <span class="n">Optional</span><span class="p">[</span><span class="nb">float</span><span class="p">]</span> <span class="o">=</span> <span class="mf">0.1</span><span class="p">,</span>
    1034. 

  1034. <span class="p">)</span> <span class="o">-&gt;</span> <span class="n">Tuple</span><span class="p">:</span>
    1035. 

  1035.     <span class="sd">&quot;&quot;&quot;</span>
    1036. 

  1036. <span class="sd">    Compute the list of precision, recall, MaP and f1 for every recall IoU threshold and for every class.</span>
    1037. 

  1037. 

  1038. <span class="sd">    :param preds_matched:      Tensor of shape (num_predictions, n_iou_thresholds)</span>
    1039. 

  1039. <span class="sd">                                    True when prediction (i) is matched with a target with respect to the (j)th IoU threshold</span>
    1040. 

  1040. <span class="sd">    :param preds_to_ignore     Tensor of shape (num_predictions, n_iou_thresholds)</span>
    1041. 

  1041. <span class="sd">                                    True when prediction (i) is matched with a crowd target with respect to the (j)th IoU threshold</span>
    1042. 

  1042. <span class="sd">    :param preds_scores:       Tensor of shape (num_predictions), confidence score for every prediction</span>
    1043. 

  1043. <span class="sd">    :param preds_cls:          Tensor of shape (num_predictions), predicted class for every prediction</span>
    1044. 

  1044. <span class="sd">    :param targets_cls:        Tensor of shape (num_targets), ground truth class for every target box to be detected</span>
    1045. 

  1045. <span class="sd">    :param recall_thresholds:   Recall thresholds used to compute MaP.</span>
    1046. 

  1046. <span class="sd">    :param score_threshold:    Minimum confidence score to consider a prediction for the computation of</span>
    1047. 

  1047. <span class="sd">                                    precision, recall and f1 (not MaP)</span>
    1048. 

  1048. <span class="sd">    :param device:             Device</span>
    1049. 

  1049. 

  1050. <span class="sd">    :return:</span>
    1051. 

  1051. <span class="sd">        :ap, precision, recall, f1: Tensors of shape (n_class, nb_iou_thrs)</span>
    1052. 

  1052. <span class="sd">        :unique_classes:            Vector with all unique target classes</span>
    1053. 

  1053. <span class="sd">    &quot;&quot;&quot;</span>
    1054. 

  1054.     <span class="n">preds_matched</span><span class="p">,</span> <span class="n">preds_to_ignore</span> <span class="o">=</span> <span class="n">preds_matched</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">),</span> <span class="n">preds_to_ignore</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">)</span>
    1055. 

  1055.     <span class="n">preds_scores</span><span class="p">,</span> <span class="n">preds_cls</span><span class="p">,</span> <span class="n">targets_cls</span> <span class="o">=</span> <span class="n">preds_scores</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">),</span> <span class="n">preds_cls</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">),</span> <span class="n">targets_cls</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">)</span>
    1056. 

  1056. 

  1057.     <span class="n">recall_thresholds</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">linspace</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">101</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span> <span class="k">if</span> <span class="n">recall_thresholds</span> <span class="ow">is</span> <span class="kc">None</span> <span class="k">else</span> <span class="n">recall_thresholds</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">)</span>
    1058. 

  1058. 

  1059.     <span class="n">unique_classes</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">unique</span><span class="p">(</span><span class="n">targets_cls</span><span class="p">)</span>
    1060. 

  1060.     <span class="n">n_class</span><span class="p">,</span> <span class="n">nb_iou_thrs</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="n">unique_classes</span><span class="p">),</span> <span class="n">preds_matched</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span>
    1061. 

  1061. 

  1062.     <span class="n">ap</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">((</span><span class="n">n_class</span><span class="p">,</span> <span class="n">nb_iou_thrs</span><span class="p">),</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
    1063. 

  1063.     <span class="n">precision</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">((</span><span class="n">n_class</span><span class="p">,</span> <span class="n">nb_iou_thrs</span><span class="p">),</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
    1064. 

  1064.     <span class="n">recall</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">((</span><span class="n">n_class</span><span class="p">,</span> <span class="n">nb_iou_thrs</span><span class="p">),</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
    1065. 

  1065. 

  1066.     <span class="k">for</span> <span class="n">cls_i</span><span class="p">,</span> <span class="bp">cls</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="n">unique_classes</span><span class="p">):</span>
    1067. 

  1067.         <span class="n">cls_preds_idx</span><span class="p">,</span> <span class="n">cls_targets_idx</span> <span class="o">=</span> <span class="p">(</span><span class="n">preds_cls</span> <span class="o">==</span> <span class="bp">cls</span><span class="p">),</span> <span class="p">(</span><span class="n">targets_cls</span> <span class="o">==</span> <span class="bp">cls</span><span class="p">)</span>
    1068. 

  1068.         <span class="n">cls_ap</span><span class="p">,</span> <span class="n">cls_precision</span><span class="p">,</span> <span class="n">cls_recall</span> <span class="o">=</span> <span class="n">compute_detection_metrics_per_cls</span><span class="p">(</span>
    1069. 

  1069.             <span class="n">preds_matched</span><span class="o">=</span><span class="n">preds_matched</span><span class="p">[</span><span class="n">cls_preds_idx</span><span class="p">],</span>
    1070. 

  1070.             <span class="n">preds_to_ignore</span><span class="o">=</span><span class="n">preds_to_ignore</span><span class="p">[</span><span class="n">cls_preds_idx</span><span class="p">],</span>
    1071. 

  1071.             <span class="n">preds_scores</span><span class="o">=</span><span class="n">preds_scores</span><span class="p">[</span><span class="n">cls_preds_idx</span><span class="p">],</span>
    1072. 

  1072.             <span class="n">n_targets</span><span class="o">=</span><span class="n">cls_targets_idx</span><span class="o">.</span><span class="n">sum</span><span class="p">(),</span>
    1073. 

  1073.             <span class="n">recall_thresholds</span><span class="o">=</span><span class="n">recall_thresholds</span><span class="p">,</span>
    1074. 

  1074.             <span class="n">score_threshold</span><span class="o">=</span><span class="n">score_threshold</span><span class="p">,</span>
    1075. 

  1075.             <span class="n">device</span><span class="o">=</span><span class="n">device</span>
    1076. 

  1076.         <span class="p">)</span>
    1077. 

  1077.         <span class="n">ap</span><span class="p">[</span><span class="n">cls_i</span><span class="p">,</span> <span class="p">:]</span> <span class="o">=</span> <span class="n">cls_ap</span>
    1078. 

  1078.         <span class="n">precision</span><span class="p">[</span><span class="n">cls_i</span><span class="p">,</span> <span class="p">:]</span> <span class="o">=</span> <span class="n">cls_precision</span>
    1079. 

  1079.         <span class="n">recall</span><span class="p">[</span><span class="n">cls_i</span><span class="p">,</span> <span class="p">:]</span> <span class="o">=</span> <span class="n">cls_recall</span>
    1080. 

  1080. 

  1081.     <span class="n">f1</span> <span class="o">=</span> <span class="mi">2</span> <span class="o">*</span> <span class="n">precision</span> <span class="o">*</span> <span class="n">recall</span> <span class="o">/</span> <span class="p">(</span><span class="n">precision</span> <span class="o">+</span> <span class="n">recall</span> <span class="o">+</span> <span class="mf">1e-16</span><span class="p">)</span>
    1082. 

  1082. 

  1083.     <span class="k">return</span> <span class="n">ap</span><span class="p">,</span> <span class="n">precision</span><span class="p">,</span> <span class="n">recall</span><span class="p">,</span> <span class="n">f1</span><span class="p">,</span> <span class="n">unique_classes</span></div>
    1084. 

  1084. 

  1085. 

  1086. <div class="viewcode-block" id="compute_detection_metrics_per_cls"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.compute_detection_metrics_per_cls">[docs]</a><span class="k">def</span> <span class="nf">compute_detection_metrics_per_cls</span><span class="p">(</span>
    1087. 

  1087.         <span class="n">preds_matched</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    1088. 

  1088.         <span class="n">preds_to_ignore</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    1089. 

  1089.         <span class="n">preds_scores</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    1090. 

  1090.         <span class="n">n_targets</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
    1091. 

  1091.         <span class="n">recall_thresholds</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span>
    1092. 

  1092.         <span class="n">score_threshold</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span>
    1093. 

  1093.         <span class="n">device</span><span class="p">:</span> <span class="nb">str</span><span class="p">,</span>
    1094. 

  1094. <span class="p">):</span>
    1095. 

  1095.     <span class="sd">&quot;&quot;&quot;</span>
    1096. 

  1096. <span class="sd">    Compute the list of precision, recall and MaP of a given class for every recall IoU threshold.</span>
    1097. 

  1097. 

  1098. <span class="sd">        :param preds_matched:      Tensor of shape (num_predictions, n_iou_thresholds)</span>
    1099. 

  1099. <span class="sd">                                        True when prediction (i) is matched with a target</span>
    1100. 

  1100. <span class="sd">                                        with respect to the(j)th IoU threshold</span>
    1101. 

  1101. <span class="sd">        :param preds_to_ignore     Tensor of shape (num_predictions, n_iou_thresholds)</span>
    1102. 

  1102. <span class="sd">                                        True when prediction (i) is matched with a crowd target</span>
    1103. 

  1103. <span class="sd">                                        with respect to the (j)th IoU threshold</span>
    1104. 

  1104. <span class="sd">        :param preds_scores:       Tensor of shape (num_predictions), confidence score for every prediction</span>
    1105. 

  1105. <span class="sd">        :param n_targets:          Number of target boxes of this class</span>
    1106. 

  1106. <span class="sd">        :param recall_thresholds:  Tensor of shape (max_n_rec_thresh) list of recall thresholds used to compute MaP</span>
    1107. 

  1107. <span class="sd">        :param score_threshold:    Minimum confidence score to consider a prediction for the computation of</span>
    1108. 

  1108. <span class="sd">                                        precision and recall (not MaP)</span>
    1109. 

  1109. <span class="sd">        :param device:             Device</span>
    1110. 

  1110. 

  1111. <span class="sd">        :return ap, precision, recall:  Tensors of shape (nb_iou_thrs)</span>
    1112. 

  1112. <span class="sd">    &quot;&quot;&quot;</span>
    1113. 

  1113.     <span class="n">nb_iou_thrs</span> <span class="o">=</span> <span class="n">preds_matched</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span>
    1114. 

  1114. 

  1115.     <span class="n">tps</span> <span class="o">=</span> <span class="n">preds_matched</span>
    1116. 

  1116.     <span class="n">fps</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">logical_and</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">logical_not</span><span class="p">(</span><span class="n">preds_matched</span><span class="p">),</span> <span class="n">torch</span><span class="o">.</span><span class="n">logical_not</span><span class="p">(</span><span class="n">preds_to_ignore</span><span class="p">))</span>
    1117. 

  1117. 

  1118.     <span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="n">tps</span><span class="p">)</span> <span class="o">==</span> <span class="mi">0</span><span class="p">:</span>
    1119. 

  1119.         <span class="k">return</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span><span class="n">nb_iou_thrs</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
    1120. 

  1120. 

  1121.     <span class="c1"># Sort by decreasing score</span>
    1122. 

  1122.     <span class="n">dtype</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">uint8</span> <span class="k">if</span> <span class="n">preds_scores</span><span class="o">.</span><span class="n">is_cuda</span> <span class="ow">and</span> <span class="n">preds_scores</span><span class="o">.</span><span class="n">dtype</span> <span class="ow">is</span> <span class="n">torch</span><span class="o">.</span><span class="n">bool</span> <span class="k">else</span> <span class="n">preds_scores</span><span class="o">.</span><span class="n">dtype</span>
    1123. 

  1123.     <span class="n">sort_ind</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">argsort</span><span class="p">(</span><span class="n">preds_scores</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">dtype</span><span class="p">),</span> <span class="n">descending</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
    1124. 

  1124.     <span class="n">tps</span> <span class="o">=</span> <span class="n">tps</span><span class="p">[</span><span class="n">sort_ind</span><span class="p">,</span> <span class="p">:]</span>
    1125. 

  1125.     <span class="n">fps</span> <span class="o">=</span> <span class="n">fps</span><span class="p">[</span><span class="n">sort_ind</span><span class="p">,</span> <span class="p">:]</span>
    1126. 

  1126.     <span class="n">preds_scores</span> <span class="o">=</span> <span class="n">preds_scores</span><span class="p">[</span><span class="n">sort_ind</span><span class="p">]</span>
    1127. 

  1127. 

  1128.     <span class="c1"># Rolling sum over the predictions</span>
    1129. 

  1129.     <span class="n">rolling_tps</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cumsum</span><span class="p">(</span><span class="n">tps</span><span class="p">,</span> <span class="n">axis</span><span class="o">=</span><span class="mi">0</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">float</span><span class="p">)</span>
    1130. 

  1130.     <span class="n">rolling_fps</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cumsum</span><span class="p">(</span><span class="n">fps</span><span class="p">,</span> <span class="n">axis</span><span class="o">=</span><span class="mi">0</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">float</span><span class="p">)</span>
    1131. 

  1131. 

  1132.     <span class="n">rolling_recalls</span> <span class="o">=</span> <span class="n">rolling_tps</span> <span class="o">/</span> <span class="n">n_targets</span>
    1133. 

  1133.     <span class="n">rolling_precisions</span> <span class="o">=</span> <span class="n">rolling_tps</span> <span class="o">/</span> <span class="p">(</span><span class="n">rolling_tps</span> <span class="o">+</span> <span class="n">rolling_fps</span> <span class="o">+</span> <span class="n">torch</span><span class="o">.</span><span class="n">finfo</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">float64</span><span class="p">)</span><span class="o">.</span><span class="n">eps</span><span class="p">)</span>
    1134. 

  1134. 

  1135.     <span class="c1"># Reversed cummax to only have decreasing values</span>
    1136. 

  1136.     <span class="n">rolling_precisions</span> <span class="o">=</span> <span class="n">rolling_precisions</span><span class="o">.</span><span class="n">flip</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span><span class="o">.</span><span class="n">cummax</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span><span class="o">.</span><span class="n">values</span><span class="o">.</span><span class="n">flip</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span>
    1137. 

  1137. 

  1138.     <span class="c1"># ==================</span>
    1139. 

  1139.     <span class="c1"># RECALL &amp; PRECISION</span>
    1140. 

  1140. 

  1141.     <span class="c1"># We want the rolling precision/recall at index i so that: preds_scores[i-1] &gt;= score_threshold &gt; preds_scores[i]</span>
    1142. 

  1142.     <span class="c1"># Note: torch.searchsorted works on increasing sequence and preds_scores is decreasing, so we work with &quot;-&quot;</span>
    1143. 

  1143.     <span class="n">lowest_score_above_threshold</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">searchsorted</span><span class="p">(</span><span class="o">-</span><span class="n">preds_scores</span><span class="p">,</span> <span class="o">-</span><span class="n">score_threshold</span><span class="p">,</span> <span class="n">right</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span>
    1144. 

  1144. 

  1145.     <span class="k">if</span> <span class="n">lowest_score_above_threshold</span> <span class="o">==</span> <span class="mi">0</span><span class="p">:</span>  <span class="c1"># Here score_threshold &gt; preds_scores[0], so no pred is above the threshold</span>
    1146. 

  1146.         <span class="n">recall</span> <span class="o">=</span> <span class="mi">0</span>
    1147. 

  1147.         <span class="n">precision</span> <span class="o">=</span> <span class="mi">0</span>  <span class="c1"># the precision is not really defined when no pred but we need to give it a value</span>
    1148. 

  1148.     <span class="k">else</span><span class="p">:</span>
    1149. 

  1149.         <span class="n">recall</span> <span class="o">=</span> <span class="n">rolling_recalls</span><span class="p">[</span><span class="n">lowest_score_above_threshold</span> <span class="o">-</span> <span class="mi">1</span><span class="p">]</span>
    1150. 

  1150.         <span class="n">precision</span> <span class="o">=</span> <span class="n">rolling_precisions</span><span class="p">[</span><span class="n">lowest_score_above_threshold</span> <span class="o">-</span> <span class="mi">1</span><span class="p">]</span>
    1151. 

  1151. 

  1152.     <span class="c1"># ==================</span>
    1153. 

  1153.     <span class="c1"># AVERAGE PRECISION</span>
    1154. 

  1154. 

  1155.     <span class="c1"># shape = (nb_iou_thrs, n_recall_thresholds)</span>
    1156. 

  1156.     <span class="n">recall_thresholds</span> <span class="o">=</span> <span class="n">recall_thresholds</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">)</span><span class="o">.</span><span class="n">repeat</span><span class="p">(</span><span class="n">nb_iou_thrs</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
    1157. 

  1157. 

  1158.     <span class="c1"># We want the index i so that: rolling_recalls[i-1] &lt; recall_thresholds[k] &lt;= rolling_recalls[i]</span>
    1159. 

  1159.     <span class="c1"># Note:  when recall_thresholds[k] &gt; max(rolling_recalls), i = len(rolling_recalls)</span>
    1160. 

  1160.     <span class="c1"># Note2: we work with transpose (.T) to apply torch.searchsorted on first dim instead of the last one</span>
    1161. 

  1161.     <span class="n">recall_threshold_idx</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">searchsorted</span><span class="p">(</span><span class="n">rolling_recalls</span><span class="o">.</span><span class="n">T</span><span class="p">,</span> <span class="n">recall_thresholds</span><span class="p">,</span> <span class="n">right</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span><span class="o">.</span><span class="n">T</span>
    1162. 

  1162. 

  1163.     <span class="c1"># When recall_thresholds[k] &gt; max(rolling_recalls), rolling_precisions[i] is not defined, and we want precision = 0</span>
    1164. 

  1164.     <span class="n">rolling_precisions</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">((</span><span class="n">rolling_precisions</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="n">nb_iou_thrs</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)),</span> <span class="n">dim</span><span class="o">=</span><span class="mi">0</span><span class="p">)</span>
    1165. 

  1165. 

  1166.     <span class="c1"># shape = (n_recall_thresholds, nb_iou_thrs)</span>
    1167. 

  1167.     <span class="n">sampled_precision_points</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">gather</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">rolling_precisions</span><span class="p">,</span> <span class="n">index</span><span class="o">=</span><span class="n">recall_threshold_idx</span><span class="p">,</span> <span class="n">dim</span><span class="o">=</span><span class="mi">0</span><span class="p">)</span>
    1168. 

  1168. 

  1169.     <span class="c1"># Average over the recall_thresholds</span>
    1170. 

  1170.     <span class="n">ap</span> <span class="o">=</span> <span class="n">sampled_precision_points</span><span class="o">.</span><span class="n">mean</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span>
    1171. 

  1171. 

  1172.     <span class="k">return</span> <span class="n">ap</span><span class="p">,</span> <span class="n">precision</span><span class="p">,</span> <span class="n">recall</span></div>
    1173. 

  1173. </pre></div>
    1174. 

  1174. 

  1175.            </div>
    1176. 

  1176.           </div>
    1177. 

  1177.           <footer>
    1178. 

  1178. 

  1179.   <hr/>
    1180. 

  1180. 

  1181.   <div role="contentinfo">
    1182. 

  1182.     <p>&#169; Copyright 2021, SuperGradients team.</p>
    1183. 

  1183.   </div>
    1184. 

  1184. 

  1185.   Built with <a href="https://www.sphinx-doc.org/">Sphinx</a> using a
    1186. 

  1186.     <a href="https://github.com/readthedocs/sphinx_rtd_theme">theme</a>
    1187. 

  1187.     provided by <a href="https://readthedocs.org">Read the Docs</a>.
    1188. 

  1188.    
    1189. 

  1189. 

  1190. </footer>
    1191. 

  1191.         </div>
    1192. 

  1192.       </div>
    1193. 

  1193.     </section>
    1194. 

  1194.   </div>
    1195. 

  1195.   <script>
    1196. 

  1196.       jQuery(function () {
    1197. 

  1197.           SphinxRtdTheme.Navigation.enable(true);
    1198. 

  1198.       });
    1199. 

  1199.   </script> 
    1200. 

  1200. 

  1201. </body>
    1202. 

  1202. </html>
    1203.
Discard
Prev
Next
Tip!

Press p or to see the previous file or, n or to see the next file