Hsueh1001 commited on
Commit
3c60371
1 Parent(s): f52d286

Update config.json

Browse files
Files changed (1) hide show
  1. config.json +181 -1212
config.json CHANGED
@@ -1,1240 +1,209 @@
1
  {
2
- "_name_or_path": "C:\\Users\\eatin\\Desktop\\ML\\yolov7",
3
- "anchors": [
4
- [
5
- 12,
6
- 16,
7
- 19,
8
- 36,
9
- 40,
10
- 28
11
- ],
12
- [
13
- 36,
14
- 75,
15
- 76,
16
- 55,
17
- 72,
18
- 146
19
- ],
20
- [
21
- 142,
22
- 110,
23
- 192,
24
- 243,
25
- 459,
26
- 401
27
- ]
28
- ],
29
  "architectures": [
30
  "YolosForObjectDetection"
31
  ],
32
  "attention_probs_dropout_prob": 0.0,
33
  "auxiliary_loss": false,
34
- "backbone": [
35
- [
36
- -1,
37
- 1,
38
- "Conv",
39
- [
40
- 32,
41
- 3,
42
- 1
43
- ]
44
- ],
45
- [
46
- -1,
47
- 1,
48
- "Conv",
49
- [
50
- 64,
51
- 3,
52
- 2
53
- ]
54
- ],
55
- [
56
- -1,
57
- 1,
58
- "Conv",
59
- [
60
- 64,
61
- 3,
62
- 1
63
- ]
64
- ],
65
- [
66
- -1,
67
- 1,
68
- "Conv",
69
- [
70
- 128,
71
- 3,
72
- 2
73
- ]
74
- ],
75
- [
76
- -1,
77
- 1,
78
- "Conv",
79
- [
80
- 64,
81
- 1,
82
- 1
83
- ]
84
- ],
85
- [
86
- -2,
87
- 1,
88
- "Conv",
89
- [
90
- 64,
91
- 1,
92
- 1
93
- ]
94
- ],
95
- [
96
- -1,
97
- 1,
98
- "Conv",
99
- [
100
- 64,
101
- 3,
102
- 1
103
- ]
104
- ],
105
- [
106
- -1,
107
- 1,
108
- "Conv",
109
- [
110
- 64,
111
- 3,
112
- 1
113
- ]
114
- ],
115
- [
116
- -1,
117
- 1,
118
- "Conv",
119
- [
120
- 64,
121
- 3,
122
- 1
123
- ]
124
- ],
125
- [
126
- -1,
127
- 1,
128
- "Conv",
129
- [
130
- 64,
131
- 3,
132
- 1
133
- ]
134
- ],
135
- [
136
- [
137
- -1,
138
- -3,
139
- -5,
140
- -6
141
- ],
142
- 1,
143
- "Concat",
144
- [
145
- 1
146
- ]
147
- ],
148
- [
149
- -1,
150
- 1,
151
- "Conv",
152
- [
153
- 256,
154
- 1,
155
- 1
156
- ]
157
- ],
158
- [
159
- -1,
160
- 1,
161
- "MP",
162
- []
163
- ],
164
- [
165
- -1,
166
- 1,
167
- "Conv",
168
- [
169
- 128,
170
- 1,
171
- 1
172
- ]
173
- ],
174
- [
175
- -3,
176
- 1,
177
- "Conv",
178
- [
179
- 128,
180
- 1,
181
- 1
182
- ]
183
- ],
184
- [
185
- -1,
186
- 1,
187
- "Conv",
188
- [
189
- 128,
190
- 3,
191
- 2
192
- ]
193
- ],
194
- [
195
- [
196
- -1,
197
- -3
198
- ],
199
- 1,
200
- "Concat",
201
- [
202
- 1
203
- ]
204
- ],
205
- [
206
- -1,
207
- 1,
208
- "Conv",
209
- [
210
- 128,
211
- 1,
212
- 1
213
- ]
214
- ],
215
- [
216
- -2,
217
- 1,
218
- "Conv",
219
- [
220
- 128,
221
- 1,
222
- 1
223
- ]
224
- ],
225
- [
226
- -1,
227
- 1,
228
- "Conv",
229
- [
230
- 128,
231
- 3,
232
- 1
233
- ]
234
- ],
235
- [
236
- -1,
237
- 1,
238
- "Conv",
239
- [
240
- 128,
241
- 3,
242
- 1
243
- ]
244
- ],
245
- [
246
- -1,
247
- 1,
248
- "Conv",
249
- [
250
- 128,
251
- 3,
252
- 1
253
- ]
254
- ],
255
- [
256
- -1,
257
- 1,
258
- "Conv",
259
- [
260
- 128,
261
- 3,
262
- 1
263
- ]
264
- ],
265
- [
266
- [
267
- -1,
268
- -3,
269
- -5,
270
- -6
271
- ],
272
- 1,
273
- "Concat",
274
- [
275
- 1
276
- ]
277
- ],
278
- [
279
- -1,
280
- 1,
281
- "Conv",
282
- [
283
- 512,
284
- 1,
285
- 1
286
- ]
287
- ],
288
- [
289
- -1,
290
- 1,
291
- "MP",
292
- []
293
- ],
294
- [
295
- -1,
296
- 1,
297
- "Conv",
298
- [
299
- 256,
300
- 1,
301
- 1
302
- ]
303
- ],
304
- [
305
- -3,
306
- 1,
307
- "Conv",
308
- [
309
- 256,
310
- 1,
311
- 1
312
- ]
313
- ],
314
- [
315
- -1,
316
- 1,
317
- "Conv",
318
- [
319
- 256,
320
- 3,
321
- 2
322
- ]
323
- ],
324
- [
325
- [
326
- -1,
327
- -3
328
- ],
329
- 1,
330
- "Concat",
331
- [
332
- 1
333
- ]
334
- ],
335
- [
336
- -1,
337
- 1,
338
- "Conv",
339
- [
340
- 256,
341
- 1,
342
- 1
343
- ]
344
- ],
345
- [
346
- -2,
347
- 1,
348
- "Conv",
349
- [
350
- 256,
351
- 1,
352
- 1
353
- ]
354
- ],
355
- [
356
- -1,
357
- 1,
358
- "Conv",
359
- [
360
- 256,
361
- 3,
362
- 1
363
- ]
364
- ],
365
- [
366
- -1,
367
- 1,
368
- "Conv",
369
- [
370
- 256,
371
- 3,
372
- 1
373
- ]
374
- ],
375
- [
376
- -1,
377
- 1,
378
- "Conv",
379
- [
380
- 256,
381
- 3,
382
- 1
383
- ]
384
- ],
385
- [
386
- -1,
387
- 1,
388
- "Conv",
389
- [
390
- 256,
391
- 3,
392
- 1
393
- ]
394
- ],
395
- [
396
- [
397
- -1,
398
- -3,
399
- -5,
400
- -6
401
- ],
402
- 1,
403
- "Concat",
404
- [
405
- 1
406
- ]
407
- ],
408
- [
409
- -1,
410
- 1,
411
- "Conv",
412
- [
413
- 1024,
414
- 1,
415
- 1
416
- ]
417
- ],
418
- [
419
- -1,
420
- 1,
421
- "MP",
422
- []
423
- ],
424
- [
425
- -1,
426
- 1,
427
- "Conv",
428
- [
429
- 512,
430
- 1,
431
- 1
432
- ]
433
- ],
434
- [
435
- -3,
436
- 1,
437
- "Conv",
438
- [
439
- 512,
440
- 1,
441
- 1
442
- ]
443
- ],
444
- [
445
- -1,
446
- 1,
447
- "Conv",
448
- [
449
- 512,
450
- 3,
451
- 2
452
- ]
453
- ],
454
- [
455
- [
456
- -1,
457
- -3
458
- ],
459
- 1,
460
- "Concat",
461
- [
462
- 1
463
- ]
464
- ],
465
- [
466
- -1,
467
- 1,
468
- "Conv",
469
- [
470
- 256,
471
- 1,
472
- 1
473
- ]
474
- ],
475
- [
476
- -2,
477
- 1,
478
- "Conv",
479
- [
480
- 256,
481
- 1,
482
- 1
483
- ]
484
- ],
485
- [
486
- -1,
487
- 1,
488
- "Conv",
489
- [
490
- 256,
491
- 3,
492
- 1
493
- ]
494
- ],
495
- [
496
- -1,
497
- 1,
498
- "Conv",
499
- [
500
- 256,
501
- 3,
502
- 1
503
- ]
504
- ],
505
- [
506
- -1,
507
- 1,
508
- "Conv",
509
- [
510
- 256,
511
- 3,
512
- 1
513
- ]
514
- ],
515
- [
516
- -1,
517
- 1,
518
- "Conv",
519
- [
520
- 256,
521
- 3,
522
- 1
523
- ]
524
- ],
525
- [
526
- [
527
- -1,
528
- -3,
529
- -5,
530
- -6
531
- ],
532
- 1,
533
- "Concat",
534
- [
535
- 1
536
- ]
537
- ],
538
- [
539
- -1,
540
- 1,
541
- "Conv",
542
- [
543
- 1024,
544
- 1,
545
- 1
546
- ]
547
- ]
548
- ],
549
  "bbox_cost": 5,
550
  "bbox_loss_coefficient": 5,
551
  "class_cost": 1,
552
- "conf_threshold": 0.1,
553
- "depth_multiple": 1.0,
554
  "eos_coefficient": 0.1,
555
  "giou_cost": 2,
556
  "giou_loss_coefficient": 2,
557
- "head": [
558
- [
559
- -1,
560
- 1,
561
- "SPPCSPC",
562
- [
563
- 512
564
- ]
565
- ],
566
- [
567
- -1,
568
- 1,
569
- "Conv",
570
- [
571
- 256,
572
- 1,
573
- 1
574
- ]
575
- ],
576
- [
577
- -1,
578
- 1,
579
- "nn.Upsample",
580
- [
581
- "None",
582
- 2,
583
- "nearest"
584
- ]
585
- ],
586
- [
587
- 37,
588
- 1,
589
- "Conv",
590
- [
591
- 256,
592
- 1,
593
- 1
594
- ]
595
- ],
596
- [
597
- [
598
- -1,
599
- -2
600
- ],
601
- 1,
602
- "Concat",
603
- [
604
- 1
605
- ]
606
- ],
607
- [
608
- -1,
609
- 1,
610
- "Conv",
611
- [
612
- 256,
613
- 1,
614
- 1
615
- ]
616
- ],
617
- [
618
- -2,
619
- 1,
620
- "Conv",
621
- [
622
- 256,
623
- 1,
624
- 1
625
- ]
626
- ],
627
- [
628
- -1,
629
- 1,
630
- "Conv",
631
- [
632
- 128,
633
- 3,
634
- 1
635
- ]
636
- ],
637
- [
638
- -1,
639
- 1,
640
- "Conv",
641
- [
642
- 128,
643
- 3,
644
- 1
645
- ]
646
- ],
647
- [
648
- -1,
649
- 1,
650
- "Conv",
651
- [
652
- 128,
653
- 3,
654
- 1
655
- ]
656
- ],
657
- [
658
- -1,
659
- 1,
660
- "Conv",
661
- [
662
- 128,
663
- 3,
664
- 1
665
- ]
666
- ],
667
- [
668
- [
669
- -1,
670
- -2,
671
- -3,
672
- -4,
673
- -5,
674
- -6
675
- ],
676
- 1,
677
- "Concat",
678
- [
679
- 1
680
- ]
681
- ],
682
- [
683
- -1,
684
- 1,
685
- "Conv",
686
- [
687
- 256,
688
- 1,
689
- 1
690
- ]
691
- ],
692
- [
693
- -1,
694
- 1,
695
- "Conv",
696
- [
697
- 128,
698
- 1,
699
- 1
700
- ]
701
- ],
702
- [
703
- -1,
704
- 1,
705
- "nn.Upsample",
706
- [
707
- "None",
708
- 2,
709
- "nearest"
710
- ]
711
- ],
712
- [
713
- 24,
714
- 1,
715
- "Conv",
716
- [
717
- 128,
718
- 1,
719
- 1
720
- ]
721
- ],
722
- [
723
- [
724
- -1,
725
- -2
726
- ],
727
- 1,
728
- "Concat",
729
- [
730
- 1
731
- ]
732
- ],
733
- [
734
- -1,
735
- 1,
736
- "Conv",
737
- [
738
- 128,
739
- 1,
740
- 1
741
- ]
742
- ],
743
- [
744
- -2,
745
- 1,
746
- "Conv",
747
- [
748
- 128,
749
- 1,
750
- 1
751
- ]
752
- ],
753
- [
754
- -1,
755
- 1,
756
- "Conv",
757
- [
758
- 64,
759
- 3,
760
- 1
761
- ]
762
- ],
763
- [
764
- -1,
765
- 1,
766
- "Conv",
767
- [
768
- 64,
769
- 3,
770
- 1
771
- ]
772
- ],
773
- [
774
- -1,
775
- 1,
776
- "Conv",
777
- [
778
- 64,
779
- 3,
780
- 1
781
- ]
782
- ],
783
- [
784
- -1,
785
- 1,
786
- "Conv",
787
- [
788
- 64,
789
- 3,
790
- 1
791
- ]
792
- ],
793
- [
794
- [
795
- -1,
796
- -2,
797
- -3,
798
- -4,
799
- -5,
800
- -6
801
- ],
802
- 1,
803
- "Concat",
804
- [
805
- 1
806
- ]
807
- ],
808
- [
809
- -1,
810
- 1,
811
- "Conv",
812
- [
813
- 128,
814
- 1,
815
- 1
816
- ]
817
- ],
818
- [
819
- -1,
820
- 1,
821
- "MP",
822
- []
823
- ],
824
- [
825
- -1,
826
- 1,
827
- "Conv",
828
- [
829
- 128,
830
- 1,
831
- 1
832
- ]
833
- ],
834
- [
835
- -3,
836
- 1,
837
- "Conv",
838
- [
839
- 128,
840
- 1,
841
- 1
842
- ]
843
- ],
844
- [
845
- -1,
846
- 1,
847
- "Conv",
848
- [
849
- 128,
850
- 3,
851
- 2
852
- ]
853
- ],
854
- [
855
- [
856
- -1,
857
- -3,
858
- 63
859
- ],
860
- 1,
861
- "Concat",
862
- [
863
- 1
864
- ]
865
- ],
866
- [
867
- -1,
868
- 1,
869
- "Conv",
870
- [
871
- 256,
872
- 1,
873
- 1
874
- ]
875
- ],
876
- [
877
- -2,
878
- 1,
879
- "Conv",
880
- [
881
- 256,
882
- 1,
883
- 1
884
- ]
885
- ],
886
- [
887
- -1,
888
- 1,
889
- "Conv",
890
- [
891
- 128,
892
- 3,
893
- 1
894
- ]
895
- ],
896
- [
897
- -1,
898
- 1,
899
- "Conv",
900
- [
901
- 128,
902
- 3,
903
- 1
904
- ]
905
- ],
906
- [
907
- -1,
908
- 1,
909
- "Conv",
910
- [
911
- 128,
912
- 3,
913
- 1
914
- ]
915
- ],
916
- [
917
- -1,
918
- 1,
919
- "Conv",
920
- [
921
- 128,
922
- 3,
923
- 1
924
- ]
925
- ],
926
- [
927
- [
928
- -1,
929
- -2,
930
- -3,
931
- -4,
932
- -5,
933
- -6
934
- ],
935
- 1,
936
- "Concat",
937
- [
938
- 1
939
- ]
940
- ],
941
- [
942
- -1,
943
- 1,
944
- "Conv",
945
- [
946
- 256,
947
- 1,
948
- 1
949
- ]
950
- ],
951
- [
952
- -1,
953
- 1,
954
- "MP",
955
- []
956
- ],
957
- [
958
- -1,
959
- 1,
960
- "Conv",
961
- [
962
- 256,
963
- 1,
964
- 1
965
- ]
966
- ],
967
- [
968
- -3,
969
- 1,
970
- "Conv",
971
- [
972
- 256,
973
- 1,
974
- 1
975
- ]
976
- ],
977
- [
978
- -1,
979
- 1,
980
- "Conv",
981
- [
982
- 256,
983
- 3,
984
- 2
985
- ]
986
- ],
987
- [
988
- [
989
- -1,
990
- -3,
991
- 51
992
- ],
993
- 1,
994
- "Concat",
995
- [
996
- 1
997
- ]
998
- ],
999
- [
1000
- -1,
1001
- 1,
1002
- "Conv",
1003
- [
1004
- 512,
1005
- 1,
1006
- 1
1007
- ]
1008
- ],
1009
- [
1010
- -2,
1011
- 1,
1012
- "Conv",
1013
- [
1014
- 512,
1015
- 1,
1016
- 1
1017
- ]
1018
- ],
1019
- [
1020
- -1,
1021
- 1,
1022
- "Conv",
1023
- [
1024
- 256,
1025
- 3,
1026
- 1
1027
- ]
1028
- ],
1029
- [
1030
- -1,
1031
- 1,
1032
- "Conv",
1033
- [
1034
- 256,
1035
- 3,
1036
- 1
1037
- ]
1038
- ],
1039
- [
1040
- -1,
1041
- 1,
1042
- "Conv",
1043
- [
1044
- 256,
1045
- 3,
1046
- 1
1047
- ]
1048
- ],
1049
- [
1050
- -1,
1051
- 1,
1052
- "Conv",
1053
- [
1054
- 256,
1055
- 3,
1056
- 1
1057
- ]
1058
- ],
1059
- [
1060
- [
1061
- -1,
1062
- -2,
1063
- -3,
1064
- -4,
1065
- -5,
1066
- -6
1067
- ],
1068
- 1,
1069
- "Concat",
1070
- [
1071
- 1
1072
- ]
1073
- ],
1074
- [
1075
- -1,
1076
- 1,
1077
- "Conv",
1078
- [
1079
- 512,
1080
- 1,
1081
- 1
1082
- ]
1083
- ],
1084
- [
1085
- 75,
1086
- 1,
1087
- "RepConv",
1088
- [
1089
- 256,
1090
- 3,
1091
- 1
1092
- ]
1093
- ],
1094
- [
1095
- 88,
1096
- 1,
1097
- "RepConv",
1098
- [
1099
- 512,
1100
- 3,
1101
- 1
1102
- ]
1103
- ],
1104
- [
1105
- 101,
1106
- 1,
1107
- "RepConv",
1108
- [
1109
- 1024,
1110
- 3,
1111
- 1
1112
- ]
1113
- ],
1114
- [
1115
- [
1116
- 102,
1117
- 103,
1118
- 104
1119
- ],
1120
- 1,
1121
- "IDetect",
1122
- [
1123
- "nc",
1124
- "anchors"
1125
- ]
1126
- ]
1127
- ],
1128
  "hidden_act": "gelu",
1129
  "hidden_dropout_prob": 0.0,
1130
  "hidden_size": 768,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1131
  "image_size": [
1132
- 512,
1133
- 864
1134
  ],
1135
  "initializer_range": 0.02,
1136
- "input_size": [
1137
- 640,
1138
- 640
1139
- ],
1140
  "intermediate_size": 3072,
1141
- "iou_threshold": 0.1,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1142
  "layer_norm_eps": 1e-12,
1143
  "model_type": "yolos",
1144
- "names": [
1145
- "person",
1146
- "bicycle",
1147
- "car",
1148
- "motorcycle",
1149
- "airplane",
1150
- "bus",
1151
- "train",
1152
- "truck",
1153
- "boat",
1154
- "traffic light",
1155
- "fire hydrant",
1156
- "stop sign",
1157
- "parking meter",
1158
- "bench",
1159
- "bird",
1160
- "cat",
1161
- "dog",
1162
- "horse",
1163
- "sheep",
1164
- "cow",
1165
- "elephant",
1166
- "bear",
1167
- "zebra",
1168
- "giraffe",
1169
- "backpack",
1170
- "umbrella",
1171
- "handbag",
1172
- "tie",
1173
- "suitcase",
1174
- "frisbee",
1175
- "skis",
1176
- "snowboard",
1177
- "sports ball",
1178
- "kite",
1179
- "baseball bat",
1180
- "baseball glove",
1181
- "skateboard",
1182
- "surfboard",
1183
- "tennis racket",
1184
- "bottle",
1185
- "wine glass",
1186
- "cup",
1187
- "fork",
1188
- "knife",
1189
- "spoon",
1190
- "bowl",
1191
- "banana",
1192
- "apple",
1193
- "sandwich",
1194
- "orange",
1195
- "broccoli",
1196
- "carrot",
1197
- "hot dog",
1198
- "pizza",
1199
- "donut",
1200
- "cake",
1201
- "chair",
1202
- "couch",
1203
- "potted plant",
1204
- "bed",
1205
- "dining table",
1206
- "toilet",
1207
- "tv",
1208
- "laptop",
1209
- "mouse",
1210
- "remote",
1211
- "keyboard",
1212
- "cell phone",
1213
- "microwave",
1214
- "oven",
1215
- "toaster",
1216
- "sink",
1217
- "refrigerator",
1218
- "book",
1219
- "clock",
1220
- "vase",
1221
- "scissors",
1222
- "teddy bear",
1223
- "hair drier",
1224
- "toothbrush"
1225
- ],
1226
- "nc": 80,
1227
- "nms_threshold": 0.1,
1228
  "num_attention_heads": 12,
1229
  "num_channels": 3,
1230
  "num_detection_tokens": 100,
1231
  "num_hidden_layers": 12,
1232
  "patch_size": 16,
1233
  "qkv_bias": true,
1234
- "save_txt": true,
1235
  "torch_dtype": "float32",
1236
- "transformers_version": "4.36.2",
1237
- "use_mid_position_embeddings": true,
1238
- "weights": "yolov7.pt",
1239
- "width_multiple": 1.0
1240
- }
 
1
  {
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2
  "architectures": [
3
  "YolosForObjectDetection"
4
  ],
5
  "attention_probs_dropout_prob": 0.0,
6
  "auxiliary_loss": false,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7
  "bbox_cost": 5,
8
  "bbox_loss_coefficient": 5,
9
  "class_cost": 1,
 
 
10
  "eos_coefficient": 0.1,
11
  "giou_cost": 2,
12
  "giou_loss_coefficient": 2,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
13
  "hidden_act": "gelu",
14
  "hidden_dropout_prob": 0.0,
15
  "hidden_size": 768,
16
+ "id2label": {
17
+ "0": "N/A",
18
+ "1": "person",
19
+ "2": "bicycle",
20
+ "3": "car",
21
+ "4": "motorcycle",
22
+ "5": "airplane",
23
+ "6": "bus",
24
+ "7": "train",
25
+ "8": "truck",
26
+ "9": "boat",
27
+ "10": "traffic light",
28
+ "11": "fire hydrant",
29
+ "12": "N/A",
30
+ "13": "stop sign",
31
+ "14": "parking meter",
32
+ "15": "bench",
33
+ "16": "bird",
34
+ "17": "cat",
35
+ "18": "dog",
36
+ "19": "horse",
37
+ "20": "sheep",
38
+ "21": "cow",
39
+ "22": "elephant",
40
+ "23": "bear",
41
+ "24": "zebra",
42
+ "25": "giraffe",
43
+ "26": "N/A",
44
+ "27": "backpack",
45
+ "28": "umbrella",
46
+ "29": "N/A",
47
+ "30": "N/A",
48
+ "31": "handbag",
49
+ "32": "tie",
50
+ "33": "suitcase",
51
+ "34": "frisbee",
52
+ "35": "skis",
53
+ "36": "snowboard",
54
+ "37": "sports ball",
55
+ "38": "kite",
56
+ "39": "baseball bat",
57
+ "40": "baseball glove",
58
+ "41": "skateboard",
59
+ "42": "surfboard",
60
+ "43": "tennis racket",
61
+ "44": "bottle",
62
+ "45": "N/A",
63
+ "46": "wine glass",
64
+ "47": "cup",
65
+ "48": "fork",
66
+ "49": "knife",
67
+ "50": "spoon",
68
+ "51": "bowl",
69
+ "52": "banana",
70
+ "53": "apple",
71
+ "54": "sandwich",
72
+ "55": "orange",
73
+ "56": "broccoli",
74
+ "57": "carrot",
75
+ "58": "hot dog",
76
+ "59": "pizza",
77
+ "60": "donut",
78
+ "61": "cake",
79
+ "62": "chair",
80
+ "63": "couch",
81
+ "64": "potted plant",
82
+ "65": "bed",
83
+ "66": "N/A",
84
+ "67": "dining table",
85
+ "68": "N/A",
86
+ "69": "N/A",
87
+ "70": "toilet",
88
+ "71": "N/A",
89
+ "72": "tv",
90
+ "73": "laptop",
91
+ "74": "mouse",
92
+ "75": "remote",
93
+ "76": "keyboard",
94
+ "77": "cell phone",
95
+ "78": "microwave",
96
+ "79": "oven",
97
+ "80": "toaster",
98
+ "81": "sink",
99
+ "82": "refrigerator",
100
+ "83": "N/A",
101
+ "84": "book",
102
+ "85": "clock",
103
+ "86": "vase",
104
+ "87": "scissors",
105
+ "88": "teddy bear",
106
+ "89": "hair drier",
107
+ "90": "toothbrush"
108
+ },
109
  "image_size": [
110
+ 800,
111
+ 1344
112
  ],
113
  "initializer_range": 0.02,
 
 
 
 
114
  "intermediate_size": 3072,
115
+ "label2id": {
116
+ "N/A": 83,
117
+ "airplane": 5,
118
+ "apple": 53,
119
+ "backpack": 27,
120
+ "banana": 52,
121
+ "baseball bat": 39,
122
+ "baseball glove": 40,
123
+ "bear": 23,
124
+ "bed": 65,
125
+ "bench": 15,
126
+ "bicycle": 2,
127
+ "bird": 16,
128
+ "boat": 9,
129
+ "book": 84,
130
+ "bottle": 44,
131
+ "bowl": 51,
132
+ "broccoli": 56,
133
+ "bus": 6,
134
+ "cake": 61,
135
+ "car": 3,
136
+ "carrot": 57,
137
+ "cat": 17,
138
+ "cell phone": 77,
139
+ "chair": 62,
140
+ "clock": 85,
141
+ "couch": 63,
142
+ "cow": 21,
143
+ "cup": 47,
144
+ "dining table": 67,
145
+ "dog": 18,
146
+ "donut": 60,
147
+ "elephant": 22,
148
+ "fire hydrant": 11,
149
+ "fork": 48,
150
+ "frisbee": 34,
151
+ "giraffe": 25,
152
+ "hair drier": 89,
153
+ "handbag": 31,
154
+ "horse": 19,
155
+ "hot dog": 58,
156
+ "keyboard": 76,
157
+ "kite": 38,
158
+ "knife": 49,
159
+ "laptop": 73,
160
+ "microwave": 78,
161
+ "motorcycle": 4,
162
+ "mouse": 74,
163
+ "orange": 55,
164
+ "oven": 79,
165
+ "parking meter": 14,
166
+ "person": 1,
167
+ "pizza": 59,
168
+ "potted plant": 64,
169
+ "refrigerator": 82,
170
+ "remote": 75,
171
+ "sandwich": 54,
172
+ "scissors": 87,
173
+ "sheep": 20,
174
+ "sink": 81,
175
+ "skateboard": 41,
176
+ "skis": 35,
177
+ "snowboard": 36,
178
+ "spoon": 50,
179
+ "sports ball": 37,
180
+ "stop sign": 13,
181
+ "suitcase": 33,
182
+ "surfboard": 42,
183
+ "teddy bear": 88,
184
+ "tennis racket": 43,
185
+ "tie": 32,
186
+ "toaster": 80,
187
+ "toilet": 70,
188
+ "toothbrush": 90,
189
+ "traffic light": 10,
190
+ "train": 7,
191
+ "truck": 8,
192
+ "tv": 72,
193
+ "umbrella": 28,
194
+ "vase": 86,
195
+ "wine glass": 46,
196
+ "zebra": 24
197
+ },
198
  "layer_norm_eps": 1e-12,
199
  "model_type": "yolos",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
200
  "num_attention_heads": 12,
201
  "num_channels": 3,
202
  "num_detection_tokens": 100,
203
  "num_hidden_layers": 12,
204
  "patch_size": 16,
205
  "qkv_bias": true,
 
206
  "torch_dtype": "float32",
207
+ "transformers_version": "4.19.0.dev0",
208
+ "use_mid_position_embeddings": true
209
+ }