HaithemH commited on
Commit
c4ebe87
·
verified ·
1 Parent(s): f4a27ab

Upload VJEPA2ForVideoClassification

Browse files
Files changed (2) hide show
  1. config.json +432 -434
  2. model.safetensors +2 -2
config.json CHANGED
@@ -11,453 +11,451 @@
11
  "hidden_dropout_prob": 0.0,
12
  "hidden_size": 1024,
13
  "id2label": {
14
- "0": "ApplyLipstick",
15
  "1": "ApplyEyeMakeup",
16
- "2": "Biking",
17
- "3": "BrushingTeeth",
18
- "4": "CuttingInKitchen",
19
- "5": "Haircut",
20
- "6": "HandstandWalking",
21
- "7": "HandstandPushups",
22
  "8": "Hammering",
23
- "9": "Knitting",
24
- "10": "Typing",
25
- "11": "BlowDryHair",
26
- "12": "PlayingDaf",
27
- "13": "HeadMassage",
28
- "14": "WalkingWithDog",
29
- "15": "Sitting",
30
- "16": "Standing",
31
- "17": "Walking",
32
- "18": "DrinkAngry",
33
- "19": "DrinkDisgust",
34
- "20": "DrinkHappy",
35
- "21": "DrinkNeutral",
36
- "22": "DrinkSad",
37
- "23": "PutonglassesAngry",
38
- "24": "PutonglassesHappy",
39
- "25": "PutonglassesNeutral",
40
- "26": "PutonglassesSad",
41
- "27": "PutonjacketAngry",
42
- "28": "PutonjacketHappy",
43
- "29": "PutonjacketNeutral",
44
- "30": "PutonjacketSad",
45
- "31": "ReadAngry",
46
- "32": "ReadHappy",
47
- "33": "ReadNeutral",
48
- "34": "ReadSad",
49
- "35": "SitdownAngry",
50
- "36": "SitdownHappy",
51
- "37": "SitdownNeutral",
52
- "38": "SitdownSad",
53
- "39": "StandupAngry",
54
- "40": "StandupHappy",
55
- "41": "StandupNeutral",
56
- "42": "StandupSad",
57
- "43": "TakeoffglassesAngry",
58
- "44": "TakeoffglassesHappy",
59
- "45": "TakeoffglassesNeutral",
60
- "46": "TakeoffglassesSad",
61
- "47": "TakeoffjacketAngry",
62
- "48": "TakeoffjacketHappy",
63
- "49": "TakeoffjacketNeutral",
64
- "50": "TakeoffjacketSad",
65
- "51": "WriteAngry",
66
- "52": "WriteHappy",
67
- "53": "WriteNeutral",
68
- "54": "ViolenceFighting",
69
- "55": "Fighting",
70
- "56": "Waving",
71
- "57": "Clapping",
72
- "58": "WalkingWhileUsingPhone",
73
- "59": "Jump",
74
- "60": "SitDown",
75
- "61": "LyingDown",
76
- "62": "StandUp",
77
- "63": "FallDown",
78
- "64": "ClosingABook",
79
- "65": "ClosingABox",
80
- "66": "ClosingAClosetCabinet",
81
- "67": "ClosingADoor",
82
- "68": "ClosingALaptop",
83
- "69": "ClosingARefrigerator",
84
- "70": "ClosingAWindow",
85
- "71": "DrinkingFromACupGlassBottle",
86
- "72": "EatingASandwich",
87
- "73": "FixingADoor",
88
- "74": "FixingADoorknob",
89
- "75": "FixingALight",
90
- "76": "FixingAVacuum",
91
- "77": "FixingTheirHair",
92
- "78": "GraspingOntoADoorknob",
93
- "79": "HoldingABag",
94
- "80": "HoldingABlanket",
95
- "81": "HoldingABook",
96
- "82": "HoldingABox",
97
- "83": "HoldingABroom",
98
- "84": "HoldingACupGlassBottleOfSomething",
99
- "85": "HoldingADish",
100
- "86": "HoldingALaptop",
101
- "87": "HoldingAMirror",
102
- "88": "HoldingAPhoneCamera",
103
- "89": "HoldingAPicture",
104
- "90": "HoldingAPillow",
105
- "91": "HoldingASandwich",
106
- "92": "HoldingAShoeShoes",
107
- "93": "HoldingATowelS",
108
- "94": "HoldingAVacuum",
109
- "95": "HoldingSomeClothes",
110
- "96": "HoldingSomeFood",
111
- "97": "HoldingSomeMedicine",
112
- "98": "LaughingAtAPicture",
113
- "99": "LaughingAtTelevision",
114
- "100": "LyingOnABed",
115
- "101": "LyingOnASofaCouch",
116
- "102": "LyingOnTheFloor",
117
- "103": "MakingASandwich",
118
- "104": "OpeningABag",
119
- "105": "OpeningABook",
120
- "106": "OpeningABox",
121
- "107": "OpeningAClosetCabinet",
122
- "108": "OpeningADoor",
123
- "109": "OpeningALaptop",
124
- "110": "OpeningARefrigerator",
125
- "111": "OpeningAWindow",
126
- "112": "PlayingWithAPhoneCamera",
127
- "113": "PouringSomethingIntoACupGlassBottle",
128
- "114": "PuttingABagSomewhere",
129
- "115": "PuttingABlanketSomewhere",
130
- "116": "PuttingABookSomewhere",
131
- "117": "PuttingABoxSomewhere",
132
- "118": "PuttingABroomSomewhere",
133
- "119": "PuttingACupGlassBottleSomewhere",
134
- "120": "PuttingADishEsSomewhere",
135
- "121": "PuttingALaptopSomewhere",
136
- "122": "PuttingAPhoneCameraSomewhere",
137
- "123": "PuttingAPictureSomewhere",
138
- "124": "PuttingAPillowSomewhere",
139
- "125": "PuttingASandwichSomewhere",
140
- "126": "PuttingATowelSSomewhere",
141
- "127": "PuttingClothesSomewhere",
142
- "128": "PuttingGroceriesSomewhere",
143
- "129": "PuttingOnShoeShoes",
144
- "130": "PuttingShoesSomewhere",
145
- "131": "PuttingSomeFoodSomewhere",
146
- "132": "PuttingSomethingOnAShelf",
147
- "133": "PuttingSomethingOnATable",
148
- "134": "PuttingTheirPaperNotebookSomewhere",
149
- "135": "ReachingForAndGrabbingAPicture",
150
- "136": "SittingAtATable",
151
- "137": "SittingInABed",
152
- "138": "SittingInAChair",
153
- "139": "SittingOnATable",
154
- "140": "SittingOnSofaCouch",
155
- "141": "SittingOnTheFloor",
156
- "142": "SmilingAtABook",
157
- "143": "SmilingInAMirror",
158
- "144": "SnugglingWithABlanket",
159
- "145": "SnugglingWithAPillow",
160
- "146": "SomeoneIsAwakeningInBed",
161
- "147": "SomeoneIsAwakeningSomewhere",
162
- "148": "SomeoneIsCookingSomething",
163
- "149": "SomeoneIsDressing",
164
- "150": "SomeoneIsEatingSomething",
165
- "151": "SomeoneIsGoingFromStandingToSitting",
166
- "152": "SomeoneIsHoldingAPaperNotebook",
167
- "153": "SomeoneIsLaughing",
168
- "154": "SomeoneIsRunningSomewhere",
169
- "155": "SomeoneIsSmiling",
170
- "156": "SomeoneIsSneezing",
171
- "157": "SomeoneIsStandingUpFromSomewhere",
172
  "158": "SomeoneIsUndressing",
173
- "159": "StandingOnAChair",
174
- "160": "TakingABagFromSomewhere",
175
- "161": "TakingABlanketFromSomewhere",
176
- "162": "TakingABookFromSomewhere",
177
- "163": "TakingABoxFromSomewhere",
178
- "164": "TakingABroomFromSomewhere",
179
- "165": "TakingACupGlassBottleFromSomewhere",
180
- "166": "TakingADishEsFromSomewhere",
181
- "167": "TakingALaptopFromSomewhere",
182
  "168": "TakingAPhoneCameraFromSomewhere",
183
- "169": "TakingAPictureOfSomething",
184
- "170": "TakingAPillowFromSomewhere",
185
- "171": "TakingASandwichFromSomewhere",
186
- "172": "TakingATowelSFromSomewhere",
187
- "173": "TakingAVacuumFromSomewhere",
188
- "174": "TakingConsumingSomeMedicine",
189
- "175": "TakingFoodFromSomewhere",
190
- "176": "TakingOffSomeShoes",
191
- "177": "TakingPaperNotebookFromSomewhere",
192
- "178": "TakingShoesFromSomewhere",
193
- "179": "TakingSomeClothesFromSomewhere",
194
- "180": "TakingSomethingFromABox",
195
- "181": "TalkingOnAPhoneCamera",
196
- "182": "ThrowingABagSomewhere",
197
- "183": "ThrowingABlanketSomewhere",
198
- "184": "ThrowingABookSomewhere",
199
- "185": "ThrowingABoxSomewhere",
200
- "186": "ThrowingABroomSomewhere",
201
- "187": "ThrowingAPillowSomewhere",
202
- "188": "ThrowingATowelSSomewhere",
203
- "189": "ThrowingClothesSomewhere",
204
- "190": "ThrowingFoodSomewhere",
205
- "191": "ThrowingShoesSomewhere",
206
- "192": "ThrowingSomethingOnTheFloor",
207
- "193": "TidyingAShelfOrSomethingOnAShelf",
208
- "194": "TidyingSomeClothes",
209
- "195": "TidyingSomethingOnTheFloor",
210
- "196": "TidyingUpABlanketS",
211
- "197": "TidyingUpAClosetCabinet",
212
- "198": "TidyingUpATable",
213
- "199": "TidyingUpATowelS",
214
- "200": "TidyingUpWithABroom",
215
- "201": "TurningOffALight",
216
- "202": "TurningOnALight",
217
- "203": "WalkingThroughADoorway",
218
- "204": "WashADishDishes",
219
- "205": "WashingACupGlassBottle",
220
- "206": "WashingAMirror",
221
- "207": "WashingATable",
222
- "208": "WashingAWindow",
223
- "209": "WashingSomeClothes",
224
- "210": "WashingSomethingWithATowel",
225
- "211": "WashingTheirHands",
226
- "212": "WatchingALaptopOrSomethingOnALaptop",
227
- "213": "WatchingLookingAtAPicture",
228
- "214": "WatchingLookingOutsideOfAWindow",
229
- "215": "WatchingReadingLookingAtABook",
230
- "216": "WatchingSomethingSomeoneThemselvesInAMirror",
231
- "217": "WatchingTelevision",
232
- "218": "WorkingAtATable",
233
- "219": "WorkingOnPaperNotebook",
234
- "220": "WorkingPlayingOnALaptop"
235
  },
236
  "image_size": 256,
237
  "in_chans": 3,
238
  "initializer_range": 0.02,
239
  "label2id": {
240
  "ApplyEyeMakeup": 1,
241
- "ApplyLipstick": 0,
242
- "Biking": 2,
243
- "BlowDryHair": 11,
244
- "BrushingTeeth": 3,
245
- "Clapping": 57,
246
- "ClosingABook": 64,
247
- "ClosingABox": 65,
248
- "ClosingAClosetCabinet": 66,
249
- "ClosingADoor": 67,
250
- "ClosingALaptop": 68,
251
- "ClosingARefrigerator": 69,
252
- "ClosingAWindow": 70,
253
- "CuttingInKitchen": 4,
254
- "DrinkAngry": 18,
255
- "DrinkDisgust": 19,
256
- "DrinkHappy": 20,
257
- "DrinkNeutral": 21,
258
- "DrinkSad": 22,
259
- "DrinkingFromACupGlassBottle": 71,
260
- "EatingASandwich": 72,
261
- "FallDown": 63,
262
- "Fighting": 55,
263
- "FixingADoor": 73,
264
- "FixingADoorknob": 74,
265
- "FixingALight": 75,
266
- "FixingAVacuum": 76,
267
- "FixingTheirHair": 77,
268
- "GraspingOntoADoorknob": 78,
269
- "Haircut": 5,
270
  "Hammering": 8,
271
- "HandstandPushups": 7,
272
- "HandstandWalking": 6,
273
- "HeadMassage": 13,
274
- "HoldingABag": 79,
275
- "HoldingABlanket": 80,
276
- "HoldingABook": 81,
277
- "HoldingABox": 82,
278
- "HoldingABroom": 83,
279
- "HoldingACupGlassBottleOfSomething": 84,
280
- "HoldingADish": 85,
281
- "HoldingALaptop": 86,
282
- "HoldingAMirror": 87,
283
- "HoldingAPhoneCamera": 88,
284
- "HoldingAPicture": 89,
285
- "HoldingAPillow": 90,
286
- "HoldingASandwich": 91,
287
- "HoldingAShoeShoes": 92,
288
- "HoldingATowelS": 93,
289
- "HoldingAVacuum": 94,
290
- "HoldingSomeClothes": 95,
291
- "HoldingSomeFood": 96,
292
- "HoldingSomeMedicine": 97,
293
- "Jump": 59,
294
- "Knitting": 9,
295
- "LaughingAtAPicture": 98,
296
- "LaughingAtTelevision": 99,
297
- "LyingDown": 61,
298
- "LyingOnABed": 100,
299
- "LyingOnASofaCouch": 101,
300
- "LyingOnTheFloor": 102,
301
- "MakingASandwich": 103,
302
- "OpeningABag": 104,
303
- "OpeningABook": 105,
304
- "OpeningABox": 106,
305
- "OpeningAClosetCabinet": 107,
306
- "OpeningADoor": 108,
307
- "OpeningALaptop": 109,
308
- "OpeningARefrigerator": 110,
309
- "OpeningAWindow": 111,
310
- "PlayingDaf": 12,
311
- "PlayingWithAPhoneCamera": 112,
312
- "PouringSomethingIntoACupGlassBottle": 113,
313
- "PutonglassesAngry": 23,
314
- "PutonglassesHappy": 24,
315
- "PutonglassesNeutral": 25,
316
- "PutonglassesSad": 26,
317
- "PutonjacketAngry": 27,
318
- "PutonjacketHappy": 28,
319
- "PutonjacketNeutral": 29,
320
- "PutonjacketSad": 30,
321
- "PuttingABagSomewhere": 114,
322
- "PuttingABlanketSomewhere": 115,
323
- "PuttingABookSomewhere": 116,
324
- "PuttingABoxSomewhere": 117,
325
- "PuttingABroomSomewhere": 118,
326
- "PuttingACupGlassBottleSomewhere": 119,
327
- "PuttingADishEsSomewhere": 120,
328
- "PuttingALaptopSomewhere": 121,
329
- "PuttingAPhoneCameraSomewhere": 122,
330
- "PuttingAPictureSomewhere": 123,
331
- "PuttingAPillowSomewhere": 124,
332
- "PuttingASandwichSomewhere": 125,
333
- "PuttingATowelSSomewhere": 126,
334
- "PuttingClothesSomewhere": 127,
335
- "PuttingGroceriesSomewhere": 128,
336
- "PuttingOnShoeShoes": 129,
337
- "PuttingShoesSomewhere": 130,
338
- "PuttingSomeFoodSomewhere": 131,
339
- "PuttingSomethingOnAShelf": 132,
340
- "PuttingSomethingOnATable": 133,
341
- "PuttingTheirPaperNotebookSomewhere": 134,
342
- "ReachingForAndGrabbingAPicture": 135,
343
- "ReadAngry": 31,
344
- "ReadHappy": 32,
345
- "ReadNeutral": 33,
346
- "ReadSad": 34,
347
- "SitDown": 60,
348
- "SitdownAngry": 35,
349
- "SitdownHappy": 36,
350
- "SitdownNeutral": 37,
351
- "SitdownSad": 38,
352
- "Sitting": 15,
353
- "SittingAtATable": 136,
354
- "SittingInABed": 137,
355
- "SittingInAChair": 138,
356
- "SittingOnATable": 139,
357
- "SittingOnSofaCouch": 140,
358
- "SittingOnTheFloor": 141,
359
- "SmilingAtABook": 142,
360
- "SmilingInAMirror": 143,
361
- "SnugglingWithABlanket": 144,
362
- "SnugglingWithAPillow": 145,
363
- "SomeoneIsAwakeningInBed": 146,
364
- "SomeoneIsAwakeningSomewhere": 147,
365
- "SomeoneIsCookingSomething": 148,
366
- "SomeoneIsDressing": 149,
367
- "SomeoneIsEatingSomething": 150,
368
- "SomeoneIsGoingFromStandingToSitting": 151,
369
- "SomeoneIsHoldingAPaperNotebook": 152,
370
- "SomeoneIsLaughing": 153,
371
- "SomeoneIsRunningSomewhere": 154,
372
- "SomeoneIsSmiling": 155,
373
- "SomeoneIsSneezing": 156,
374
- "SomeoneIsStandingUpFromSomewhere": 157,
375
  "SomeoneIsUndressing": 158,
376
- "StandUp": 62,
377
- "Standing": 16,
378
- "StandingOnAChair": 159,
379
- "StandupAngry": 39,
380
- "StandupHappy": 40,
381
- "StandupNeutral": 41,
382
- "StandupSad": 42,
383
- "TakeoffglassesAngry": 43,
384
- "TakeoffglassesHappy": 44,
385
- "TakeoffglassesNeutral": 45,
386
- "TakeoffglassesSad": 46,
387
- "TakeoffjacketAngry": 47,
388
- "TakeoffjacketHappy": 48,
389
- "TakeoffjacketNeutral": 49,
390
- "TakeoffjacketSad": 50,
391
- "TakingABagFromSomewhere": 160,
392
- "TakingABlanketFromSomewhere": 161,
393
- "TakingABookFromSomewhere": 162,
394
- "TakingABoxFromSomewhere": 163,
395
- "TakingABroomFromSomewhere": 164,
396
- "TakingACupGlassBottleFromSomewhere": 165,
397
- "TakingADishEsFromSomewhere": 166,
398
- "TakingALaptopFromSomewhere": 167,
399
  "TakingAPhoneCameraFromSomewhere": 168,
400
- "TakingAPictureOfSomething": 169,
401
- "TakingAPillowFromSomewhere": 170,
402
- "TakingASandwichFromSomewhere": 171,
403
- "TakingATowelSFromSomewhere": 172,
404
- "TakingAVacuumFromSomewhere": 173,
405
- "TakingConsumingSomeMedicine": 174,
406
- "TakingFoodFromSomewhere": 175,
407
- "TakingOffSomeShoes": 176,
408
- "TakingPaperNotebookFromSomewhere": 177,
409
- "TakingShoesFromSomewhere": 178,
410
- "TakingSomeClothesFromSomewhere": 179,
411
- "TakingSomethingFromABox": 180,
412
- "TalkingOnAPhoneCamera": 181,
413
- "ThrowingABagSomewhere": 182,
414
- "ThrowingABlanketSomewhere": 183,
415
- "ThrowingABookSomewhere": 184,
416
- "ThrowingABoxSomewhere": 185,
417
- "ThrowingABroomSomewhere": 186,
418
- "ThrowingAPillowSomewhere": 187,
419
- "ThrowingATowelSSomewhere": 188,
420
- "ThrowingClothesSomewhere": 189,
421
- "ThrowingFoodSomewhere": 190,
422
- "ThrowingShoesSomewhere": 191,
423
- "ThrowingSomethingOnTheFloor": 192,
424
- "TidyingAShelfOrSomethingOnAShelf": 193,
425
- "TidyingSomeClothes": 194,
426
- "TidyingSomethingOnTheFloor": 195,
427
- "TidyingUpABlanketS": 196,
428
- "TidyingUpAClosetCabinet": 197,
429
- "TidyingUpATable": 198,
430
- "TidyingUpATowelS": 199,
431
- "TidyingUpWithABroom": 200,
432
- "TurningOffALight": 201,
433
- "TurningOnALight": 202,
434
- "Typing": 10,
435
- "ViolenceFighting": 54,
436
- "Walking": 17,
437
- "WalkingThroughADoorway": 203,
438
- "WalkingWhileUsingPhone": 58,
439
- "WalkingWithDog": 14,
440
- "WashADishDishes": 204,
441
- "WashingACupGlassBottle": 205,
442
- "WashingAMirror": 206,
443
- "WashingATable": 207,
444
- "WashingAWindow": 208,
445
- "WashingSomeClothes": 209,
446
- "WashingSomethingWithATowel": 210,
447
- "WashingTheirHands": 211,
448
- "WatchingALaptopOrSomethingOnALaptop": 212,
449
- "WatchingLookingAtAPicture": 213,
450
- "WatchingLookingOutsideOfAWindow": 214,
451
- "WatchingReadingLookingAtABook": 215,
452
- "WatchingSomethingSomeoneThemselvesInAMirror": 216,
453
- "WatchingTelevision": 217,
454
- "Waving": 56,
455
- "WorkingAtATable": 218,
456
- "WorkingOnPaperNotebook": 219,
457
- "WorkingPlayingOnALaptop": 220,
458
- "WriteAngry": 51,
459
- "WriteHappy": 52,
460
- "WriteNeutral": 53
461
  },
462
  "layer_norm_eps": 1e-06,
463
  "mlp_ratio": 4,
 
11
  "hidden_dropout_prob": 0.0,
12
  "hidden_size": 1024,
13
  "id2label": {
14
+ "0": "Jump",
15
  "1": "ApplyEyeMakeup",
16
+ "2": "ApplyLipstick",
17
+ "3": "Biking",
18
+ "4": "BlowDryHair",
19
+ "5": "BrushingTeeth",
20
+ "6": "CuttingInKitchen",
21
+ "7": "Haircut",
22
  "8": "Hammering",
23
+ "9": "HandstandPushups",
24
+ "10": "HandstandWalking",
25
+ "11": "HeadMassage",
26
+ "12": "Knitting",
27
+ "13": "PlayingDaf",
28
+ "14": "Typing",
29
+ "15": "WalkingWithDog",
30
+ "16": "FallDown",
31
+ "17": "LyingDown",
32
+ "18": "SitDown",
33
+ "19": "Sitting",
34
+ "20": "StandUp",
35
+ "21": "Standing",
36
+ "22": "Walking",
37
+ "23": "DrinkAngry",
38
+ "24": "DrinkDisgust",
39
+ "25": "DrinkHappy",
40
+ "26": "DrinkNeutral",
41
+ "27": "DrinkSad",
42
+ "28": "PutonglassesAngry",
43
+ "29": "PutonglassesHappy",
44
+ "30": "PutonglassesNeutral",
45
+ "31": "PutonglassesSad",
46
+ "32": "PutonjacketAngry",
47
+ "33": "PutonjacketHappy",
48
+ "34": "PutonjacketNeutral",
49
+ "35": "PutonjacketSad",
50
+ "36": "ReadAngry",
51
+ "37": "ReadHappy",
52
+ "38": "ReadNeutral",
53
+ "39": "ReadSad",
54
+ "40": "SitdownAngry",
55
+ "41": "SitdownHappy",
56
+ "42": "SitdownNeutral",
57
+ "43": "SitdownSad",
58
+ "44": "StandupAngry",
59
+ "45": "StandupHappy",
60
+ "46": "StandupNeutral",
61
+ "47": "StandupSad",
62
+ "48": "TakeoffglassesAngry",
63
+ "49": "TakeoffglassesHappy",
64
+ "50": "TakeoffglassesNeutral",
65
+ "51": "TakeoffglassesSad",
66
+ "52": "TakeoffjacketAngry",
67
+ "53": "TakeoffjacketHappy",
68
+ "54": "TakeoffjacketNeutral",
69
+ "55": "TakeoffjacketSad",
70
+ "56": "WriteAngry",
71
+ "57": "WriteHappy",
72
+ "58": "WriteNeutral",
73
+ "59": "Waving",
74
+ "60": "Fighting",
75
+ "61": "Clapping",
76
+ "62": "WalkingWhileUsingPhone",
77
+ "63": "ClosingABook",
78
+ "64": "ClosingABox",
79
+ "65": "ClosingAClosetCabinet",
80
+ "66": "ClosingADoor",
81
+ "67": "ClosingALaptop",
82
+ "68": "ClosingARefrigerator",
83
+ "69": "ClosingAWindow",
84
+ "70": "DrinkingFromACupGlassBottle",
85
+ "71": "EatingASandwich",
86
+ "72": "FixingADoor",
87
+ "73": "FixingADoorknob",
88
+ "74": "FixingALight",
89
+ "75": "FixingAVacuum",
90
+ "76": "FixingTheirHair",
91
+ "77": "GraspingOntoADoorknob",
92
+ "78": "HoldingABag",
93
+ "79": "HoldingABlanket",
94
+ "80": "HoldingABook",
95
+ "81": "HoldingABox",
96
+ "82": "HoldingABroom",
97
+ "83": "HoldingACupGlassBottleOfSomething",
98
+ "84": "HoldingADish",
99
+ "85": "HoldingALaptop",
100
+ "86": "HoldingAMirror",
101
+ "87": "HoldingAPhoneCamera",
102
+ "88": "HoldingAPicture",
103
+ "89": "HoldingAPillow",
104
+ "90": "HoldingASandwich",
105
+ "91": "SomeoneIsRunningSomewhere",
106
+ "92": "SomeoneIsSmiling",
107
+ "93": "SomeoneIsSneezing",
108
+ "94": "SomeoneIsStandingUpFromSomewhere",
109
+ "95": "HoldingAShoeShoes",
110
+ "96": "HoldingATowelS",
111
+ "97": "HoldingAVacuum",
112
+ "98": "HoldingSomeClothes",
113
+ "99": "HoldingSomeFood",
114
+ "100": "HoldingSomeMedicine",
115
+ "101": "LaughingAtAPicture",
116
+ "102": "LaughingAtTelevision",
117
+ "103": "LyingOnABed",
118
+ "104": "LyingOnASofaCouch",
119
+ "105": "LyingOnTheFloor",
120
+ "106": "MakingASandwich",
121
+ "107": "OpeningABag",
122
+ "108": "OpeningABook",
123
+ "109": "OpeningABox",
124
+ "110": "OpeningAClosetCabinet",
125
+ "111": "OpeningADoor",
126
+ "112": "OpeningALaptop",
127
+ "113": "OpeningARefrigerator",
128
+ "114": "OpeningAWindow",
129
+ "115": "PlayingWithAPhoneCamera",
130
+ "116": "PouringSomethingIntoACupGlassBottle",
131
+ "117": "PuttingABagSomewhere",
132
+ "118": "PuttingABlanketSomewhere",
133
+ "119": "PuttingABookSomewhere",
134
+ "120": "PuttingABoxSomewhere",
135
+ "121": "PuttingABroomSomewhere",
136
+ "122": "PuttingACupGlassBottleSomewhere",
137
+ "123": "PuttingADishEsSomewhere",
138
+ "124": "PuttingALaptopSomewhere",
139
+ "125": "PuttingAPhoneCameraSomewhere",
140
+ "126": "PuttingAPictureSomewhere",
141
+ "127": "PuttingAPillowSomewhere",
142
+ "128": "PuttingASandwichSomewhere",
143
+ "129": "PuttingATowelSSomewhere",
144
+ "130": "PuttingClothesSomewhere",
145
+ "131": "PuttingGroceriesSomewhere",
146
+ "132": "PuttingOnShoeShoes",
147
+ "133": "PuttingShoesSomewhere",
148
+ "134": "PuttingSomeFoodSomewhere",
149
+ "135": "PuttingSomethingOnAShelf",
150
+ "136": "PuttingSomethingOnATable",
151
+ "137": "PuttingTheirPaperNotebookSomewhere",
152
+ "138": "ReachingForAndGrabbingAPicture",
153
+ "139": "SittingAtATable",
154
+ "140": "SittingInABed",
155
+ "141": "SittingInAChair",
156
+ "142": "SittingOnATable",
157
+ "143": "SittingOnSofaCouch",
158
+ "144": "SittingOnTheFloor",
159
+ "145": "SmilingAtABook",
160
+ "146": "SmilingInAMirror",
161
+ "147": "SnugglingWithABlanket",
162
+ "148": "SnugglingWithAPillow",
163
+ "149": "SomeoneIsAwakeningInBed",
164
+ "150": "SomeoneIsAwakeningSomewhere",
165
+ "151": "SomeoneIsCookingSomething",
166
+ "152": "SomeoneIsDressing",
167
+ "153": "SomeoneIsEatingSomething",
168
+ "154": "SomeoneIsGoingFromStandingToSitting",
169
+ "155": "SomeoneIsHoldingAPaperNotebook",
170
+ "156": "SomeoneIsLaughing",
171
+ "157": "StandingOnAChair",
172
  "158": "SomeoneIsUndressing",
173
+ "159": "TakingABlanketFromSomewhere",
174
+ "160": "TakingABookFromSomewhere",
175
+ "161": "TakingABoxFromSomewhere",
176
+ "162": "TakingADishEsFromSomewhere",
177
+ "163": "TakingALaptopFromSomewhere",
178
+ "164": "TakingAPictureOfSomething",
179
+ "165": "TakingABagFromSomewhere",
180
+ "166": "TakingABroomFromSomewhere",
181
+ "167": "TakingACupGlassBottleFromSomewhere",
182
  "168": "TakingAPhoneCameraFromSomewhere",
183
+ "169": "TakingAPillowFromSomewhere",
184
+ "170": "TakingASandwichFromSomewhere",
185
+ "171": "TakingATowelSFromSomewhere",
186
+ "172": "TakingAVacuumFromSomewhere",
187
+ "173": "TakingConsumingSomeMedicine",
188
+ "174": "TakingFoodFromSomewhere",
189
+ "175": "TakingOffSomeShoes",
190
+ "176": "TakingPaperNotebookFromSomewhere",
191
+ "177": "TakingShoesFromSomewhere",
192
+ "178": "TakingSomeClothesFromSomewhere",
193
+ "179": "TakingSomethingFromABox",
194
+ "180": "TalkingOnAPhoneCamera",
195
+ "181": "ThrowingABagSomewhere",
196
+ "182": "ThrowingABlanketSomewhere",
197
+ "183": "ThrowingABookSomewhere",
198
+ "184": "ThrowingABoxSomewhere",
199
+ "185": "ThrowingABroomSomewhere",
200
+ "186": "ThrowingAPillowSomewhere",
201
+ "187": "ThrowingATowelSSomewhere",
202
+ "188": "ThrowingClothesSomewhere",
203
+ "189": "ThrowingFoodSomewhere",
204
+ "190": "ThrowingShoesSomewhere",
205
+ "191": "ThrowingSomethingOnTheFloor",
206
+ "192": "TidyingAShelfOrSomethingOnAShelf",
207
+ "193": "TidyingSomeClothes",
208
+ "194": "TidyingSomethingOnTheFloor",
209
+ "195": "TidyingUpABlanketS",
210
+ "196": "TidyingUpAClosetCabinet",
211
+ "197": "TidyingUpATable",
212
+ "198": "TidyingUpATowelS",
213
+ "199": "TidyingUpWithABroom",
214
+ "200": "TurningOffALight",
215
+ "201": "TurningOnALight",
216
+ "202": "WalkingThroughADoorway",
217
+ "203": "WashADishDishes",
218
+ "204": "WashingACupGlassBottle",
219
+ "205": "WashingAMirror",
220
+ "206": "WashingATable",
221
+ "207": "WashingAWindow",
222
+ "208": "WashingSomeClothes",
223
+ "209": "WashingSomethingWithATowel",
224
+ "210": "WashingTheirHands",
225
+ "211": "WatchingALaptopOrSomethingOnALaptop",
226
+ "212": "WatchingLookingAtAPicture",
227
+ "213": "WatchingLookingOutsideOfAWindow",
228
+ "214": "WatchingReadingLookingAtABook",
229
+ "215": "WatchingSomethingSomeoneThemselvesInAMirror",
230
+ "216": "WatchingTelevision",
231
+ "217": "WorkingAtATable",
232
+ "218": "WorkingOnPaperNotebook",
233
+ "219": "WorkingPlayingOnALaptop"
 
234
  },
235
  "image_size": 256,
236
  "in_chans": 3,
237
  "initializer_range": 0.02,
238
  "label2id": {
239
  "ApplyEyeMakeup": 1,
240
+ "ApplyLipstick": 2,
241
+ "Biking": 3,
242
+ "BlowDryHair": 4,
243
+ "BrushingTeeth": 5,
244
+ "Clapping": 61,
245
+ "ClosingABook": 63,
246
+ "ClosingABox": 64,
247
+ "ClosingAClosetCabinet": 65,
248
+ "ClosingADoor": 66,
249
+ "ClosingALaptop": 67,
250
+ "ClosingARefrigerator": 68,
251
+ "ClosingAWindow": 69,
252
+ "CuttingInKitchen": 6,
253
+ "DrinkAngry": 23,
254
+ "DrinkDisgust": 24,
255
+ "DrinkHappy": 25,
256
+ "DrinkNeutral": 26,
257
+ "DrinkSad": 27,
258
+ "DrinkingFromACupGlassBottle": 70,
259
+ "EatingASandwich": 71,
260
+ "FallDown": 16,
261
+ "Fighting": 60,
262
+ "FixingADoor": 72,
263
+ "FixingADoorknob": 73,
264
+ "FixingALight": 74,
265
+ "FixingAVacuum": 75,
266
+ "FixingTheirHair": 76,
267
+ "GraspingOntoADoorknob": 77,
268
+ "Haircut": 7,
269
  "Hammering": 8,
270
+ "HandstandPushups": 9,
271
+ "HandstandWalking": 10,
272
+ "HeadMassage": 11,
273
+ "HoldingABag": 78,
274
+ "HoldingABlanket": 79,
275
+ "HoldingABook": 80,
276
+ "HoldingABox": 81,
277
+ "HoldingABroom": 82,
278
+ "HoldingACupGlassBottleOfSomething": 83,
279
+ "HoldingADish": 84,
280
+ "HoldingALaptop": 85,
281
+ "HoldingAMirror": 86,
282
+ "HoldingAPhoneCamera": 87,
283
+ "HoldingAPicture": 88,
284
+ "HoldingAPillow": 89,
285
+ "HoldingASandwich": 90,
286
+ "HoldingAShoeShoes": 95,
287
+ "HoldingATowelS": 96,
288
+ "HoldingAVacuum": 97,
289
+ "HoldingSomeClothes": 98,
290
+ "HoldingSomeFood": 99,
291
+ "HoldingSomeMedicine": 100,
292
+ "Jump": 0,
293
+ "Knitting": 12,
294
+ "LaughingAtAPicture": 101,
295
+ "LaughingAtTelevision": 102,
296
+ "LyingDown": 17,
297
+ "LyingOnABed": 103,
298
+ "LyingOnASofaCouch": 104,
299
+ "LyingOnTheFloor": 105,
300
+ "MakingASandwich": 106,
301
+ "OpeningABag": 107,
302
+ "OpeningABook": 108,
303
+ "OpeningABox": 109,
304
+ "OpeningAClosetCabinet": 110,
305
+ "OpeningADoor": 111,
306
+ "OpeningALaptop": 112,
307
+ "OpeningARefrigerator": 113,
308
+ "OpeningAWindow": 114,
309
+ "PlayingDaf": 13,
310
+ "PlayingWithAPhoneCamera": 115,
311
+ "PouringSomethingIntoACupGlassBottle": 116,
312
+ "PutonglassesAngry": 28,
313
+ "PutonglassesHappy": 29,
314
+ "PutonglassesNeutral": 30,
315
+ "PutonglassesSad": 31,
316
+ "PutonjacketAngry": 32,
317
+ "PutonjacketHappy": 33,
318
+ "PutonjacketNeutral": 34,
319
+ "PutonjacketSad": 35,
320
+ "PuttingABagSomewhere": 117,
321
+ "PuttingABlanketSomewhere": 118,
322
+ "PuttingABookSomewhere": 119,
323
+ "PuttingABoxSomewhere": 120,
324
+ "PuttingABroomSomewhere": 121,
325
+ "PuttingACupGlassBottleSomewhere": 122,
326
+ "PuttingADishEsSomewhere": 123,
327
+ "PuttingALaptopSomewhere": 124,
328
+ "PuttingAPhoneCameraSomewhere": 125,
329
+ "PuttingAPictureSomewhere": 126,
330
+ "PuttingAPillowSomewhere": 127,
331
+ "PuttingASandwichSomewhere": 128,
332
+ "PuttingATowelSSomewhere": 129,
333
+ "PuttingClothesSomewhere": 130,
334
+ "PuttingGroceriesSomewhere": 131,
335
+ "PuttingOnShoeShoes": 132,
336
+ "PuttingShoesSomewhere": 133,
337
+ "PuttingSomeFoodSomewhere": 134,
338
+ "PuttingSomethingOnAShelf": 135,
339
+ "PuttingSomethingOnATable": 136,
340
+ "PuttingTheirPaperNotebookSomewhere": 137,
341
+ "ReachingForAndGrabbingAPicture": 138,
342
+ "ReadAngry": 36,
343
+ "ReadHappy": 37,
344
+ "ReadNeutral": 38,
345
+ "ReadSad": 39,
346
+ "SitDown": 18,
347
+ "SitdownAngry": 40,
348
+ "SitdownHappy": 41,
349
+ "SitdownNeutral": 42,
350
+ "SitdownSad": 43,
351
+ "Sitting": 19,
352
+ "SittingAtATable": 139,
353
+ "SittingInABed": 140,
354
+ "SittingInAChair": 141,
355
+ "SittingOnATable": 142,
356
+ "SittingOnSofaCouch": 143,
357
+ "SittingOnTheFloor": 144,
358
+ "SmilingAtABook": 145,
359
+ "SmilingInAMirror": 146,
360
+ "SnugglingWithABlanket": 147,
361
+ "SnugglingWithAPillow": 148,
362
+ "SomeoneIsAwakeningInBed": 149,
363
+ "SomeoneIsAwakeningSomewhere": 150,
364
+ "SomeoneIsCookingSomething": 151,
365
+ "SomeoneIsDressing": 152,
366
+ "SomeoneIsEatingSomething": 153,
367
+ "SomeoneIsGoingFromStandingToSitting": 154,
368
+ "SomeoneIsHoldingAPaperNotebook": 155,
369
+ "SomeoneIsLaughing": 156,
370
+ "SomeoneIsRunningSomewhere": 91,
371
+ "SomeoneIsSmiling": 92,
372
+ "SomeoneIsSneezing": 93,
373
+ "SomeoneIsStandingUpFromSomewhere": 94,
374
  "SomeoneIsUndressing": 158,
375
+ "StandUp": 20,
376
+ "Standing": 21,
377
+ "StandingOnAChair": 157,
378
+ "StandupAngry": 44,
379
+ "StandupHappy": 45,
380
+ "StandupNeutral": 46,
381
+ "StandupSad": 47,
382
+ "TakeoffglassesAngry": 48,
383
+ "TakeoffglassesHappy": 49,
384
+ "TakeoffglassesNeutral": 50,
385
+ "TakeoffglassesSad": 51,
386
+ "TakeoffjacketAngry": 52,
387
+ "TakeoffjacketHappy": 53,
388
+ "TakeoffjacketNeutral": 54,
389
+ "TakeoffjacketSad": 55,
390
+ "TakingABagFromSomewhere": 165,
391
+ "TakingABlanketFromSomewhere": 159,
392
+ "TakingABookFromSomewhere": 160,
393
+ "TakingABoxFromSomewhere": 161,
394
+ "TakingABroomFromSomewhere": 166,
395
+ "TakingACupGlassBottleFromSomewhere": 167,
396
+ "TakingADishEsFromSomewhere": 162,
397
+ "TakingALaptopFromSomewhere": 163,
398
  "TakingAPhoneCameraFromSomewhere": 168,
399
+ "TakingAPictureOfSomething": 164,
400
+ "TakingAPillowFromSomewhere": 169,
401
+ "TakingASandwichFromSomewhere": 170,
402
+ "TakingATowelSFromSomewhere": 171,
403
+ "TakingAVacuumFromSomewhere": 172,
404
+ "TakingConsumingSomeMedicine": 173,
405
+ "TakingFoodFromSomewhere": 174,
406
+ "TakingOffSomeShoes": 175,
407
+ "TakingPaperNotebookFromSomewhere": 176,
408
+ "TakingShoesFromSomewhere": 177,
409
+ "TakingSomeClothesFromSomewhere": 178,
410
+ "TakingSomethingFromABox": 179,
411
+ "TalkingOnAPhoneCamera": 180,
412
+ "ThrowingABagSomewhere": 181,
413
+ "ThrowingABlanketSomewhere": 182,
414
+ "ThrowingABookSomewhere": 183,
415
+ "ThrowingABoxSomewhere": 184,
416
+ "ThrowingABroomSomewhere": 185,
417
+ "ThrowingAPillowSomewhere": 186,
418
+ "ThrowingATowelSSomewhere": 187,
419
+ "ThrowingClothesSomewhere": 188,
420
+ "ThrowingFoodSomewhere": 189,
421
+ "ThrowingShoesSomewhere": 190,
422
+ "ThrowingSomethingOnTheFloor": 191,
423
+ "TidyingAShelfOrSomethingOnAShelf": 192,
424
+ "TidyingSomeClothes": 193,
425
+ "TidyingSomethingOnTheFloor": 194,
426
+ "TidyingUpABlanketS": 195,
427
+ "TidyingUpAClosetCabinet": 196,
428
+ "TidyingUpATable": 197,
429
+ "TidyingUpATowelS": 198,
430
+ "TidyingUpWithABroom": 199,
431
+ "TurningOffALight": 200,
432
+ "TurningOnALight": 201,
433
+ "Typing": 14,
434
+ "Walking": 22,
435
+ "WalkingThroughADoorway": 202,
436
+ "WalkingWhileUsingPhone": 62,
437
+ "WalkingWithDog": 15,
438
+ "WashADishDishes": 203,
439
+ "WashingACupGlassBottle": 204,
440
+ "WashingAMirror": 205,
441
+ "WashingATable": 206,
442
+ "WashingAWindow": 207,
443
+ "WashingSomeClothes": 208,
444
+ "WashingSomethingWithATowel": 209,
445
+ "WashingTheirHands": 210,
446
+ "WatchingALaptopOrSomethingOnALaptop": 211,
447
+ "WatchingLookingAtAPicture": 212,
448
+ "WatchingLookingOutsideOfAWindow": 213,
449
+ "WatchingReadingLookingAtABook": 214,
450
+ "WatchingSomethingSomeoneThemselvesInAMirror": 215,
451
+ "WatchingTelevision": 216,
452
+ "Waving": 59,
453
+ "WorkingAtATable": 217,
454
+ "WorkingOnPaperNotebook": 218,
455
+ "WorkingPlayingOnALaptop": 219,
456
+ "WriteAngry": 56,
457
+ "WriteHappy": 57,
458
+ "WriteNeutral": 58
 
459
  },
460
  "layer_norm_eps": 1e-06,
461
  "mlp_ratio": 4,
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8f00c1bc53c993368bdb2d9af030be09924e79af3d311427060e530d0f7fc299
3
- size 1502211292
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48b31dca615bb1b91099d78e09808899eecae07bda2ac7f8b73517e0063e6eaa
3
+ size 1502207192