File size: 50,107 Bytes
2a8b904
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# Loss functions

import torch
import torch.nn as nn
import torch.nn.functional as F

from utils.general import bbox_iou, bbox_alpha_iou, box_iou, box_giou, box_diou, box_ciou, xywh2xyxy
from utils.torch_utils import is_parallel


def smooth_BCE(eps=0.1):  # https://github.com/ultralytics/yolov3/issues/238#issuecomment-598028441
    # return positive, negative label smoothing BCE targets
    return 1.0 - 0.5 * eps, 0.5 * eps


class BCEBlurWithLogitsLoss(nn.Module):
    # BCEwithLogitLoss() with reduced missing label effects.
    def __init__(self, alpha=0.05):
        super(BCEBlurWithLogitsLoss, self).__init__()
        self.loss_fcn = nn.BCEWithLogitsLoss(reduction='none')  # must be nn.BCEWithLogitsLoss()
        self.alpha = alpha

    def forward(self, pred, true):
        loss = self.loss_fcn(pred, true)
        pred = torch.sigmoid(pred)  # prob from logits
        dx = pred - true  # reduce only missing label effects
        # dx = (pred - true).abs()  # reduce missing label and false label effects
        alpha_factor = 1 - torch.exp((dx - 1) / (self.alpha + 1e-4))
        loss *= alpha_factor
        return loss.mean()


class SigmoidBin(nn.Module):
    stride = None  # strides computed during build
    export = False  # onnx export

    def __init__(self, bin_count=10, min=0.0, max=1.0, reg_scale = 2.0, use_loss_regression=True, use_fw_regression=True, BCE_weight=1.0, smooth_eps=0.0):
        super(SigmoidBin, self).__init__()
        
        self.bin_count = bin_count
        self.length = bin_count + 1
        self.min = min
        self.max = max
        self.scale = float(max - min)
        self.shift = self.scale / 2.0

        self.use_loss_regression = use_loss_regression
        self.use_fw_regression = use_fw_regression
        self.reg_scale = reg_scale
        self.BCE_weight = BCE_weight

        start = min + (self.scale/2.0) / self.bin_count
        end = max - (self.scale/2.0) / self.bin_count
        step = self.scale / self.bin_count
        self.step = step
        #print(f" start = {start}, end = {end}, step = {step} ")

        bins = torch.range(start, end + 0.0001, step).float() 
        self.register_buffer('bins', bins) 
               

        self.cp = 1.0 - 0.5 * smooth_eps
        self.cn = 0.5 * smooth_eps

        self.BCEbins = nn.BCEWithLogitsLoss(pos_weight=torch.Tensor([BCE_weight]))
        self.MSELoss = nn.MSELoss()

    def get_length(self):
        return self.length

    def forward(self, pred):
        assert pred.shape[-1] == self.length, 'pred.shape[-1]=%d is not equal to self.length=%d' % (pred.shape[-1], self.length)

        pred_reg = (pred[..., 0] * self.reg_scale - self.reg_scale/2.0) * self.step
        pred_bin = pred[..., 1:(1+self.bin_count)]

        _, bin_idx = torch.max(pred_bin, dim=-1)
        bin_bias = self.bins[bin_idx]

        if self.use_fw_regression:
            result = pred_reg + bin_bias
        else:
            result = bin_bias
        result = result.clamp(min=self.min, max=self.max)

        return result


    def training_loss(self, pred, target):
        assert pred.shape[-1] == self.length, 'pred.shape[-1]=%d is not equal to self.length=%d' % (pred.shape[-1], self.length)
        assert pred.shape[0] == target.shape[0], 'pred.shape=%d is not equal to the target.shape=%d' % (pred.shape[0], target.shape[0])
        device = pred.device

        pred_reg = (pred[..., 0].sigmoid() * self.reg_scale - self.reg_scale/2.0) * self.step
        pred_bin = pred[..., 1:(1+self.bin_count)]

        diff_bin_target = torch.abs(target[..., None] - self.bins)
        _, bin_idx = torch.min(diff_bin_target, dim=-1)
    
        bin_bias = self.bins[bin_idx]
        bin_bias.requires_grad = False
        result = pred_reg + bin_bias

        target_bins = torch.full_like(pred_bin, self.cn, device=device)  # targets
        n = pred.shape[0] 
        target_bins[range(n), bin_idx] = self.cp

        loss_bin = self.BCEbins(pred_bin, target_bins) # BCE

        if self.use_loss_regression:
            loss_regression = self.MSELoss(result, target)  # MSE        
            loss = loss_bin + loss_regression
        else:
            loss = loss_bin

        out_result = result.clamp(min=self.min, max=self.max)

        return loss, out_result


class FocalLoss(nn.Module):
    # Wraps focal loss around existing loss_fcn(), i.e. criteria = FocalLoss(nn.BCEWithLogitsLoss(), gamma=1.5)
    def __init__(self, loss_fcn, gamma=1.5, alpha=0.25):
        super(FocalLoss, self).__init__()
        self.loss_fcn = loss_fcn  # must be nn.BCEWithLogitsLoss()
        self.gamma = gamma
        self.alpha = alpha
        self.reduction = loss_fcn.reduction
        self.loss_fcn.reduction = 'none'  # required to apply FL to each element

    def forward(self, pred, true):
        loss = self.loss_fcn(pred, true)
        # p_t = torch.exp(-loss)
        # loss *= self.alpha * (1.000001 - p_t) ** self.gamma  # non-zero power for gradient stability

        # TF implementation https://github.com/tensorflow/addons/blob/v0.7.1/tensorflow_addons/losses/focal_loss.py
        pred_prob = torch.sigmoid(pred)  # prob from logits
        p_t = true * pred_prob + (1 - true) * (1 - pred_prob)
        alpha_factor = true * self.alpha + (1 - true) * (1 - self.alpha)
        modulating_factor = (1.0 - p_t) ** self.gamma
        loss *= alpha_factor * modulating_factor

        if self.reduction == 'mean':
            return loss.mean()
        elif self.reduction == 'sum':
            return loss.sum()
        else:  # 'none'
            return loss


class QFocalLoss(nn.Module):
    # Wraps Quality focal loss around existing loss_fcn(), i.e. criteria = FocalLoss(nn.BCEWithLogitsLoss(), gamma=1.5)
    def __init__(self, loss_fcn, gamma=1.5, alpha=0.25):
        super(QFocalLoss, self).__init__()
        self.loss_fcn = loss_fcn  # must be nn.BCEWithLogitsLoss()
        self.gamma = gamma
        self.alpha = alpha
        self.reduction = loss_fcn.reduction
        self.loss_fcn.reduction = 'none'  # required to apply FL to each element

    def forward(self, pred, true):
        loss = self.loss_fcn(pred, true)

        pred_prob = torch.sigmoid(pred)  # prob from logits
        alpha_factor = true * self.alpha + (1 - true) * (1 - self.alpha)
        modulating_factor = torch.abs(true - pred_prob) ** self.gamma
        loss *= alpha_factor * modulating_factor

        if self.reduction == 'mean':
            return loss.mean()
        elif self.reduction == 'sum':
            return loss.sum()
        else:  # 'none'
            return loss

class RankSort(torch.autograd.Function):
    @staticmethod
    def forward(ctx, logits, targets, delta_RS=0.50, eps=1e-10): 

        classification_grads=torch.zeros(logits.shape).cuda()
        
        #Filter fg logits
        fg_labels = (targets > 0.)
        fg_logits = logits[fg_labels]
        fg_targets = targets[fg_labels]
        fg_num = len(fg_logits)

        #Do not use bg with scores less than minimum fg logit
        #since changing its score does not have an effect on precision
        threshold_logit = torch.min(fg_logits)-delta_RS
        relevant_bg_labels=((targets==0) & (logits>=threshold_logit))
        
        relevant_bg_logits = logits[relevant_bg_labels] 
        relevant_bg_grad=torch.zeros(len(relevant_bg_logits)).cuda()
        sorting_error=torch.zeros(fg_num).cuda()
        ranking_error=torch.zeros(fg_num).cuda()
        fg_grad=torch.zeros(fg_num).cuda()
        
        #sort the fg logits
        order=torch.argsort(fg_logits)
        #Loops over each positive following the order
        for ii in order:
            # Difference Transforms (x_ij)
            fg_relations=fg_logits-fg_logits[ii] 
            bg_relations=relevant_bg_logits-fg_logits[ii]

            if delta_RS > 0:
                fg_relations=torch.clamp(fg_relations/(2*delta_RS)+0.5,min=0,max=1)
                bg_relations=torch.clamp(bg_relations/(2*delta_RS)+0.5,min=0,max=1)
            else:
                fg_relations = (fg_relations >= 0).float()
                bg_relations = (bg_relations >= 0).float()

            # Rank of ii among pos and false positive number (bg with larger scores)
            rank_pos=torch.sum(fg_relations)
            FP_num=torch.sum(bg_relations)

            # Rank of ii among all examples
            rank=rank_pos+FP_num
                            
            # Ranking error of example ii. target_ranking_error is always 0. (Eq. 7)
            ranking_error[ii]=FP_num/rank      

            # Current sorting error of example ii. (Eq. 7)
            current_sorting_error = torch.sum(fg_relations*(1-fg_targets))/rank_pos

            #Find examples in the target sorted order for example ii         
            iou_relations = (fg_targets >= fg_targets[ii])
            target_sorted_order = iou_relations * fg_relations

            #The rank of ii among positives in sorted order
            rank_pos_target = torch.sum(target_sorted_order)

            #Compute target sorting error. (Eq. 8)
            #Since target ranking error is 0, this is also total target error 
            target_sorting_error= torch.sum(target_sorted_order*(1-fg_targets))/rank_pos_target

            #Compute sorting error on example ii
            sorting_error[ii] = current_sorting_error - target_sorting_error
  
            #Identity Update for Ranking Error 
            if FP_num > eps:
                #For ii the update is the ranking error
                fg_grad[ii] -= ranking_error[ii]
                #For negatives, distribute error via ranking pmf (i.e. bg_relations/FP_num)
                relevant_bg_grad += (bg_relations*(ranking_error[ii]/FP_num))

            #Find the positives that are misranked (the cause of the error)
            #These are the ones with smaller IoU but larger logits
            missorted_examples = (~ iou_relations) * fg_relations

            #Denominotor of sorting pmf 
            sorting_pmf_denom = torch.sum(missorted_examples)

            #Identity Update for Sorting Error 
            if sorting_pmf_denom > eps:
                #For ii the update is the sorting error
                fg_grad[ii] -= sorting_error[ii]
                #For positives, distribute error via sorting pmf (i.e. missorted_examples/sorting_pmf_denom)
                fg_grad += (missorted_examples*(sorting_error[ii]/sorting_pmf_denom))

        #Normalize gradients by number of positives 
        classification_grads[fg_labels]= (fg_grad/fg_num)
        classification_grads[relevant_bg_labels]= (relevant_bg_grad/fg_num)

        ctx.save_for_backward(classification_grads)

        return ranking_error.mean(), sorting_error.mean()

    @staticmethod
    def backward(ctx, out_grad1, out_grad2):
        g1, =ctx.saved_tensors
        return g1*out_grad1, None, None, None

class aLRPLoss(torch.autograd.Function):
    @staticmethod
    def forward(ctx, logits, targets, regression_losses, delta=1., eps=1e-5): 
        classification_grads=torch.zeros(logits.shape).cuda()
        
        #Filter fg logits
        fg_labels = (targets == 1)
        fg_logits = logits[fg_labels]
        fg_num = len(fg_logits)

        #Do not use bg with scores less than minimum fg logit
        #since changing its score does not have an effect on precision
        threshold_logit = torch.min(fg_logits)-delta

        #Get valid bg logits
        relevant_bg_labels=((targets==0)&(logits>=threshold_logit))
        relevant_bg_logits=logits[relevant_bg_labels] 
        relevant_bg_grad=torch.zeros(len(relevant_bg_logits)).cuda()
        rank=torch.zeros(fg_num).cuda()
        prec=torch.zeros(fg_num).cuda()
        fg_grad=torch.zeros(fg_num).cuda()
        
        max_prec=0                                           
        #sort the fg logits
        order=torch.argsort(fg_logits)
        #Loops over each positive following the order
        for ii in order:
            #x_ij s as score differences with fgs
            fg_relations=fg_logits-fg_logits[ii] 
            #Apply piecewise linear function and determine relations with fgs
            fg_relations=torch.clamp(fg_relations/(2*delta)+0.5,min=0,max=1)
            #Discard i=j in the summation in rank_pos
            fg_relations[ii]=0

            #x_ij s as score differences with bgs
            bg_relations=relevant_bg_logits-fg_logits[ii]
            #Apply piecewise linear function and determine relations with bgs
            bg_relations=torch.clamp(bg_relations/(2*delta)+0.5,min=0,max=1)

            #Compute the rank of the example within fgs and number of bgs with larger scores
            rank_pos=1+torch.sum(fg_relations)
            FP_num=torch.sum(bg_relations)
            #Store the total since it is normalizer also for aLRP Regression error
            rank[ii]=rank_pos+FP_num
                            
            #Compute precision for this example to compute classification loss 
            prec[ii]=rank_pos/rank[ii]                
            #For stability, set eps to a infinitesmall value (e.g. 1e-6), then compute grads
            if FP_num > eps:   
                fg_grad[ii] = -(torch.sum(fg_relations*regression_losses)+FP_num)/rank[ii]
                relevant_bg_grad += (bg_relations*(-fg_grad[ii]/FP_num))   
                    
        #aLRP with grad formulation fg gradient
        classification_grads[fg_labels]= fg_grad
        #aLRP with grad formulation bg gradient
        classification_grads[relevant_bg_labels]= relevant_bg_grad 
 
        classification_grads /= (fg_num)
    
        cls_loss=1-prec.mean()
        ctx.save_for_backward(classification_grads)

        return cls_loss, rank, order

    @staticmethod
    def backward(ctx, out_grad1, out_grad2, out_grad3):
        g1, =ctx.saved_tensors
        return g1*out_grad1, None, None, None, None
    
    
class APLoss(torch.autograd.Function):
    @staticmethod
    def forward(ctx, logits, targets, delta=1.): 
        classification_grads=torch.zeros(logits.shape).cuda()
        
        #Filter fg logits
        fg_labels = (targets == 1)
        fg_logits = logits[fg_labels]
        fg_num = len(fg_logits)

        #Do not use bg with scores less than minimum fg logit
        #since changing its score does not have an effect on precision
        threshold_logit = torch.min(fg_logits)-delta

        #Get valid bg logits
        relevant_bg_labels=((targets==0)&(logits>=threshold_logit))
        relevant_bg_logits=logits[relevant_bg_labels] 
        relevant_bg_grad=torch.zeros(len(relevant_bg_logits)).cuda()
        rank=torch.zeros(fg_num).cuda()
        prec=torch.zeros(fg_num).cuda()
        fg_grad=torch.zeros(fg_num).cuda()
        
        max_prec=0                                           
        #sort the fg logits
        order=torch.argsort(fg_logits)
        #Loops over each positive following the order
        for ii in order:
            #x_ij s as score differences with fgs
            fg_relations=fg_logits-fg_logits[ii] 
            #Apply piecewise linear function and determine relations with fgs
            fg_relations=torch.clamp(fg_relations/(2*delta)+0.5,min=0,max=1)
            #Discard i=j in the summation in rank_pos
            fg_relations[ii]=0

            #x_ij s as score differences with bgs
            bg_relations=relevant_bg_logits-fg_logits[ii]
            #Apply piecewise linear function and determine relations with bgs
            bg_relations=torch.clamp(bg_relations/(2*delta)+0.5,min=0,max=1)

            #Compute the rank of the example within fgs and number of bgs with larger scores
            rank_pos=1+torch.sum(fg_relations)
            FP_num=torch.sum(bg_relations)
            #Store the total since it is normalizer also for aLRP Regression error
            rank[ii]=rank_pos+FP_num
                            
            #Compute precision for this example 
            current_prec=rank_pos/rank[ii]
            
            #Compute interpolated AP and store gradients for relevant bg examples
            if (max_prec<=current_prec):
                max_prec=current_prec
                relevant_bg_grad += (bg_relations/rank[ii])
            else:
                relevant_bg_grad += (bg_relations/rank[ii])*(((1-max_prec)/(1-current_prec)))
            
            #Store fg gradients
            fg_grad[ii]=-(1-max_prec)
            prec[ii]=max_prec 

        #aLRP with grad formulation fg gradient
        classification_grads[fg_labels]= fg_grad
        #aLRP with grad formulation bg gradient
        classification_grads[relevant_bg_labels]= relevant_bg_grad 
 
        classification_grads /= fg_num
    
        cls_loss=1-prec.mean()
        ctx.save_for_backward(classification_grads)

        return cls_loss

    @staticmethod
    def backward(ctx, out_grad1):
        g1, =ctx.saved_tensors
        return g1*out_grad1, None, None


class ComputeLoss:
    # Compute losses
    def __init__(self, model, autobalance=False):
        super(ComputeLoss, self).__init__()
        device = next(model.parameters()).device  # get model device
        h = model.hyp  # hyperparameters

        # Define criteria
        BCEcls = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['cls_pw']], device=device))
        BCEobj = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['obj_pw']], device=device))

        # Class label smoothing https://arxiv.org/pdf/1902.04103.pdf eqn 3
        self.cp, self.cn = smooth_BCE(eps=h.get('label_smoothing', 0.0))  # positive, negative BCE targets

        # Focal loss
        g = h['fl_gamma']  # focal loss gamma
        if g > 0:
            BCEcls, BCEobj = FocalLoss(BCEcls, g), FocalLoss(BCEobj, g)

        det = model.module.model[-1] if is_parallel(model) else model.model[-1]  # Detect() module
        self.balance = {3: [4.0, 1.0, 0.4]}.get(det.nl, [4.0, 1.0, 0.25, 0.06, .02])  # P3-P7
        #self.balance = {3: [4.0, 1.0, 0.4]}.get(det.nl, [4.0, 1.0, 0.25, 0.1, .05])  # P3-P7
        #self.balance = {3: [4.0, 1.0, 0.4]}.get(det.nl, [4.0, 1.0, 0.5, 0.4, .1])  # P3-P7
        self.ssi = list(det.stride).index(16) if autobalance else 0  # stride 16 index
        self.BCEcls, self.BCEobj, self.gr, self.hyp, self.autobalance = BCEcls, BCEobj, model.gr, h, autobalance
        for k in 'na', 'nc', 'nl', 'anchors':
            setattr(self, k, getattr(det, k))

    def __call__(self, p, targets):  # predictions, targets, model
        device = targets.device
        lcls, lbox, lobj = torch.zeros(1, device=device), torch.zeros(1, device=device), torch.zeros(1, device=device)
        tcls, tbox, indices, anchors = self.build_targets(p, targets)  # targets

        # Losses
        for i, pi in enumerate(p):  # layer index, layer predictions
            b, a, gj, gi = indices[i]  # image, anchor, gridy, gridx
            tobj = torch.zeros_like(pi[..., 0], device=device)  # target obj

            n = b.shape[0]  # number of targets
            if n:
                ps = pi[b, a, gj, gi]  # prediction subset corresponding to targets

                # Regression
                pxy = ps[:, :2].sigmoid() * 2. - 0.5
                pwh = (ps[:, 2:4].sigmoid() * 2) ** 2 * anchors[i]
                pbox = torch.cat((pxy, pwh), 1)  # predicted box
                iou = bbox_iou(pbox.T, tbox[i], x1y1x2y2=False, CIoU=True)  # iou(prediction, target)
                lbox += (1.0 - iou).mean()  # iou loss

                # Objectness
                tobj[b, a, gj, gi] = (1.0 - self.gr) + self.gr * iou.detach().clamp(0).type(tobj.dtype)  # iou ratio

                # Classification
                if self.nc > 1:  # cls loss (only if multiple classes)
                    t = torch.full_like(ps[:, 5:], self.cn, device=device)  # targets
                    t[range(n), tcls[i]] = self.cp
                    #t[t==self.cp] = iou.detach().clamp(0).type(t.dtype)
                    lcls += self.BCEcls(ps[:, 5:], t)  # BCE

                # Append targets to text file
                # with open('targets.txt', 'a') as file:
                #     [file.write('%11.5g ' * 4 % tuple(x) + '\n') for x in torch.cat((txy[i], twh[i]), 1)]

            obji = self.BCEobj(pi[..., 4], tobj)
            lobj += obji * self.balance[i]  # obj loss
            if self.autobalance:
                self.balance[i] = self.balance[i] * 0.9999 + 0.0001 / obji.detach().item()

        if self.autobalance:
            self.balance = [x / self.balance[self.ssi] for x in self.balance]
        lbox *= self.hyp['box']
        lobj *= self.hyp['obj']
        lcls *= self.hyp['cls']
        bs = tobj.shape[0]  # batch size

        loss = lbox + lobj + lcls
        return loss * bs, torch.cat((lbox, lobj, lcls, loss)).detach()

    def build_targets(self, p, targets):
        # Build targets for compute_loss(), input targets(image,class,x,y,w,h)
        na, nt = self.na, targets.shape[0]  # number of anchors, targets
        tcls, tbox, indices, anch = [], [], [], []
        gain = torch.ones(7, device=targets.device)  # normalized to gridspace gain
        ai = torch.arange(na, device=targets.device).float().view(na, 1).repeat(1, nt)  # same as .repeat_interleave(nt)
        targets = torch.cat((targets.repeat(na, 1, 1), ai[:, :, None]), 2)  # append anchor indices

        g = 0.5  # bias
        off = torch.tensor([[0, 0],
                            [1, 0], [0, 1], [-1, 0], [0, -1],  # j,k,l,m
                            # [1, 1], [1, -1], [-1, 1], [-1, -1],  # jk,jm,lk,lm
                            ], device=targets.device).float() * g  # offsets

        for i in range(self.nl):
            anchors = self.anchors[i]
            gain[2:6] = torch.tensor(p[i].shape)[[3, 2, 3, 2]]  # xyxy gain

            # Match targets to anchors
            t = targets * gain
            if nt:
                # Matches
                r = t[:, :, 4:6] / anchors[:, None]  # wh ratio
                j = torch.max(r, 1. / r).max(2)[0] < self.hyp['anchor_t']  # compare
                # j = wh_iou(anchors, t[:, 4:6]) > model.hyp['iou_t']  # iou(3,n)=wh_iou(anchors(3,2), gwh(n,2))
                t = t[j]  # filter

                # Offsets
                gxy = t[:, 2:4]  # grid xy
                gxi = gain[[2, 3]] - gxy  # inverse
                j, k = ((gxy % 1. < g) & (gxy > 1.)).T
                l, m = ((gxi % 1. < g) & (gxi > 1.)).T
                j = torch.stack((torch.ones_like(j), j, k, l, m))
                t = t.repeat((5, 1, 1))[j]
                offsets = (torch.zeros_like(gxy)[None] + off[:, None])[j]
            else:
                t = targets[0]
                offsets = 0

            # Define
            b, c = t[:, :2].long().T  # image, class
            gxy = t[:, 2:4]  # grid xy
            gwh = t[:, 4:6]  # grid wh
            gij = (gxy - offsets).long()
            gi, gj = gij.T  # grid xy indices

            # Append
            a = t[:, 6].long()  # anchor indices
            indices.append((b, a, gj.clamp_(0, gain[3] - 1), gi.clamp_(0, gain[2] - 1)))  # image, anchor, grid indices
            tbox.append(torch.cat((gxy - gij, gwh), 1))  # box
            anch.append(anchors[a])  # anchors
            tcls.append(c)  # class

        return tcls, tbox, indices, anch


class ComputeLossOTA:
    # Compute losses
    def __init__(self, model, autobalance=False):
        super(ComputeLossOTA, self).__init__()
        device = next(model.parameters()).device  # get model device
        h = model.hyp  # hyperparameters

        # Define criteria
        BCEcls = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['cls_pw']], device=device))
        BCEobj = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['obj_pw']], device=device))

        # Class label smoothing https://arxiv.org/pdf/1902.04103.pdf eqn 3
        self.cp, self.cn = smooth_BCE(eps=h.get('label_smoothing', 0.0))  # positive, negative BCE targets

        # Focal loss
        g = h['fl_gamma']  # focal loss gamma
        if g > 0:
            BCEcls, BCEobj = FocalLoss(BCEcls, g), FocalLoss(BCEobj, g)

        det = model.module.model[-1] if is_parallel(model) else model.model[-1]  # Detect() module
        self.balance = {3: [4.0, 1.0, 0.4]}.get(det.nl, [4.0, 1.0, 0.25, 0.06, .02])  # P3-P7
        self.ssi = list(det.stride).index(16) if autobalance else 0  # stride 16 index
        self.BCEcls, self.BCEobj, self.gr, self.hyp, self.autobalance = BCEcls, BCEobj, model.gr, h, autobalance
        for k in 'na', 'nc', 'nl', 'anchors', 'stride':
            setattr(self, k, getattr(det, k))

    def __call__(self, p, targets, imgs):  # predictions, targets, model   
        device = targets.device
        lcls, lbox, lobj = torch.zeros(1, device=device), torch.zeros(1, device=device), torch.zeros(1, device=device)
        bs, as_, gjs, gis, targets, anchors = self.build_targets(p, targets, imgs)
        pre_gen_gains = [torch.tensor(pp.shape, device=device)[[3, 2, 3, 2]] for pp in p] 
    

        # Losses
        for i, pi in enumerate(p):  # layer index, layer predictions
            b, a, gj, gi = bs[i], as_[i], gjs[i], gis[i]  # image, anchor, gridy, gridx
            tobj = torch.zeros_like(pi[..., 0], device=device)  # target obj

            n = b.shape[0]  # number of targets
            if n:
                ps = pi[b, a, gj, gi]  # prediction subset corresponding to targets

                # Regression
                grid = torch.stack([gi, gj], dim=1)
                pxy = ps[:, :2].sigmoid() * 2. - 0.5
                #pxy = ps[:, :2].sigmoid() * 3. - 1.
                pwh = (ps[:, 2:4].sigmoid() * 2) ** 2 * anchors[i]
                pbox = torch.cat((pxy, pwh), 1)  # predicted box
                selected_tbox = targets[i][:, 2:6] * pre_gen_gains[i]
                selected_tbox[:, :2] -= grid
                iou = bbox_iou(pbox.T, selected_tbox, x1y1x2y2=False, CIoU=True)  # iou(prediction, target)
                lbox += (1.0 - iou).mean()  # iou loss

                # Objectness
                tobj[b, a, gj, gi] = (1.0 - self.gr) + self.gr * iou.detach().clamp(0).type(tobj.dtype)  # iou ratio

                # Classification
                selected_tcls = targets[i][:, 1].long()
                if self.nc > 1:  # cls loss (only if multiple classes)
                    t = torch.full_like(ps[:, 5:], self.cn, device=device)  # targets
                    t[range(n), selected_tcls] = self.cp
                    lcls += self.BCEcls(ps[:, 5:], t)  # BCE

                # Append targets to text file
                # with open('targets.txt', 'a') as file:
                #     [file.write('%11.5g ' * 4 % tuple(x) + '\n') for x in torch.cat((txy[i], twh[i]), 1)]

            obji = self.BCEobj(pi[..., 4], tobj)
            lobj += obji * self.balance[i]  # obj loss
            if self.autobalance:
                self.balance[i] = self.balance[i] * 0.9999 + 0.0001 / obji.detach().item()

        if self.autobalance:
            self.balance = [x / self.balance[self.ssi] for x in self.balance]
        lbox *= self.hyp['box']
        lobj *= self.hyp['obj']
        lcls *= self.hyp['cls']
        bs = tobj.shape[0]  # batch size

        loss = lbox + lobj + lcls
        return loss * bs, torch.cat((lbox, lobj, lcls, loss)).detach()

    def build_targets(self, p, targets, imgs):
        
        #indices, anch = self.find_positive(p, targets)
        indices, anch = self.find_3_positive(p, targets)
        #indices, anch = self.find_4_positive(p, targets)
        #indices, anch = self.find_5_positive(p, targets)
        #indices, anch = self.find_9_positive(p, targets)

        matching_bs = [[] for pp in p]
        matching_as = [[] for pp in p]
        matching_gjs = [[] for pp in p]
        matching_gis = [[] for pp in p]
        matching_targets = [[] for pp in p]
        matching_anchs = [[] for pp in p]
        
        nl = len(p)    
    
        for batch_idx in range(p[0].shape[0]):
        
            b_idx = targets[:, 0]==batch_idx
            this_target = targets[b_idx]
            if this_target.shape[0] == 0:
                continue
                
            txywh = this_target[:, 2:6] * imgs[batch_idx].shape[1]
            txyxy = xywh2xyxy(txywh)

            pxyxys = []
            p_cls = []
            p_obj = []
            from_which_layer = []
            all_b = []
            all_a = []
            all_gj = []
            all_gi = []
            all_anch = []
            
            for i, pi in enumerate(p):
                
                b, a, gj, gi = indices[i]
                idx = (b == batch_idx)
                b, a, gj, gi = b[idx], a[idx], gj[idx], gi[idx]                
                all_b.append(b)
                all_a.append(a)
                all_gj.append(gj)
                all_gi.append(gi)
                all_anch.append(anch[i][idx])
                from_which_layer.append(torch.ones(size=(len(b),)) * i)
                
                fg_pred = pi[b, a, gj, gi]                
                p_obj.append(fg_pred[:, 4:5])
                p_cls.append(fg_pred[:, 5:])
                
                grid = torch.stack([gi, gj], dim=1)
                pxy = (fg_pred[:, :2].sigmoid() * 2. - 0.5 + grid) * self.stride[i] #/ 8.
                #pxy = (fg_pred[:, :2].sigmoid() * 3. - 1. + grid) * self.stride[i]
                pwh = (fg_pred[:, 2:4].sigmoid() * 2) ** 2 * anch[i][idx] * self.stride[i] #/ 8.
                pxywh = torch.cat([pxy, pwh], dim=-1)
                pxyxy = xywh2xyxy(pxywh)
                pxyxys.append(pxyxy)
            
            pxyxys = torch.cat(pxyxys, dim=0)
            if pxyxys.shape[0] == 0:
                continue
            p_obj = torch.cat(p_obj, dim=0)
            p_cls = torch.cat(p_cls, dim=0)
            from_which_layer = torch.cat(from_which_layer, dim=0)
            all_b = torch.cat(all_b, dim=0)
            all_a = torch.cat(all_a, dim=0)
            all_gj = torch.cat(all_gj, dim=0)
            all_gi = torch.cat(all_gi, dim=0)
            all_anch = torch.cat(all_anch, dim=0)
        
            pair_wise_iou = box_iou(txyxy, pxyxys)

            pair_wise_iou_loss = -torch.log(pair_wise_iou + 1e-8)

            top_k, _ = torch.topk(pair_wise_iou, min(10, pair_wise_iou.shape[1]), dim=1)
            dynamic_ks = torch.clamp(top_k.sum(1).int(), min=1)

            gt_cls_per_image = (
                F.one_hot(this_target[:, 1].to(torch.int64), self.nc)
                .float()
                .unsqueeze(1)
                .repeat(1, pxyxys.shape[0], 1)
            )

            num_gt = this_target.shape[0]
            cls_preds_ = (
                p_cls.float().unsqueeze(0).repeat(num_gt, 1, 1).sigmoid_()
                * p_obj.unsqueeze(0).repeat(num_gt, 1, 1).sigmoid_()
            )

            y = cls_preds_.sqrt_()
            pair_wise_cls_loss = F.binary_cross_entropy_with_logits(
               torch.log(y/(1-y)) , gt_cls_per_image, reduction="none"
            ).sum(-1)
            del cls_preds_
        
            cost = (
                pair_wise_cls_loss
                + 3.0 * pair_wise_iou_loss
            )

            matching_matrix = torch.zeros_like(cost)

            for gt_idx in range(num_gt):
                _, pos_idx = torch.topk(
                    cost[gt_idx], k=dynamic_ks[gt_idx].item(), largest=False
                )
                matching_matrix[gt_idx][pos_idx] = 1.0

            del top_k, dynamic_ks
            anchor_matching_gt = matching_matrix.sum(0)
            if (anchor_matching_gt > 1).sum() > 0:
                _, cost_argmin = torch.min(cost[:, anchor_matching_gt > 1], dim=0)
                matching_matrix[:, anchor_matching_gt > 1] *= 0.0
                matching_matrix[cost_argmin, anchor_matching_gt > 1] = 1.0
            fg_mask_inboxes = matching_matrix.sum(0) > 0.0
            matched_gt_inds = matching_matrix[:, fg_mask_inboxes].argmax(0)
        
            from_which_layer = from_which_layer[fg_mask_inboxes]
            all_b = all_b[fg_mask_inboxes]
            all_a = all_a[fg_mask_inboxes]
            all_gj = all_gj[fg_mask_inboxes]
            all_gi = all_gi[fg_mask_inboxes]
            all_anch = all_anch[fg_mask_inboxes]
        
            this_target = this_target[matched_gt_inds]
        
            for i in range(nl):
                layer_idx = from_which_layer == i
                matching_bs[i].append(all_b[layer_idx])
                matching_as[i].append(all_a[layer_idx])
                matching_gjs[i].append(all_gj[layer_idx])
                matching_gis[i].append(all_gi[layer_idx])
                matching_targets[i].append(this_target[layer_idx])
                matching_anchs[i].append(all_anch[layer_idx])

        for i in range(nl):
            matching_bs[i] = torch.cat(matching_bs[i], dim=0)
            matching_as[i] = torch.cat(matching_as[i], dim=0)
            matching_gjs[i] = torch.cat(matching_gjs[i], dim=0)
            matching_gis[i] = torch.cat(matching_gis[i], dim=0)
            matching_targets[i] = torch.cat(matching_targets[i], dim=0)
            matching_anchs[i] = torch.cat(matching_anchs[i], dim=0)

        return matching_bs, matching_as, matching_gjs, matching_gis, matching_targets, matching_anchs           

    def find_3_positive(self, p, targets):
        # Build targets for compute_loss(), input targets(image,class,x,y,w,h)
        na, nt = self.na, targets.shape[0]  # number of anchors, targets
        indices, anch = [], []
        gain = torch.ones(7, device=targets.device)  # normalized to gridspace gain
        ai = torch.arange(na, device=targets.device).float().view(na, 1).repeat(1, nt)  # same as .repeat_interleave(nt)
        targets = torch.cat((targets.repeat(na, 1, 1), ai[:, :, None]), 2)  # append anchor indices

        g = 0.5  # bias
        off = torch.tensor([[0, 0],
                            [1, 0], [0, 1], [-1, 0], [0, -1],  # j,k,l,m
                            # [1, 1], [1, -1], [-1, 1], [-1, -1],  # jk,jm,lk,lm
                            ], device=targets.device).float() * g  # offsets

        for i in range(self.nl):
            anchors = self.anchors[i]
            gain[2:6] = torch.tensor(p[i].shape)[[3, 2, 3, 2]]  # xyxy gain

            # Match targets to anchors
            t = targets * gain
            if nt:
                # Matches
                r = t[:, :, 4:6] / anchors[:, None]  # wh ratio
                j = torch.max(r, 1. / r).max(2)[0] < self.hyp['anchor_t']  # compare
                # j = wh_iou(anchors, t[:, 4:6]) > model.hyp['iou_t']  # iou(3,n)=wh_iou(anchors(3,2), gwh(n,2))
                t = t[j]  # filter

                # Offsets
                gxy = t[:, 2:4]  # grid xy
                gxi = gain[[2, 3]] - gxy  # inverse
                j, k = ((gxy % 1. < g) & (gxy > 1.)).T
                l, m = ((gxi % 1. < g) & (gxi > 1.)).T
                j = torch.stack((torch.ones_like(j), j, k, l, m))
                t = t.repeat((5, 1, 1))[j]
                offsets = (torch.zeros_like(gxy)[None] + off[:, None])[j]
            else:
                t = targets[0]
                offsets = 0

            # Define
            b, c = t[:, :2].long().T  # image, class
            gxy = t[:, 2:4]  # grid xy
            gwh = t[:, 4:6]  # grid wh
            gij = (gxy - offsets).long()
            gi, gj = gij.T  # grid xy indices

            # Append
            a = t[:, 6].long()  # anchor indices
            indices.append((b, a, gj.clamp_(0, gain[3] - 1), gi.clamp_(0, gain[2] - 1)))  # image, anchor, grid indices
            anch.append(anchors[a])  # anchors

        return indices, anch
    

class ComputeLossBinOTA:
    # Compute losses
    def __init__(self, model, autobalance=False):
        super(ComputeLossBinOTA, self).__init__()
        device = next(model.parameters()).device  # get model device
        h = model.hyp  # hyperparameters

        # Define criteria
        BCEcls = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['cls_pw']], device=device))
        BCEobj = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['obj_pw']], device=device))
        #MSEangle = nn.MSELoss().to(device)

        # Class label smoothing https://arxiv.org/pdf/1902.04103.pdf eqn 3
        self.cp, self.cn = smooth_BCE(eps=h.get('label_smoothing', 0.0))  # positive, negative BCE targets

        # Focal loss
        g = h['fl_gamma']  # focal loss gamma
        if g > 0:
            BCEcls, BCEobj = FocalLoss(BCEcls, g), FocalLoss(BCEobj, g)

        det = model.module.model[-1] if is_parallel(model) else model.model[-1]  # Detect() module
        self.balance = {3: [4.0, 1.0, 0.4]}.get(det.nl, [4.0, 1.0, 0.25, 0.06, .02])  # P3-P7
        self.ssi = list(det.stride).index(16) if autobalance else 0  # stride 16 index
        self.BCEcls, self.BCEobj, self.gr, self.hyp, self.autobalance = BCEcls, BCEobj, model.gr, h, autobalance
        for k in 'na', 'nc', 'nl', 'anchors', 'stride', 'bin_count':
            setattr(self, k, getattr(det, k))

        #xy_bin_sigmoid = SigmoidBin(bin_count=11, min=-0.5, max=1.5, use_loss_regression=False).to(device)
        wh_bin_sigmoid = SigmoidBin(bin_count=self.bin_count, min=0.0, max=4.0, use_loss_regression=False).to(device)
        #angle_bin_sigmoid = SigmoidBin(bin_count=31, min=-1.1, max=1.1, use_loss_regression=False).to(device)
        self.wh_bin_sigmoid = wh_bin_sigmoid

    def __call__(self, p, targets, imgs):  # predictions, targets, model   
        device = targets.device
        lcls, lbox, lobj = torch.zeros(1, device=device), torch.zeros(1, device=device), torch.zeros(1, device=device)
        bs, as_, gjs, gis, targets, anchors = self.build_targets(p, targets, imgs)
        pre_gen_gains = [torch.tensor(pp.shape, device=device)[[3, 2, 3, 2]] for pp in p] 
    

        # Losses
        for i, pi in enumerate(p):  # layer index, layer predictions
            b, a, gj, gi = bs[i], as_[i], gjs[i], gis[i]  # image, anchor, gridy, gridx
            tobj = torch.zeros_like(pi[..., 0], device=device)  # target obj

            obj_idx = self.wh_bin_sigmoid.get_length()*2 + 2     # x,y, w-bce, h-bce     # xy_bin_sigmoid.get_length()*2

            n = b.shape[0]  # number of targets
            if n:
                ps = pi[b, a, gj, gi]  # prediction subset corresponding to targets

                # Regression
                grid = torch.stack([gi, gj], dim=1)
                selected_tbox = targets[i][:, 2:6] * pre_gen_gains[i]
                selected_tbox[:, :2] -= grid
                
                #pxy = ps[:, :2].sigmoid() * 2. - 0.5
                ##pxy = ps[:, :2].sigmoid() * 3. - 1.
                #pwh = (ps[:, 2:4].sigmoid() * 2) ** 2 * anchors[i]
                #pbox = torch.cat((pxy, pwh), 1)  # predicted box

                #x_loss, px = xy_bin_sigmoid.training_loss(ps[..., 0:12], tbox[i][..., 0])
                #y_loss, py = xy_bin_sigmoid.training_loss(ps[..., 12:24], tbox[i][..., 1])
                w_loss, pw = self.wh_bin_sigmoid.training_loss(ps[..., 2:(3+self.bin_count)], selected_tbox[..., 2] / anchors[i][..., 0])
                h_loss, ph = self.wh_bin_sigmoid.training_loss(ps[..., (3+self.bin_count):obj_idx], selected_tbox[..., 3] / anchors[i][..., 1])

                pw *= anchors[i][..., 0]
                ph *= anchors[i][..., 1]

                px = ps[:, 0].sigmoid() * 2. - 0.5
                py = ps[:, 1].sigmoid() * 2. - 0.5

                lbox += w_loss + h_loss # + x_loss + y_loss

                #print(f"\n px = {px.shape}, py = {py.shape}, pw = {pw.shape}, ph = {ph.shape} \n")

                pbox = torch.cat((px.unsqueeze(1), py.unsqueeze(1), pw.unsqueeze(1), ph.unsqueeze(1)), 1).to(device)  # predicted box

                
                
                
                iou = bbox_iou(pbox.T, selected_tbox, x1y1x2y2=False, CIoU=True)  # iou(prediction, target)
                lbox += (1.0 - iou).mean()  # iou loss

                # Objectness
                tobj[b, a, gj, gi] = (1.0 - self.gr) + self.gr * iou.detach().clamp(0).type(tobj.dtype)  # iou ratio

                # Classification
                selected_tcls = targets[i][:, 1].long()
                if self.nc > 1:  # cls loss (only if multiple classes)
                    t = torch.full_like(ps[:, (1+obj_idx):], self.cn, device=device)  # targets
                    t[range(n), selected_tcls] = self.cp
                    lcls += self.BCEcls(ps[:, (1+obj_idx):], t)  # BCE

                # Append targets to text file
                # with open('targets.txt', 'a') as file:
                #     [file.write('%11.5g ' * 4 % tuple(x) + '\n') for x in torch.cat((txy[i], twh[i]), 1)]

            obji = self.BCEobj(pi[..., obj_idx], tobj)
            lobj += obji * self.balance[i]  # obj loss
            if self.autobalance:
                self.balance[i] = self.balance[i] * 0.9999 + 0.0001 / obji.detach().item()

        if self.autobalance:
            self.balance = [x / self.balance[self.ssi] for x in self.balance]
        lbox *= self.hyp['box']
        lobj *= self.hyp['obj']
        lcls *= self.hyp['cls']
        bs = tobj.shape[0]  # batch size

        loss = lbox + lobj + lcls
        return loss * bs, torch.cat((lbox, lobj, lcls, loss)).detach()

    def build_targets(self, p, targets, imgs):
        
        #indices, anch = self.find_positive(p, targets)
        indices, anch = self.find_3_positive(p, targets)
        #indices, anch = self.find_4_positive(p, targets)
        #indices, anch = self.find_5_positive(p, targets)
        #indices, anch = self.find_9_positive(p, targets)

        matching_bs = [[] for pp in p]
        matching_as = [[] for pp in p]
        matching_gjs = [[] for pp in p]
        matching_gis = [[] for pp in p]
        matching_targets = [[] for pp in p]
        matching_anchs = [[] for pp in p]
        
        nl = len(p)    
    
        for batch_idx in range(p[0].shape[0]):
        
            b_idx = targets[:, 0]==batch_idx
            this_target = targets[b_idx]
            if this_target.shape[0] == 0:
                continue
                
            txywh = this_target[:, 2:6] * imgs[batch_idx].shape[1]
            txyxy = xywh2xyxy(txywh)

            pxyxys = []
            p_cls = []
            p_obj = []
            from_which_layer = []
            all_b = []
            all_a = []
            all_gj = []
            all_gi = []
            all_anch = []
            
            for i, pi in enumerate(p):
                
                obj_idx = self.wh_bin_sigmoid.get_length()*2 + 2
                
                b, a, gj, gi = indices[i]
                idx = (b == batch_idx)
                b, a, gj, gi = b[idx], a[idx], gj[idx], gi[idx]                
                all_b.append(b)
                all_a.append(a)
                all_gj.append(gj)
                all_gi.append(gi)
                all_anch.append(anch[i][idx])
                from_which_layer.append(torch.ones(size=(len(b),)) * i)
                
                fg_pred = pi[b, a, gj, gi]                
                p_obj.append(fg_pred[:, obj_idx:(obj_idx+1)])
                p_cls.append(fg_pred[:, (obj_idx+1):])
                
                grid = torch.stack([gi, gj], dim=1)
                pxy = (fg_pred[:, :2].sigmoid() * 2. - 0.5 + grid) * self.stride[i] #/ 8.
                #pwh = (fg_pred[:, 2:4].sigmoid() * 2) ** 2 * anch[i][idx] * self.stride[i] #/ 8.
                pw = self.wh_bin_sigmoid.forward(fg_pred[..., 2:(3+self.bin_count)].sigmoid()) * anch[i][idx][:, 0] * self.stride[i]
                ph = self.wh_bin_sigmoid.forward(fg_pred[..., (3+self.bin_count):obj_idx].sigmoid()) * anch[i][idx][:, 1] * self.stride[i]
                
                pxywh = torch.cat([pxy, pw.unsqueeze(1), ph.unsqueeze(1)], dim=-1)
                pxyxy = xywh2xyxy(pxywh)
                pxyxys.append(pxyxy)
            
            pxyxys = torch.cat(pxyxys, dim=0)
            if pxyxys.shape[0] == 0:
                continue
            p_obj = torch.cat(p_obj, dim=0)
            p_cls = torch.cat(p_cls, dim=0)
            from_which_layer = torch.cat(from_which_layer, dim=0)
            all_b = torch.cat(all_b, dim=0)
            all_a = torch.cat(all_a, dim=0)
            all_gj = torch.cat(all_gj, dim=0)
            all_gi = torch.cat(all_gi, dim=0)
            all_anch = torch.cat(all_anch, dim=0)
        
            pair_wise_iou = box_iou(txyxy, pxyxys)

            pair_wise_iou_loss = -torch.log(pair_wise_iou + 1e-8)

            top_k, _ = torch.topk(pair_wise_iou, min(10, pair_wise_iou.shape[1]), dim=1)
            dynamic_ks = torch.clamp(top_k.sum(1).int(), min=1)

            gt_cls_per_image = (
                F.one_hot(this_target[:, 1].to(torch.int64), self.nc)
                .float()
                .unsqueeze(1)
                .repeat(1, pxyxys.shape[0], 1)
            )

            num_gt = this_target.shape[0]            
            cls_preds_ = (
                p_cls.float().unsqueeze(0).repeat(num_gt, 1, 1).sigmoid_()
                * p_obj.unsqueeze(0).repeat(num_gt, 1, 1).sigmoid_()
            )

            y = cls_preds_.sqrt_()
            pair_wise_cls_loss = F.binary_cross_entropy_with_logits(
               torch.log(y/(1-y)) , gt_cls_per_image, reduction="none"
            ).sum(-1)
            del cls_preds_
        
            cost = (
                pair_wise_cls_loss
                + 3.0 * pair_wise_iou_loss
            )

            matching_matrix = torch.zeros_like(cost)

            for gt_idx in range(num_gt):
                _, pos_idx = torch.topk(
                    cost[gt_idx], k=dynamic_ks[gt_idx].item(), largest=False
                )
                matching_matrix[gt_idx][pos_idx] = 1.0

            del top_k, dynamic_ks
            anchor_matching_gt = matching_matrix.sum(0)
            if (anchor_matching_gt > 1).sum() > 0:
                _, cost_argmin = torch.min(cost[:, anchor_matching_gt > 1], dim=0)
                matching_matrix[:, anchor_matching_gt > 1] *= 0.0
                matching_matrix[cost_argmin, anchor_matching_gt > 1] = 1.0
            fg_mask_inboxes = matching_matrix.sum(0) > 0.0
            matched_gt_inds = matching_matrix[:, fg_mask_inboxes].argmax(0)
        
            from_which_layer = from_which_layer[fg_mask_inboxes]
            all_b = all_b[fg_mask_inboxes]
            all_a = all_a[fg_mask_inboxes]
            all_gj = all_gj[fg_mask_inboxes]
            all_gi = all_gi[fg_mask_inboxes]
            all_anch = all_anch[fg_mask_inboxes]
        
            this_target = this_target[matched_gt_inds]
        
            for i in range(nl):
                layer_idx = from_which_layer == i
                matching_bs[i].append(all_b[layer_idx])
                matching_as[i].append(all_a[layer_idx])
                matching_gjs[i].append(all_gj[layer_idx])
                matching_gis[i].append(all_gi[layer_idx])
                matching_targets[i].append(this_target[layer_idx])
                matching_anchs[i].append(all_anch[layer_idx])

        for i in range(nl):
            matching_bs[i] = torch.cat(matching_bs[i], dim=0)
            matching_as[i] = torch.cat(matching_as[i], dim=0)
            matching_gjs[i] = torch.cat(matching_gjs[i], dim=0)
            matching_gis[i] = torch.cat(matching_gis[i], dim=0)
            matching_targets[i] = torch.cat(matching_targets[i], dim=0)
            matching_anchs[i] = torch.cat(matching_anchs[i], dim=0)

        return matching_bs, matching_as, matching_gjs, matching_gis, matching_targets, matching_anchs       

    def find_3_positive(self, p, targets):
        # Build targets for compute_loss(), input targets(image,class,x,y,w,h)
        na, nt = self.na, targets.shape[0]  # number of anchors, targets
        indices, anch = [], []
        gain = torch.ones(7, device=targets.device)  # normalized to gridspace gain
        ai = torch.arange(na, device=targets.device).float().view(na, 1).repeat(1, nt)  # same as .repeat_interleave(nt)
        targets = torch.cat((targets.repeat(na, 1, 1), ai[:, :, None]), 2)  # append anchor indices

        g = 0.5  # bias
        off = torch.tensor([[0, 0],
                            [1, 0], [0, 1], [-1, 0], [0, -1],  # j,k,l,m
                            # [1, 1], [1, -1], [-1, 1], [-1, -1],  # jk,jm,lk,lm
                            ], device=targets.device).float() * g  # offsets

        for i in range(self.nl):
            anchors = self.anchors[i]
            gain[2:6] = torch.tensor(p[i].shape)[[3, 2, 3, 2]]  # xyxy gain

            # Match targets to anchors
            t = targets * gain
            if nt:
                # Matches
                r = t[:, :, 4:6] / anchors[:, None]  # wh ratio
                j = torch.max(r, 1. / r).max(2)[0] < self.hyp['anchor_t']  # compare
                # j = wh_iou(anchors, t[:, 4:6]) > model.hyp['iou_t']  # iou(3,n)=wh_iou(anchors(3,2), gwh(n,2))
                t = t[j]  # filter

                # Offsets
                gxy = t[:, 2:4]  # grid xy
                gxi = gain[[2, 3]] - gxy  # inverse
                j, k = ((gxy % 1. < g) & (gxy > 1.)).T
                l, m = ((gxi % 1. < g) & (gxi > 1.)).T
                j = torch.stack((torch.ones_like(j), j, k, l, m))
                t = t.repeat((5, 1, 1))[j]
                offsets = (torch.zeros_like(gxy)[None] + off[:, None])[j]
            else:
                t = targets[0]
                offsets = 0

            # Define
            b, c = t[:, :2].long().T  # image, class
            gxy = t[:, 2:4]  # grid xy
            gwh = t[:, 4:6]  # grid wh
            gij = (gxy - offsets).long()
            gi, gj = gij.T  # grid xy indices

            # Append
            a = t[:, 6].long()  # anchor indices
            indices.append((b, a, gj.clamp_(0, gain[3] - 1), gi.clamp_(0, gain[2] - 1)))  # image, anchor, grid indices
            anch.append(anchors[a])  # anchors

        return indices, anch