Hugging Face's logo

facebook
/
sapiens2-pose-5b

Keypoint Detection
sapiens
Safetensors
sapiens2
human-centric
pose
Model card Files
xet
 Community
3

Add HF-compatible model.safetensors, config.json, and preprocessor_config.json

#1
by guarin HF Staff - opened
base: refs/heads/main
←
from: refs/pr/1
Discussion Files changed
+1451
-0
Files changed (3) hide show
  1. config.json +1419 -0
  2. model.safetensors +3 -0
  3. preprocessor_config.json +29 -0
config.json ADDED
@@ -0,0 +1,1419 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "apply_layernorm": true,
3
+ "architectures": [
4
+ "Sapiens2ForPoseEstimation"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "drop_path_rate": 0.0,
8
+ "flip_pairs": [
9
+ [
10
+ 1,
11
+ 2
12
+ ],
13
+ [
14
+ 3,
15
+ 4
16
+ ],
17
+ [
18
+ 5,
19
+ 6
20
+ ],
21
+ [
22
+ 7,
23
+ 8
24
+ ],
25
+ [
26
+ 9,
27
+ 10
28
+ ],
29
+ [
30
+ 11,
31
+ 12
32
+ ],
33
+ [
34
+ 13,
35
+ 14
36
+ ],
37
+ [
38
+ 15,
39
+ 18
40
+ ],
41
+ [
42
+ 16,
43
+ 19
44
+ ],
45
+ [
46
+ 17,
47
+ 20
48
+ ],
49
+ [
50
+ 21,
51
+ 42
52
+ ],
53
+ [
54
+ 22,
55
+ 43
56
+ ],
57
+ [
58
+ 23,
59
+ 44
60
+ ],
61
+ [
62
+ 24,
63
+ 45
64
+ ],
65
+ [
66
+ 25,
67
+ 46
68
+ ],
69
+ [
70
+ 26,
71
+ 47
72
+ ],
73
+ [
74
+ 27,
75
+ 48
76
+ ],
77
+ [
78
+ 28,
79
+ 49
80
+ ],
81
+ [
82
+ 29,
83
+ 50
84
+ ],
85
+ [
86
+ 30,
87
+ 51
88
+ ],
89
+ [
90
+ 31,
91
+ 52
92
+ ],
93
+ [
94
+ 32,
95
+ 53
96
+ ],
97
+ [
98
+ 33,
99
+ 54
100
+ ],
101
+ [
102
+ 34,
103
+ 55
104
+ ],
105
+ [
106
+ 35,
107
+ 56
108
+ ],
109
+ [
110
+ 36,
111
+ 57
112
+ ],
113
+ [
114
+ 37,
115
+ 58
116
+ ],
117
+ [
118
+ 38,
119
+ 59
120
+ ],
121
+ [
122
+ 39,
123
+ 60
124
+ ],
125
+ [
126
+ 40,
127
+ 61
128
+ ],
129
+ [
130
+ 41,
131
+ 62
132
+ ],
133
+ [
134
+ 63,
135
+ 64
136
+ ],
137
+ [
138
+ 65,
139
+ 66
140
+ ],
141
+ [
142
+ 67,
143
+ 68
144
+ ],
145
+ [
146
+ 78,
147
+ 87
148
+ ],
149
+ [
150
+ 79,
151
+ 88
152
+ ],
153
+ [
154
+ 80,
155
+ 89
156
+ ],
157
+ [
158
+ 81,
159
+ 90
160
+ ],
161
+ [
162
+ 82,
163
+ 91
164
+ ],
165
+ [
166
+ 83,
167
+ 93
168
+ ],
169
+ [
170
+ 84,
171
+ 92
172
+ ],
173
+ [
174
+ 85,
175
+ 95
176
+ ],
177
+ [
178
+ 86,
179
+ 94
180
+ ],
181
+ [
182
+ 96,
183
+ 120
184
+ ],
185
+ [
186
+ 97,
187
+ 121
188
+ ],
189
+ [
190
+ 98,
191
+ 122
192
+ ],
193
+ [
194
+ 99,
195
+ 123
196
+ ],
197
+ [
198
+ 100,
199
+ 124
200
+ ],
201
+ [
202
+ 101,
203
+ 125
204
+ ],
205
+ [
206
+ 102,
207
+ 126
208
+ ],
209
+ [
210
+ 103,
211
+ 127
212
+ ],
213
+ [
214
+ 104,
215
+ 128
216
+ ],
217
+ [
218
+ 105,
219
+ 129
220
+ ],
221
+ [
222
+ 106,
223
+ 130
224
+ ],
225
+ [
226
+ 107,
227
+ 131
228
+ ],
229
+ [
230
+ 108,
231
+ 132
232
+ ],
233
+ [
234
+ 109,
235
+ 133
236
+ ],
237
+ [
238
+ 110,
239
+ 134
240
+ ],
241
+ [
242
+ 111,
243
+ 135
244
+ ],
245
+ [
246
+ 112,
247
+ 136
248
+ ],
249
+ [
250
+ 113,
251
+ 137
252
+ ],
253
+ [
254
+ 114,
255
+ 138
256
+ ],
257
+ [
258
+ 115,
259
+ 139
260
+ ],
261
+ [
262
+ 116,
263
+ 140
264
+ ],
265
+ [
266
+ 117,
267
+ 141
268
+ ],
269
+ [
270
+ 118,
271
+ 142
272
+ ],
273
+ [
274
+ 119,
275
+ 143
276
+ ],
277
+ [
278
+ 144,
279
+ 161
280
+ ],
281
+ [
282
+ 145,
283
+ 162
284
+ ],
285
+ [
286
+ 146,
287
+ 163
288
+ ],
289
+ [
290
+ 147,
291
+ 164
292
+ ],
293
+ [
294
+ 148,
295
+ 165
296
+ ],
297
+ [
298
+ 149,
299
+ 166
300
+ ],
301
+ [
302
+ 150,
303
+ 167
304
+ ],
305
+ [
306
+ 151,
307
+ 168
308
+ ],
309
+ [
310
+ 152,
311
+ 169
312
+ ],
313
+ [
314
+ 153,
315
+ 170
316
+ ],
317
+ [
318
+ 154,
319
+ 171
320
+ ],
321
+ [
322
+ 155,
323
+ 172
324
+ ],
325
+ [
326
+ 156,
327
+ 173
328
+ ],
329
+ [
330
+ 157,
331
+ 174
332
+ ],
333
+ [
334
+ 158,
335
+ 175
336
+ ],
337
+ [
338
+ 159,
339
+ 176
340
+ ],
341
+ [
342
+ 160,
343
+ 177
344
+ ],
345
+ [
346
+ 180,
347
+ 181
348
+ ],
349
+ [
350
+ 182,
351
+ 185
352
+ ],
353
+ [
354
+ 183,
355
+ 186
356
+ ],
357
+ [
358
+ 184,
359
+ 187
360
+ ],
361
+ [
362
+ 188,
363
+ 189
364
+ ],
365
+ [
366
+ 192,
367
+ 193
368
+ ],
369
+ [
370
+ 194,
371
+ 195
372
+ ],
373
+ [
374
+ 196,
375
+ 199
376
+ ],
377
+ [
378
+ 197,
379
+ 198
380
+ ],
381
+ [
382
+ 200,
383
+ 203
384
+ ],
385
+ [
386
+ 201,
387
+ 202
388
+ ],
389
+ [
390
+ 204,
391
+ 205
392
+ ],
393
+ [
394
+ 208,
395
+ 209
396
+ ],
397
+ [
398
+ 210,
399
+ 211
400
+ ],
401
+ [
402
+ 212,
403
+ 215
404
+ ],
405
+ [
406
+ 213,
407
+ 214
408
+ ],
409
+ [
410
+ 216,
411
+ 219
412
+ ],
413
+ [
414
+ 217,
415
+ 218
416
+ ],
417
+ [
418
+ 220,
419
+ 246
420
+ ],
421
+ [
422
+ 221,
423
+ 247
424
+ ],
425
+ [
426
+ 222,
427
+ 248
428
+ ],
429
+ [
430
+ 223,
431
+ 249
432
+ ],
433
+ [
434
+ 224,
435
+ 250
436
+ ],
437
+ [
438
+ 225,
439
+ 251
440
+ ],
441
+ [
442
+ 226,
443
+ 252
444
+ ],
445
+ [
446
+ 227,
447
+ 253
448
+ ],
449
+ [
450
+ 228,
451
+ 254
452
+ ],
453
+ [
454
+ 229,
455
+ 255
456
+ ],
457
+ [
458
+ 230,
459
+ 256
460
+ ],
461
+ [
462
+ 231,
463
+ 257
464
+ ],
465
+ [
466
+ 232,
467
+ 258
468
+ ],
469
+ [
470
+ 233,
471
+ 259
472
+ ],
473
+ [
474
+ 234,
475
+ 260
476
+ ],
477
+ [
478
+ 235,
479
+ 261
480
+ ],
481
+ [
482
+ 236,
483
+ 262
484
+ ],
485
+ [
486
+ 237,
487
+ 263
488
+ ],
489
+ [
490
+ 238,
491
+ 264
492
+ ],
493
+ [
494
+ 239,
495
+ 265
496
+ ],
497
+ [
498
+ 240,
499
+ 266
500
+ ],
501
+ [
502
+ 241,
503
+ 267
504
+ ],
505
+ [
506
+ 242,
507
+ 268
508
+ ],
509
+ [
510
+ 243,
511
+ 269
512
+ ],
513
+ [
514
+ 244,
515
+ 270
516
+ ],
517
+ [
518
+ 245,
519
+ 271
520
+ ],
521
+ [
522
+ 272,
523
+ 281
524
+ ],
525
+ [
526
+ 273,
527
+ 286
528
+ ],
529
+ [
530
+ 274,
531
+ 285
532
+ ],
533
+ [
534
+ 275,
535
+ 284
536
+ ],
537
+ [
538
+ 276,
539
+ 283
540
+ ],
541
+ [
542
+ 277,
543
+ 282
544
+ ],
545
+ [
546
+ 278,
547
+ 289
548
+ ],
549
+ [
550
+ 279,
551
+ 288
552
+ ],
553
+ [
554
+ 280,
555
+ 287
556
+ ],
557
+ [
558
+ 290,
559
+ 299
560
+ ],
561
+ [
562
+ 291,
563
+ 304
564
+ ],
565
+ [
566
+ 292,
567
+ 303
568
+ ],
569
+ [
570
+ 293,
571
+ 302
572
+ ],
573
+ [
574
+ 294,
575
+ 301
576
+ ],
577
+ [
578
+ 295,
579
+ 300
580
+ ],
581
+ [
582
+ 296,
583
+ 307
584
+ ],
585
+ [
586
+ 297,
587
+ 306
588
+ ],
589
+ [
590
+ 298,
591
+ 305
592
+ ]
593
+ ],
594
+ "head_config": {
595
+ "_name_or_path": "",
596
+ "architectures": null,
597
+ "chunk_size_feed_forward": 0,
598
+ "conv_kernel_size": 1,
599
+ "conv_kernel_sizes": [
600
+ 1,
601
+ 1,
602
+ 1
603
+ ],
604
+ "conv_out_channels": [
605
+ 512,
606
+ 512,
607
+ 256
608
+ ],
609
+ "dtype": null,
610
+ "id2label": {
611
+ "0": "LABEL_0",
612
+ "1": "LABEL_1"
613
+ },
614
+ "is_encoder_decoder": false,
615
+ "label2id": {
616
+ "LABEL_0": 0,
617
+ "LABEL_1": 1
618
+ },
619
+ "model_type": "sapiens2_head",
620
+ "output_attentions": false,
621
+ "output_hidden_states": false,
622
+ "problem_type": null,
623
+ "return_dict": true,
624
+ "scale_conv_kernel_size": 1,
625
+ "scale_conv_kernel_sizes": null,
626
+ "scale_conv_out_channels": null,
627
+ "scale_final_hidden_sizes": null,
628
+ "scale_final_input_size": null,
629
+ "upsample_kernel_size": 4,
630
+ "upsample_kernel_sizes": [
631
+ 4,
632
+ 4
633
+ ],
634
+ "upsample_out_channels": [
635
+ 1024,
636
+ 768
637
+ ],
638
+ "use_pixel_shuffle": null
639
+ },
640
+ "hidden_act": "silu",
641
+ "hidden_size": 2432,
642
+ "id2label": {
643
+ "0": "LABEL_0",
644
+ "1": "LABEL_1",
645
+ "2": "LABEL_2",
646
+ "3": "LABEL_3",
647
+ "4": "LABEL_4",
648
+ "5": "LABEL_5",
649
+ "6": "LABEL_6",
650
+ "7": "LABEL_7",
651
+ "8": "LABEL_8",
652
+ "9": "LABEL_9",
653
+ "10": "LABEL_10",
654
+ "11": "LABEL_11",
655
+ "12": "LABEL_12",
656
+ "13": "LABEL_13",
657
+ "14": "LABEL_14",
658
+ "15": "LABEL_15",
659
+ "16": "LABEL_16",
660
+ "17": "LABEL_17",
661
+ "18": "LABEL_18",
662
+ "19": "LABEL_19",
663
+ "20": "LABEL_20",
664
+ "21": "LABEL_21",
665
+ "22": "LABEL_22",
666
+ "23": "LABEL_23",
667
+ "24": "LABEL_24",
668
+ "25": "LABEL_25",
669
+ "26": "LABEL_26",
670
+ "27": "LABEL_27",
671
+ "28": "LABEL_28",
672
+ "29": "LABEL_29",
673
+ "30": "LABEL_30",
674
+ "31": "LABEL_31",
675
+ "32": "LABEL_32",
676
+ "33": "LABEL_33",
677
+ "34": "LABEL_34",
678
+ "35": "LABEL_35",
679
+ "36": "LABEL_36",
680
+ "37": "LABEL_37",
681
+ "38": "LABEL_38",
682
+ "39": "LABEL_39",
683
+ "40": "LABEL_40",
684
+ "41": "LABEL_41",
685
+ "42": "LABEL_42",
686
+ "43": "LABEL_43",
687
+ "44": "LABEL_44",
688
+ "45": "LABEL_45",
689
+ "46": "LABEL_46",
690
+ "47": "LABEL_47",
691
+ "48": "LABEL_48",
692
+ "49": "LABEL_49",
693
+ "50": "LABEL_50",
694
+ "51": "LABEL_51",
695
+ "52": "LABEL_52",
696
+ "53": "LABEL_53",
697
+ "54": "LABEL_54",
698
+ "55": "LABEL_55",
699
+ "56": "LABEL_56",
700
+ "57": "LABEL_57",
701
+ "58": "LABEL_58",
702
+ "59": "LABEL_59",
703
+ "60": "LABEL_60",
704
+ "61": "LABEL_61",
705
+ "62": "LABEL_62",
706
+ "63": "LABEL_63",
707
+ "64": "LABEL_64",
708
+ "65": "LABEL_65",
709
+ "66": "LABEL_66",
710
+ "67": "LABEL_67",
711
+ "68": "LABEL_68",
712
+ "69": "LABEL_69",
713
+ "70": "LABEL_70",
714
+ "71": "LABEL_71",
715
+ "72": "LABEL_72",
716
+ "73": "LABEL_73",
717
+ "74": "LABEL_74",
718
+ "75": "LABEL_75",
719
+ "76": "LABEL_76",
720
+ "77": "LABEL_77",
721
+ "78": "LABEL_78",
722
+ "79": "LABEL_79",
723
+ "80": "LABEL_80",
724
+ "81": "LABEL_81",
725
+ "82": "LABEL_82",
726
+ "83": "LABEL_83",
727
+ "84": "LABEL_84",
728
+ "85": "LABEL_85",
729
+ "86": "LABEL_86",
730
+ "87": "LABEL_87",
731
+ "88": "LABEL_88",
732
+ "89": "LABEL_89",
733
+ "90": "LABEL_90",
734
+ "91": "LABEL_91",
735
+ "92": "LABEL_92",
736
+ "93": "LABEL_93",
737
+ "94": "LABEL_94",
738
+ "95": "LABEL_95",
739
+ "96": "LABEL_96",
740
+ "97": "LABEL_97",
741
+ "98": "LABEL_98",
742
+ "99": "LABEL_99",
743
+ "100": "LABEL_100",
744
+ "101": "LABEL_101",
745
+ "102": "LABEL_102",
746
+ "103": "LABEL_103",
747
+ "104": "LABEL_104",
748
+ "105": "LABEL_105",
749
+ "106": "LABEL_106",
750
+ "107": "LABEL_107",
751
+ "108": "LABEL_108",
752
+ "109": "LABEL_109",
753
+ "110": "LABEL_110",
754
+ "111": "LABEL_111",
755
+ "112": "LABEL_112",
756
+ "113": "LABEL_113",
757
+ "114": "LABEL_114",
758
+ "115": "LABEL_115",
759
+ "116": "LABEL_116",
760
+ "117": "LABEL_117",
761
+ "118": "LABEL_118",
762
+ "119": "LABEL_119",
763
+ "120": "LABEL_120",
764
+ "121": "LABEL_121",
765
+ "122": "LABEL_122",
766
+ "123": "LABEL_123",
767
+ "124": "LABEL_124",
768
+ "125": "LABEL_125",
769
+ "126": "LABEL_126",
770
+ "127": "LABEL_127",
771
+ "128": "LABEL_128",
772
+ "129": "LABEL_129",
773
+ "130": "LABEL_130",
774
+ "131": "LABEL_131",
775
+ "132": "LABEL_132",
776
+ "133": "LABEL_133",
777
+ "134": "LABEL_134",
778
+ "135": "LABEL_135",
779
+ "136": "LABEL_136",
780
+ "137": "LABEL_137",
781
+ "138": "LABEL_138",
782
+ "139": "LABEL_139",
783
+ "140": "LABEL_140",
784
+ "141": "LABEL_141",
785
+ "142": "LABEL_142",
786
+ "143": "LABEL_143",
787
+ "144": "LABEL_144",
788
+ "145": "LABEL_145",
789
+ "146": "LABEL_146",
790
+ "147": "LABEL_147",
791
+ "148": "LABEL_148",
792
+ "149": "LABEL_149",
793
+ "150": "LABEL_150",
794
+ "151": "LABEL_151",
795
+ "152": "LABEL_152",
796
+ "153": "LABEL_153",
797
+ "154": "LABEL_154",
798
+ "155": "LABEL_155",
799
+ "156": "LABEL_156",
800
+ "157": "LABEL_157",
801
+ "158": "LABEL_158",
802
+ "159": "LABEL_159",
803
+ "160": "LABEL_160",
804
+ "161": "LABEL_161",
805
+ "162": "LABEL_162",
806
+ "163": "LABEL_163",
807
+ "164": "LABEL_164",
808
+ "165": "LABEL_165",
809
+ "166": "LABEL_166",
810
+ "167": "LABEL_167",
811
+ "168": "LABEL_168",
812
+ "169": "LABEL_169",
813
+ "170": "LABEL_170",
814
+ "171": "LABEL_171",
815
+ "172": "LABEL_172",
816
+ "173": "LABEL_173",
817
+ "174": "LABEL_174",
818
+ "175": "LABEL_175",
819
+ "176": "LABEL_176",
820
+ "177": "LABEL_177",
821
+ "178": "LABEL_178",
822
+ "179": "LABEL_179",
823
+ "180": "LABEL_180",
824
+ "181": "LABEL_181",
825
+ "182": "LABEL_182",
826
+ "183": "LABEL_183",
827
+ "184": "LABEL_184",
828
+ "185": "LABEL_185",
829
+ "186": "LABEL_186",
830
+ "187": "LABEL_187",
831
+ "188": "LABEL_188",
832
+ "189": "LABEL_189",
833
+ "190": "LABEL_190",
834
+ "191": "LABEL_191",
835
+ "192": "LABEL_192",
836
+ "193": "LABEL_193",
837
+ "194": "LABEL_194",
838
+ "195": "LABEL_195",
839
+ "196": "LABEL_196",
840
+ "197": "LABEL_197",
841
+ "198": "LABEL_198",
842
+ "199": "LABEL_199",
843
+ "200": "LABEL_200",
844
+ "201": "LABEL_201",
845
+ "202": "LABEL_202",
846
+ "203": "LABEL_203",
847
+ "204": "LABEL_204",
848
+ "205": "LABEL_205",
849
+ "206": "LABEL_206",
850
+ "207": "LABEL_207",
851
+ "208": "LABEL_208",
852
+ "209": "LABEL_209",
853
+ "210": "LABEL_210",
854
+ "211": "LABEL_211",
855
+ "212": "LABEL_212",
856
+ "213": "LABEL_213",
857
+ "214": "LABEL_214",
858
+ "215": "LABEL_215",
859
+ "216": "LABEL_216",
860
+ "217": "LABEL_217",
861
+ "218": "LABEL_218",
862
+ "219": "LABEL_219",
863
+ "220": "LABEL_220",
864
+ "221": "LABEL_221",
865
+ "222": "LABEL_222",
866
+ "223": "LABEL_223",
867
+ "224": "LABEL_224",
868
+ "225": "LABEL_225",
869
+ "226": "LABEL_226",
870
+ "227": "LABEL_227",
871
+ "228": "LABEL_228",
872
+ "229": "LABEL_229",
873
+ "230": "LABEL_230",
874
+ "231": "LABEL_231",
875
+ "232": "LABEL_232",
876
+ "233": "LABEL_233",
877
+ "234": "LABEL_234",
878
+ "235": "LABEL_235",
879
+ "236": "LABEL_236",
880
+ "237": "LABEL_237",
881
+ "238": "LABEL_238",
882
+ "239": "LABEL_239",
883
+ "240": "LABEL_240",
884
+ "241": "LABEL_241",
885
+ "242": "LABEL_242",
886
+ "243": "LABEL_243",
887
+ "244": "LABEL_244",
888
+ "245": "LABEL_245",
889
+ "246": "LABEL_246",
890
+ "247": "LABEL_247",
891
+ "248": "LABEL_248",
892
+ "249": "LABEL_249",
893
+ "250": "LABEL_250",
894
+ "251": "LABEL_251",
895
+ "252": "LABEL_252",
896
+ "253": "LABEL_253",
897
+ "254": "LABEL_254",
898
+ "255": "LABEL_255",
899
+ "256": "LABEL_256",
900
+ "257": "LABEL_257",
901
+ "258": "LABEL_258",
902
+ "259": "LABEL_259",
903
+ "260": "LABEL_260",
904
+ "261": "LABEL_261",
905
+ "262": "LABEL_262",
906
+ "263": "LABEL_263",
907
+ "264": "LABEL_264",
908
+ "265": "LABEL_265",
909
+ "266": "LABEL_266",
910
+ "267": "LABEL_267",
911
+ "268": "LABEL_268",
912
+ "269": "LABEL_269",
913
+ "270": "LABEL_270",
914
+ "271": "LABEL_271",
915
+ "272": "LABEL_272",
916
+ "273": "LABEL_273",
917
+ "274": "LABEL_274",
918
+ "275": "LABEL_275",
919
+ "276": "LABEL_276",
920
+ "277": "LABEL_277",
921
+ "278": "LABEL_278",
922
+ "279": "LABEL_279",
923
+ "280": "LABEL_280",
924
+ "281": "LABEL_281",
925
+ "282": "LABEL_282",
926
+ "283": "LABEL_283",
927
+ "284": "LABEL_284",
928
+ "285": "LABEL_285",
929
+ "286": "LABEL_286",
930
+ "287": "LABEL_287",
931
+ "288": "LABEL_288",
932
+ "289": "LABEL_289",
933
+ "290": "LABEL_290",
934
+ "291": "LABEL_291",
935
+ "292": "LABEL_292",
936
+ "293": "LABEL_293",
937
+ "294": "LABEL_294",
938
+ "295": "LABEL_295",
939
+ "296": "LABEL_296",
940
+ "297": "LABEL_297",
941
+ "298": "LABEL_298",
942
+ "299": "LABEL_299",
943
+ "300": "LABEL_300",
944
+ "301": "LABEL_301",
945
+ "302": "LABEL_302",
946
+ "303": "LABEL_303",
947
+ "304": "LABEL_304",
948
+ "305": "LABEL_305",
949
+ "306": "LABEL_306",
950
+ "307": "LABEL_307"
951
+ },
952
+ "image_size": [
953
+ 1024,
954
+ 768
955
+ ],
956
+ "initializer_range": 0.02,
957
+ "intermediate_size": 9728,
958
+ "key_bias": true,
959
+ "label2id": {
960
+ "LABEL_0": 0,
961
+ "LABEL_1": 1,
962
+ "LABEL_10": 10,
963
+ "LABEL_100": 100,
964
+ "LABEL_101": 101,
965
+ "LABEL_102": 102,
966
+ "LABEL_103": 103,
967
+ "LABEL_104": 104,
968
+ "LABEL_105": 105,
969
+ "LABEL_106": 106,
970
+ "LABEL_107": 107,
971
+ "LABEL_108": 108,
972
+ "LABEL_109": 109,
973
+ "LABEL_11": 11,
974
+ "LABEL_110": 110,
975
+ "LABEL_111": 111,
976
+ "LABEL_112": 112,
977
+ "LABEL_113": 113,
978
+ "LABEL_114": 114,
979
+ "LABEL_115": 115,
980
+ "LABEL_116": 116,
981
+ "LABEL_117": 117,
982
+ "LABEL_118": 118,
983
+ "LABEL_119": 119,
984
+ "LABEL_12": 12,
985
+ "LABEL_120": 120,
986
+ "LABEL_121": 121,
987
+ "LABEL_122": 122,
988
+ "LABEL_123": 123,
989
+ "LABEL_124": 124,
990
+ "LABEL_125": 125,
991
+ "LABEL_126": 126,
992
+ "LABEL_127": 127,
993
+ "LABEL_128": 128,
994
+ "LABEL_129": 129,
995
+ "LABEL_13": 13,
996
+ "LABEL_130": 130,
997
+ "LABEL_131": 131,
998
+ "LABEL_132": 132,
999
+ "LABEL_133": 133,
1000
+ "LABEL_134": 134,
1001
+ "LABEL_135": 135,
1002
+ "LABEL_136": 136,
1003
+ "LABEL_137": 137,
1004
+ "LABEL_138": 138,
1005
+ "LABEL_139": 139,
1006
+ "LABEL_14": 14,
1007
+ "LABEL_140": 140,
1008
+ "LABEL_141": 141,
1009
+ "LABEL_142": 142,
1010
+ "LABEL_143": 143,
1011
+ "LABEL_144": 144,
1012
+ "LABEL_145": 145,
1013
+ "LABEL_146": 146,
1014
+ "LABEL_147": 147,
1015
+ "LABEL_148": 148,
1016
+ "LABEL_149": 149,
1017
+ "LABEL_15": 15,
1018
+ "LABEL_150": 150,
1019
+ "LABEL_151": 151,
1020
+ "LABEL_152": 152,
1021
+ "LABEL_153": 153,
1022
+ "LABEL_154": 154,
1023
+ "LABEL_155": 155,
1024
+ "LABEL_156": 156,
1025
+ "LABEL_157": 157,
1026
+ "LABEL_158": 158,
1027
+ "LABEL_159": 159,
1028
+ "LABEL_16": 16,
1029
+ "LABEL_160": 160,
1030
+ "LABEL_161": 161,
1031
+ "LABEL_162": 162,
1032
+ "LABEL_163": 163,
1033
+ "LABEL_164": 164,
1034
+ "LABEL_165": 165,
1035
+ "LABEL_166": 166,
1036
+ "LABEL_167": 167,
1037
+ "LABEL_168": 168,
1038
+ "LABEL_169": 169,
1039
+ "LABEL_17": 17,
1040
+ "LABEL_170": 170,
1041
+ "LABEL_171": 171,
1042
+ "LABEL_172": 172,
1043
+ "LABEL_173": 173,
1044
+ "LABEL_174": 174,
1045
+ "LABEL_175": 175,
1046
+ "LABEL_176": 176,
1047
+ "LABEL_177": 177,
1048
+ "LABEL_178": 178,
1049
+ "LABEL_179": 179,
1050
+ "LABEL_18": 18,
1051
+ "LABEL_180": 180,
1052
+ "LABEL_181": 181,
1053
+ "LABEL_182": 182,
1054
+ "LABEL_183": 183,
1055
+ "LABEL_184": 184,
1056
+ "LABEL_185": 185,
1057
+ "LABEL_186": 186,
1058
+ "LABEL_187": 187,
1059
+ "LABEL_188": 188,
1060
+ "LABEL_189": 189,
1061
+ "LABEL_19": 19,
1062
+ "LABEL_190": 190,
1063
+ "LABEL_191": 191,
1064
+ "LABEL_192": 192,
1065
+ "LABEL_193": 193,
1066
+ "LABEL_194": 194,
1067
+ "LABEL_195": 195,
1068
+ "LABEL_196": 196,
1069
+ "LABEL_197": 197,
1070
+ "LABEL_198": 198,
1071
+ "LABEL_199": 199,
1072
+ "LABEL_2": 2,
1073
+ "LABEL_20": 20,
1074
+ "LABEL_200": 200,
1075
+ "LABEL_201": 201,
1076
+ "LABEL_202": 202,
1077
+ "LABEL_203": 203,
1078
+ "LABEL_204": 204,
1079
+ "LABEL_205": 205,
1080
+ "LABEL_206": 206,
1081
+ "LABEL_207": 207,
1082
+ "LABEL_208": 208,
1083
+ "LABEL_209": 209,
1084
+ "LABEL_21": 21,
1085
+ "LABEL_210": 210,
1086
+ "LABEL_211": 211,
1087
+ "LABEL_212": 212,
1088
+ "LABEL_213": 213,
1089
+ "LABEL_214": 214,
1090
+ "LABEL_215": 215,
1091
+ "LABEL_216": 216,
1092
+ "LABEL_217": 217,
1093
+ "LABEL_218": 218,
1094
+ "LABEL_219": 219,
1095
+ "LABEL_22": 22,
1096
+ "LABEL_220": 220,
1097
+ "LABEL_221": 221,
1098
+ "LABEL_222": 222,
1099
+ "LABEL_223": 223,
1100
+ "LABEL_224": 224,
1101
+ "LABEL_225": 225,
1102
+ "LABEL_226": 226,
1103
+ "LABEL_227": 227,
1104
+ "LABEL_228": 228,
1105
+ "LABEL_229": 229,
1106
+ "LABEL_23": 23,
1107
+ "LABEL_230": 230,
1108
+ "LABEL_231": 231,
1109
+ "LABEL_232": 232,
1110
+ "LABEL_233": 233,
1111
+ "LABEL_234": 234,
1112
+ "LABEL_235": 235,
1113
+ "LABEL_236": 236,
1114
+ "LABEL_237": 237,
1115
+ "LABEL_238": 238,
1116
+ "LABEL_239": 239,
1117
+ "LABEL_24": 24,
1118
+ "LABEL_240": 240,
1119
+ "LABEL_241": 241,
1120
+ "LABEL_242": 242,
1121
+ "LABEL_243": 243,
1122
+ "LABEL_244": 244,
1123
+ "LABEL_245": 245,
1124
+ "LABEL_246": 246,
1125
+ "LABEL_247": 247,
1126
+ "LABEL_248": 248,
1127
+ "LABEL_249": 249,
1128
+ "LABEL_25": 25,
1129
+ "LABEL_250": 250,
1130
+ "LABEL_251": 251,
1131
+ "LABEL_252": 252,
1132
+ "LABEL_253": 253,
1133
+ "LABEL_254": 254,
1134
+ "LABEL_255": 255,
1135
+ "LABEL_256": 256,
1136
+ "LABEL_257": 257,
1137
+ "LABEL_258": 258,
1138
+ "LABEL_259": 259,
1139
+ "LABEL_26": 26,
1140
+ "LABEL_260": 260,
1141
+ "LABEL_261": 261,
1142
+ "LABEL_262": 262,
1143
+ "LABEL_263": 263,
1144
+ "LABEL_264": 264,
1145
+ "LABEL_265": 265,
1146
+ "LABEL_266": 266,
1147
+ "LABEL_267": 267,
1148
+ "LABEL_268": 268,
1149
+ "LABEL_269": 269,
1150
+ "LABEL_27": 27,
1151
+ "LABEL_270": 270,
1152
+ "LABEL_271": 271,
1153
+ "LABEL_272": 272,
1154
+ "LABEL_273": 273,
1155
+ "LABEL_274": 274,
1156
+ "LABEL_275": 275,
1157
+ "LABEL_276": 276,
1158
+ "LABEL_277": 277,
1159
+ "LABEL_278": 278,
1160
+ "LABEL_279": 279,
1161
+ "LABEL_28": 28,
1162
+ "LABEL_280": 280,
1163
+ "LABEL_281": 281,
1164
+ "LABEL_282": 282,
1165
+ "LABEL_283": 283,
1166
+ "LABEL_284": 284,
1167
+ "LABEL_285": 285,
1168
+ "LABEL_286": 286,
1169
+ "LABEL_287": 287,
1170
+ "LABEL_288": 288,
1171
+ "LABEL_289": 289,
1172
+ "LABEL_29": 29,
1173
+ "LABEL_290": 290,
1174
+ "LABEL_291": 291,
1175
+ "LABEL_292": 292,
1176
+ "LABEL_293": 293,
1177
+ "LABEL_294": 294,
1178
+ "LABEL_295": 295,
1179
+ "LABEL_296": 296,
1180
+ "LABEL_297": 297,
1181
+ "LABEL_298": 298,
1182
+ "LABEL_299": 299,
1183
+ "LABEL_3": 3,
1184
+ "LABEL_30": 30,
1185
+ "LABEL_300": 300,
1186
+ "LABEL_301": 301,
1187
+ "LABEL_302": 302,
1188
+ "LABEL_303": 303,
1189
+ "LABEL_304": 304,
1190
+ "LABEL_305": 305,
1191
+ "LABEL_306": 306,
1192
+ "LABEL_307": 307,
1193
+ "LABEL_31": 31,
1194
+ "LABEL_32": 32,
1195
+ "LABEL_33": 33,
1196
+ "LABEL_34": 34,
1197
+ "LABEL_35": 35,
1198
+ "LABEL_36": 36,
1199
+ "LABEL_37": 37,
1200
+ "LABEL_38": 38,
1201
+ "LABEL_39": 39,
1202
+ "LABEL_4": 4,
1203
+ "LABEL_40": 40,
1204
+ "LABEL_41": 41,
1205
+ "LABEL_42": 42,
1206
+ "LABEL_43": 43,
1207
+ "LABEL_44": 44,
1208
+ "LABEL_45": 45,
1209
+ "LABEL_46": 46,
1210
+ "LABEL_47": 47,
1211
+ "LABEL_48": 48,
1212
+ "LABEL_49": 49,
1213
+ "LABEL_5": 5,
1214
+ "LABEL_50": 50,
1215
+ "LABEL_51": 51,
1216
+ "LABEL_52": 52,
1217
+ "LABEL_53": 53,
1218
+ "LABEL_54": 54,
1219
+ "LABEL_55": 55,
1220
+ "LABEL_56": 56,
1221
+ "LABEL_57": 57,
1222
+ "LABEL_58": 58,
1223
+ "LABEL_59": 59,
1224
+ "LABEL_6": 6,
1225
+ "LABEL_60": 60,
1226
+ "LABEL_61": 61,
1227
+ "LABEL_62": 62,
1228
+ "LABEL_63": 63,
1229
+ "LABEL_64": 64,
1230
+ "LABEL_65": 65,
1231
+ "LABEL_66": 66,
1232
+ "LABEL_67": 67,
1233
+ "LABEL_68": 68,
1234
+ "LABEL_69": 69,
1235
+ "LABEL_7": 7,
1236
+ "LABEL_70": 70,
1237
+ "LABEL_71": 71,
1238
+ "LABEL_72": 72,
1239
+ "LABEL_73": 73,
1240
+ "LABEL_74": 74,
1241
+ "LABEL_75": 75,
1242
+ "LABEL_76": 76,
1243
+ "LABEL_77": 77,
1244
+ "LABEL_78": 78,
1245
+ "LABEL_79": 79,
1246
+ "LABEL_8": 8,
1247
+ "LABEL_80": 80,
1248
+ "LABEL_81": 81,
1249
+ "LABEL_82": 82,
1250
+ "LABEL_83": 83,
1251
+ "LABEL_84": 84,
1252
+ "LABEL_85": 85,
1253
+ "LABEL_86": 86,
1254
+ "LABEL_87": 87,
1255
+ "LABEL_88": 88,
1256
+ "LABEL_89": 89,
1257
+ "LABEL_9": 9,
1258
+ "LABEL_90": 90,
1259
+ "LABEL_91": 91,
1260
+ "LABEL_92": 92,
1261
+ "LABEL_93": 93,
1262
+ "LABEL_94": 94,
1263
+ "LABEL_95": 95,
1264
+ "LABEL_96": 96,
1265
+ "LABEL_97": 97,
1266
+ "LABEL_98": 98,
1267
+ "LABEL_99": 99
1268
+ },
1269
+ "layer_norm_eps": 1e-05,
1270
+ "layerscale_value": 1.0,
1271
+ "mlp_bias": true,
1272
+ "model_type": "sapiens2",
1273
+ "normalize_backbone_outputs": true,
1274
+ "num_attention_heads": 32,
1275
+ "num_channels": 3,
1276
+ "num_first_full_attention_layers": 8,
1277
+ "num_hidden_layers": 56,
1278
+ "num_key_value_attention_heads": 16,
1279
+ "num_key_value_heads_per_layer": [
1280
+ 32,
1281
+ 32,
1282
+ 32,
1283
+ 32,
1284
+ 32,
1285
+ 32,
1286
+ 32,
1287
+ 32,
1288
+ 16,
1289
+ 16,
1290
+ 16,
1291
+ 16,
1292
+ 16,
1293
+ 16,
1294
+ 16,
1295
+ 16,
1296
+ 16,
1297
+ 16,
1298
+ 16,
1299
+ 16,
1300
+ 16,
1301
+ 16,
1302
+ 16,
1303
+ 16,
1304
+ 16,
1305
+ 16,
1306
+ 16,
1307
+ 16,
1308
+ 16,
1309
+ 16,
1310
+ 16,
1311
+ 16,
1312
+ 16,
1313
+ 16,
1314
+ 16,
1315
+ 16,
1316
+ 16,
1317
+ 16,
1318
+ 16,
1319
+ 16,
1320
+ 16,
1321
+ 16,
1322
+ 16,
1323
+ 16,
1324
+ 16,
1325
+ 16,
1326
+ 16,
1327
+ 16,
1328
+ 32,
1329
+ 32,
1330
+ 32,
1331
+ 32,
1332
+ 32,
1333
+ 32,
1334
+ 32,
1335
+ 32
1336
+ ],
1337
+ "num_last_full_attention_layers": 8,
1338
+ "num_register_tokens": 8,
1339
+ "out_features": [
1340
+ "stage56"
1341
+ ],
1342
+ "out_indices": [
1343
+ 56
1344
+ ],
1345
+ "patch_size": 16,
1346
+ "pos_embed_jitter": null,
1347
+ "pos_embed_rescale": 2.0,
1348
+ "pos_embed_shift": null,
1349
+ "proj_bias": true,
1350
+ "query_bias": true,
1351
+ "reshape_hidden_states": true,
1352
+ "rms_norm_eps": 1e-06,
1353
+ "rope_theta": 100.0,
1354
+ "semantic_loss_ignore_index": 255,
1355
+ "stage_names": [
1356
+ "stem",
1357
+ "stage1",
1358
+ "stage2",
1359
+ "stage3",
1360
+ "stage4",
1361
+ "stage5",
1362
+ "stage6",
1363
+ "stage7",
1364
+ "stage8",
1365
+ "stage9",
1366
+ "stage10",
1367
+ "stage11",
1368
+ "stage12",
1369
+ "stage13",
1370
+ "stage14",
1371
+ "stage15",
1372
+ "stage16",
1373
+ "stage17",
1374
+ "stage18",
1375
+ "stage19",
1376
+ "stage20",
1377
+ "stage21",
1378
+ "stage22",
1379
+ "stage23",
1380
+ "stage24",
1381
+ "stage25",
1382
+ "stage26",
1383
+ "stage27",
1384
+ "stage28",
1385
+ "stage29",
1386
+ "stage30",
1387
+ "stage31",
1388
+ "stage32",
1389
+ "stage33",
1390
+ "stage34",
1391
+ "stage35",
1392
+ "stage36",
1393
+ "stage37",
1394
+ "stage38",
1395
+ "stage39",
1396
+ "stage40",
1397
+ "stage41",
1398
+ "stage42",
1399
+ "stage43",
1400
+ "stage44",
1401
+ "stage45",
1402
+ "stage46",
1403
+ "stage47",
1404
+ "stage48",
1405
+ "stage49",
1406
+ "stage50",
1407
+ "stage51",
1408
+ "stage52",
1409
+ "stage53",
1410
+ "stage54",
1411
+ "stage55",
1412
+ "stage56"
1413
+ ],
1414
+ "transformers_version": "5.10.0.dev0",
1415
+ "use_gated_mlp": true,
1416
+ "use_mask_token": false,
1417
+ "use_qk_norm": true,
1418
+ "value_bias": true
1419
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4848da8691c72e14d3ff71319f077363107129bf4128019eb39d072129b2a52
3
+ size 20480899148
preprocessor_config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "crop_size": {
3
+ "height": 224,
4
+ "width": 224
5
+ },
6
+ "do_center_crop": false,
7
+ "do_normalize": true,
8
+ "do_pad": false,
9
+ "do_reduce_labels": false,
10
+ "do_rescale": true,
11
+ "do_resize": true,
12
+ "image_mean": [
13
+ 0.485,
14
+ 0.456,
15
+ 0.406
16
+ ],
17
+ "image_processor_type": "Sapiens2ImageProcessor",
18
+ "image_std": [
19
+ 0.229,
20
+ 0.224,
21
+ 0.225
22
+ ],
23
+ "resample": 2,
24
+ "rescale_factor": 0.00392156862745098,
25
+ "size": {
26
+ "height": 1024,
27
+ "width": 768
28
+ }
29
+ }