├── .ipynb_checkpoints
    └── DeepSegmentation-checkpoint.ipynb
├── Data
    ├── CLASSES.pkl
    ├── classid_to_wnid.pkl
    ├── validation_1.pkl
    ├── validation_10.pkl
    ├── validation_11.pkl
    ├── validation_12.pkl
    ├── validation_13.pkl
    ├── validation_2.pkl
    ├── validation_3.pkl
    ├── validation_4.pkl
    ├── validation_5.pkl
    ├── validation_6.pkl
    ├── validation_7.pkl
    ├── validation_8.pkl
    ├── validation_9.pkl
    ├── validation_bottle.pkl
    ├── validation_cat.pkl
    └── validation_cycles.pkl
├── Images
    ├── apple.jpg
    ├── beach.jpg
    ├── beetroot.jpg
    ├── cat.jpg
    ├── dog.jpg
    ├── dolph.jpg
    ├── einstein.jpg
    ├── kitten.jpg
    ├── orange.jpg
    ├── pig.html
    ├── pig.jpg
    ├── pig_with_dog.jpg
    ├── rose.jpg
    └── sky.jpg
├── Localization.ipynb
├── Poster.pdf
├── README.md
├── Validation.ipynb
├── deconv_utils.py
├── frameworkpython
├── library
    ├── __init__.py
    ├── classifiers
    │   ├── __init__.py
    │   └── pretrained_vgg16.py
    ├── data_utils.py
    ├── fast_layers.py
    ├── frameworkpython
    ├── im2col.py
    ├── im2col_cython.c
    ├── im2col_cython.pyx
    ├── image_utils.py
    ├── layer_utils.py
    ├── layers.py
    ├── localization.py
    └── setup.py
└── validation_script.py


/Data/CLASSES.pkl:
--------------------------------------------------------------------------------
   1 | (lp0
   2 | S'tench, Tinca tinca'
   3 | p1
   4 | aS'goldfish, Carassius auratus'
   5 | p2
   6 | aS'great white shark, white shark, man-eater, man-eating shark, Carcharodon carcharias'
   7 | p3
   8 | aS'tiger shark, Galeocerdo cuvieri'
   9 | p4
  10 | aS'hammerhead, hammerhead shark'
  11 | p5
  12 | aS'electric ray, crampfish, numbfish, torpedo'
  13 | p6
  14 | aS'stingray'
  15 | p7
  16 | aS'cock'
  17 | p8
  18 | aS'hen'
  19 | p9
  20 | aS'ostrich, Struthio camelus'
  21 | p10
  22 | aS'brambling, Fringilla montifringilla'
  23 | p11
  24 | aS'goldfinch, Carduelis carduelis'
  25 | p12
  26 | aS'house finch, linnet, Carpodacus mexicanus'
  27 | p13
  28 | aS'junco, snowbird'
  29 | p14
  30 | aS'indigo bunting, indigo finch, indigo bird, Passerina cyanea'
  31 | p15
  32 | aS'robin, American robin, Turdus migratorius'
  33 | p16
  34 | aS'bulbul'
  35 | p17
  36 | aS'jay'
  37 | p18
  38 | aS'magpie'
  39 | p19
  40 | aS'chickadee'
  41 | p20
  42 | aS'water ouzel, dipper'
  43 | p21
  44 | aS'kite'
  45 | p22
  46 | aS'bald eagle, American eagle, Haliaeetus leucocephalus'
  47 | p23
  48 | aS'vulture'
  49 | p24
  50 | aS'great grey owl, great gray owl, Strix nebulosa'
  51 | p25
  52 | aS'European fire salamander, Salamandra salamandra'
  53 | p26
  54 | aS'common newt, Triturus vulgaris'
  55 | p27
  56 | aS'eft'
  57 | p28
  58 | aS'spotted salamander, Ambystoma maculatum'
  59 | p29
  60 | aS'axolotl, mud puppy, Ambystoma mexicanum'
  61 | p30
  62 | aS'bullfrog, Rana catesbeiana'
  63 | p31
  64 | aS'tree frog, tree-frog'
  65 | p32
  66 | aS'tailed frog, bell toad, ribbed toad, tailed toad, Ascaphus trui'
  67 | p33
  68 | aS'loggerhead, loggerhead turtle, Caretta caretta'
  69 | p34
  70 | aS'leatherback turtle, leatherback, leathery turtle, Dermochelys coriacea'
  71 | p35
  72 | aS'mud turtle'
  73 | p36
  74 | aS'terrapin'
  75 | p37
  76 | aS'box turtle, box tortoise'
  77 | p38
  78 | aS'banded gecko'
  79 | p39
  80 | aS'common iguana, iguana, Iguana iguana'
  81 | p40
  82 | aS'American chameleon, anole, Anolis carolinensis'
  83 | p41
  84 | aS'whiptail, whiptail lizard'
  85 | p42
  86 | aS'agama'
  87 | p43
  88 | aS'frilled lizard, Chlamydosaurus kingi'
  89 | p44
  90 | aS'alligator lizard'
  91 | p45
  92 | aS'Gila monster, Heloderma suspectum'
  93 | p46
  94 | aS'green lizard, Lacerta viridis'
  95 | p47
  96 | aS'African chameleon, Chamaeleo chamaeleon'
  97 | p48
  98 | aS'Komodo dragon, Komodo lizard, dragon lizard, giant lizard, Varanus komodoensis'
  99 | p49
 100 | aS'African crocodile, Nile crocodile, Crocodylus niloticus'
 101 | p50
 102 | aS'American alligator, Alligator mississipiensis'
 103 | p51
 104 | aS'triceratops'
 105 | p52
 106 | aS'thunder snake, worm snake, Carphophis amoenus'
 107 | p53
 108 | aS'ringneck snake, ring-necked snake, ring snake'
 109 | p54
 110 | aS'hognose snake, puff adder, sand viper'
 111 | p55
 112 | aS'green snake, grass snake'
 113 | p56
 114 | aS'king snake, kingsnake'
 115 | p57
 116 | aS'garter snake, grass snake'
 117 | p58
 118 | aS'water snake'
 119 | p59
 120 | aS'vine snake'
 121 | p60
 122 | aS'night snake, Hypsiglena torquata'
 123 | p61
 124 | aS'boa constrictor, Constrictor constrictor'
 125 | p62
 126 | aS'rock python, rock snake, Python sebae'
 127 | p63
 128 | aS'Indian cobra, Naja naja'
 129 | p64
 130 | aS'green mamba'
 131 | p65
 132 | aS'sea snake'
 133 | p66
 134 | aS'horned viper, cerastes, sand viper, horned asp, Cerastes cornutus'
 135 | p67
 136 | aS'diamondback, diamondback rattlesnake, Crotalus adamanteus'
 137 | p68
 138 | aS'sidewinder, horned rattlesnake, Crotalus cerastes'
 139 | p69
 140 | aS'trilobite'
 141 | p70
 142 | aS'harvestman, daddy longlegs, Phalangium opilio'
 143 | p71
 144 | aS'scorpion'
 145 | p72
 146 | aS'black and gold garden spider, Argiope aurantia'
 147 | p73
 148 | aS'barn spider, Araneus cavaticus'
 149 | p74
 150 | aS'garden spider, Aranea diademata'
 151 | p75
 152 | aS'black widow, Latrodectus mactans'
 153 | p76
 154 | aS'tarantula'
 155 | p77
 156 | aS'wolf spider, hunting spider'
 157 | p78
 158 | aS'tick'
 159 | p79
 160 | aS'centipede'
 161 | p80
 162 | aS'black grouse'
 163 | p81
 164 | aS'ptarmigan'
 165 | p82
 166 | aS'ruffed grouse, partridge, Bonasa umbellus'
 167 | p83
 168 | aS'prairie chicken, prairie grouse, prairie fowl'
 169 | p84
 170 | aS'peacock'
 171 | p85
 172 | aS'quail'
 173 | p86
 174 | aS'partridge'
 175 | p87
 176 | aS'African grey, African gray, Psittacus erithacus'
 177 | p88
 178 | aS'macaw'
 179 | p89
 180 | aS'sulphur-crested cockatoo, Kakatoe galerita, Cacatua galerita'
 181 | p90
 182 | aS'lorikeet'
 183 | p91
 184 | aS'coucal'
 185 | p92
 186 | aS'bee eater'
 187 | p93
 188 | aS'hornbill'
 189 | p94
 190 | aS'hummingbird'
 191 | p95
 192 | aS'jacamar'
 193 | p96
 194 | aS'toucan'
 195 | p97
 196 | aS'drake'
 197 | p98
 198 | aS'red-breasted merganser, Mergus serrator'
 199 | p99
 200 | aS'goose'
 201 | p100
 202 | aS'black swan, Cygnus atratus'
 203 | p101
 204 | aS'tusker'
 205 | p102
 206 | aS'echidna, spiny anteater, anteater'
 207 | p103
 208 | aS'platypus, duckbill, duckbilled platypus, duck-billed platypus, Ornithorhynchus anatinus'
 209 | p104
 210 | aS'wallaby, brush kangaroo'
 211 | p105
 212 | aS'koala, koala bear, kangaroo bear, native bear, Phascolarctos cinereus'
 213 | p106
 214 | aS'wombat'
 215 | p107
 216 | aS'jellyfish'
 217 | p108
 218 | aS'sea anemone, anemone'
 219 | p109
 220 | aS'brain coral'
 221 | p110
 222 | aS'flatworm, platyhelminth'
 223 | p111
 224 | aS'nematode, nematode worm, roundworm'
 225 | p112
 226 | aS'conch'
 227 | p113
 228 | aS'snail'
 229 | p114
 230 | aS'slug'
 231 | p115
 232 | aS'sea slug, nudibranch'
 233 | p116
 234 | aS'chiton, coat-of-mail shell, sea cradle, polyplacophore'
 235 | p117
 236 | aS'chambered nautilus, pearly nautilus, nautilus'
 237 | p118
 238 | aS'Dungeness crab, Cancer magister'
 239 | p119
 240 | aS'rock crab, Cancer irroratus'
 241 | p120
 242 | aS'fiddler crab'
 243 | p121
 244 | aS'king crab, Alaska crab, Alaskan king crab, Alaska king crab, Paralithodes camtschatica'
 245 | p122
 246 | aS'American lobster, Northern lobster, Maine lobster, Homarus americanus'
 247 | p123
 248 | aS'spiny lobster, langouste, rock lobster, crawfish, crayfish, sea crawfish'
 249 | p124
 250 | aS'crayfish, crawfish, crawdad, crawdaddy'
 251 | p125
 252 | aS'hermit crab'
 253 | p126
 254 | aS'isopod'
 255 | p127
 256 | aS'white stork, Ciconia ciconia'
 257 | p128
 258 | aS'black stork, Ciconia nigra'
 259 | p129
 260 | aS'spoonbill'
 261 | p130
 262 | aS'flamingo'
 263 | p131
 264 | aS'little blue heron, Egretta caerulea'
 265 | p132
 266 | aS'American egret, great white heron, Egretta albus'
 267 | p133
 268 | aS'bittern'
 269 | p134
 270 | aS'crane'
 271 | p135
 272 | aS'limpkin, Aramus pictus'
 273 | p136
 274 | aS'European gallinule, Porphyrio porphyrio'
 275 | p137
 276 | aS'American coot, marsh hen, mud hen, water hen, Fulica americana'
 277 | p138
 278 | aS'bustard'
 279 | p139
 280 | aS'ruddy turnstone, Arenaria interpres'
 281 | p140
 282 | aS'red-backed sandpiper, dunlin, Erolia alpina'
 283 | p141
 284 | aS'redshank, Tringa totanus'
 285 | p142
 286 | aS'dowitcher'
 287 | p143
 288 | aS'oystercatcher, oyster catcher'
 289 | p144
 290 | aS'pelican'
 291 | p145
 292 | aS'king penguin, Aptenodytes patagonica'
 293 | p146
 294 | aS'albatross, mollymawk'
 295 | p147
 296 | aS'grey whale, gray whale, devilfish, Eschrichtius gibbosus, Eschrichtius robustus'
 297 | p148
 298 | aS'killer whale, killer, orca, grampus, sea wolf, Orcinus orca'
 299 | p149
 300 | aS'dugong, Dugong dugon'
 301 | p150
 302 | aS'sea lion'
 303 | p151
 304 | aS'Chihuahua'
 305 | p152
 306 | aS'Japanese spaniel'
 307 | p153
 308 | aS'Maltese dog, Maltese terrier, Maltese'
 309 | p154
 310 | aS'Pekinese, Pekingese, Peke'
 311 | p155
 312 | aS'Shih-Tzu'
 313 | p156
 314 | aS'Blenheim spaniel'
 315 | p157
 316 | aS'papillon'
 317 | p158
 318 | aS'toy terrier'
 319 | p159
 320 | aS'Rhodesian ridgeback'
 321 | p160
 322 | aS'Afghan hound, Afghan'
 323 | p161
 324 | aS'basset, basset hound'
 325 | p162
 326 | aS'beagle'
 327 | p163
 328 | aS'bloodhound, sleuthhound'
 329 | p164
 330 | aS'bluetick'
 331 | p165
 332 | aS'black-and-tan coonhound'
 333 | p166
 334 | aS'Walker hound, Walker foxhound'
 335 | p167
 336 | aS'English foxhound'
 337 | p168
 338 | aS'redbone'
 339 | p169
 340 | aS'borzoi, Russian wolfhound'
 341 | p170
 342 | aS'Irish wolfhound'
 343 | p171
 344 | aS'Italian greyhound'
 345 | p172
 346 | aS'whippet'
 347 | p173
 348 | aS'Ibizan hound, Ibizan Podenco'
 349 | p174
 350 | aS'Norwegian elkhound, elkhound'
 351 | p175
 352 | aS'otterhound, otter hound'
 353 | p176
 354 | aS'Saluki, gazelle hound'
 355 | p177
 356 | aS'Scottish deerhound, deerhound'
 357 | p178
 358 | aS'Weimaraner'
 359 | p179
 360 | aS'Staffordshire bullterrier, Staffordshire bull terrier'
 361 | p180
 362 | aS'American Staffordshire terrier, Staffordshire terrier, American pit bull terrier, pit bull terrier'
 363 | p181
 364 | aS'Bedlington terrier'
 365 | p182
 366 | aS'Border terrier'
 367 | p183
 368 | aS'Kerry blue terrier'
 369 | p184
 370 | aS'Irish terrier'
 371 | p185
 372 | aS'Norfolk terrier'
 373 | p186
 374 | aS'Norwich terrier'
 375 | p187
 376 | aS'Yorkshire terrier'
 377 | p188
 378 | aS'wire-haired fox terrier'
 379 | p189
 380 | aS'Lakeland terrier'
 381 | p190
 382 | aS'Sealyham terrier, Sealyham'
 383 | p191
 384 | aS'Airedale, Airedale terrier'
 385 | p192
 386 | aS'cairn, cairn terrier'
 387 | p193
 388 | aS'Australian terrier'
 389 | p194
 390 | aS'Dandie Dinmont, Dandie Dinmont terrier'
 391 | p195
 392 | aS'Boston bull, Boston terrier'
 393 | p196
 394 | aS'miniature schnauzer'
 395 | p197
 396 | aS'giant schnauzer'
 397 | p198
 398 | aS'standard schnauzer'
 399 | p199
 400 | aS'Scotch terrier, Scottish terrier, Scottie'
 401 | p200
 402 | aS'Tibetan terrier, chrysanthemum dog'
 403 | p201
 404 | aS'silky terrier, Sydney silky'
 405 | p202
 406 | aS'soft-coated wheaten terrier'
 407 | p203
 408 | aS'West Highland white terrier'
 409 | p204
 410 | aS'Lhasa, Lhasa apso'
 411 | p205
 412 | aS'flat-coated retriever'
 413 | p206
 414 | aS'curly-coated retriever'
 415 | p207
 416 | aS'golden retriever'
 417 | p208
 418 | aS'Labrador retriever'
 419 | p209
 420 | aS'Chesapeake Bay retriever'
 421 | p210
 422 | aS'German short-haired pointer'
 423 | p211
 424 | aS'vizsla, Hungarian pointer'
 425 | p212
 426 | aS'English setter'
 427 | p213
 428 | aS'Irish setter, red setter'
 429 | p214
 430 | aS'Gordon setter'
 431 | p215
 432 | aS'Brittany spaniel'
 433 | p216
 434 | aS'clumber, clumber spaniel'
 435 | p217
 436 | aS'English springer, English springer spaniel'
 437 | p218
 438 | aS'Welsh springer spaniel'
 439 | p219
 440 | aS'cocker spaniel, English cocker spaniel, cocker'
 441 | p220
 442 | aS'Sussex spaniel'
 443 | p221
 444 | aS'Irish water spaniel'
 445 | p222
 446 | aS'kuvasz'
 447 | p223
 448 | aS'schipperke'
 449 | p224
 450 | aS'groenendael'
 451 | p225
 452 | aS'malinois'
 453 | p226
 454 | aS'briard'
 455 | p227
 456 | aS'kelpie'
 457 | p228
 458 | aS'komondor'
 459 | p229
 460 | aS'Old English sheepdog, bobtail'
 461 | p230
 462 | aS'Shetland sheepdog, Shetland sheep dog, Shetland'
 463 | p231
 464 | aS'collie'
 465 | p232
 466 | aS'Border collie'
 467 | p233
 468 | aS'Bouvier des Flandres, Bouviers des Flandres'
 469 | p234
 470 | aS'Rottweiler'
 471 | p235
 472 | aS'German shepherd, German shepherd dog, German police dog, alsatian'
 473 | p236
 474 | aS'Doberman, Doberman pinscher'
 475 | p237
 476 | aS'miniature pinscher'
 477 | p238
 478 | aS'Greater Swiss Mountain dog'
 479 | p239
 480 | aS'Bernese mountain dog'
 481 | p240
 482 | aS'Appenzeller'
 483 | p241
 484 | aS'EntleBucher'
 485 | p242
 486 | aS'boxer'
 487 | p243
 488 | aS'bull mastiff'
 489 | p244
 490 | aS'Tibetan mastiff'
 491 | p245
 492 | aS'French bulldog'
 493 | p246
 494 | aS'Great Dane'
 495 | p247
 496 | aS'Saint Bernard, St Bernard'
 497 | p248
 498 | aS'Eskimo dog, husky'
 499 | p249
 500 | aS'malamute, malemute, Alaskan malamute'
 501 | p250
 502 | aS'Siberian husky'
 503 | p251
 504 | aS'dalmatian, coach dog, carriage dog'
 505 | p252
 506 | aS'affenpinscher, monkey pinscher, monkey dog'
 507 | p253
 508 | aS'basenji'
 509 | p254
 510 | aS'pug, pug-dog'
 511 | p255
 512 | aS'Leonberg'
 513 | p256
 514 | aS'Newfoundland, Newfoundland dog'
 515 | p257
 516 | aS'Great Pyrenees'
 517 | p258
 518 | aS'Samoyed, Samoyede'
 519 | p259
 520 | aS'Pomeranian'
 521 | p260
 522 | aS'chow, chow chow'
 523 | p261
 524 | aS'keeshond'
 525 | p262
 526 | aS'Brabancon griffon'
 527 | p263
 528 | aS'Pembroke, Pembroke Welsh corgi'
 529 | p264
 530 | aS'Cardigan, Cardigan Welsh corgi'
 531 | p265
 532 | aS'toy poodle'
 533 | p266
 534 | aS'miniature poodle'
 535 | p267
 536 | aS'standard poodle'
 537 | p268
 538 | aS'Mexican hairless'
 539 | p269
 540 | aS'timber wolf, grey wolf, gray wolf, Canis lupus'
 541 | p270
 542 | aS'white wolf, Arctic wolf, Canis lupus tundrarum'
 543 | p271
 544 | aS'red wolf, maned wolf, Canis rufus, Canis niger'
 545 | p272
 546 | aS'coyote, prairie wolf, brush wolf, Canis latrans'
 547 | p273
 548 | aS'dingo, warrigal, warragal, Canis dingo'
 549 | p274
 550 | aS'dhole, Cuon alpinus'
 551 | p275
 552 | aS'African hunting dog, hyena dog, Cape hunting dog, Lycaon pictus'
 553 | p276
 554 | aS'hyena, hyaena'
 555 | p277
 556 | aS'red fox, Vulpes vulpes'
 557 | p278
 558 | aS'kit fox, Vulpes macrotis'
 559 | p279
 560 | aS'Arctic fox, white fox, Alopex lagopus'
 561 | p280
 562 | aS'grey fox, gray fox, Urocyon cinereoargenteus'
 563 | p281
 564 | aS'tabby, tabby cat'
 565 | p282
 566 | aS'tiger cat'
 567 | p283
 568 | aS'Persian cat'
 569 | p284
 570 | aS'Siamese cat, Siamese'
 571 | p285
 572 | aS'Egyptian cat'
 573 | p286
 574 | aS'cougar, puma, catamount, mountain lion, painter, panther, Felis concolor'
 575 | p287
 576 | aS'lynx, catamount'
 577 | p288
 578 | aS'leopard, Panthera pardus'
 579 | p289
 580 | aS'snow leopard, ounce, Panthera uncia'
 581 | p290
 582 | aS'jaguar, panther, Panthera onca, Felis onca'
 583 | p291
 584 | aS'lion, king of beasts, Panthera leo'
 585 | p292
 586 | aS'tiger, Panthera tigris'
 587 | p293
 588 | aS'cheetah, chetah, Acinonyx jubatus'
 589 | p294
 590 | aS'brown bear, bruin, Ursus arctos'
 591 | p295
 592 | aS'American black bear, black bear, Ursus americanus, Euarctos americanus'
 593 | p296
 594 | aS'ice bear, polar bear, Ursus Maritimus, Thalarctos maritimus'
 595 | p297
 596 | aS'sloth bear, Melursus ursinus, Ursus ursinus'
 597 | p298
 598 | aS'mongoose'
 599 | p299
 600 | aS'meerkat, mierkat'
 601 | p300
 602 | aS'tiger beetle'
 603 | p301
 604 | aS'ladybug, ladybeetle, lady beetle, ladybird, ladybird beetle'
 605 | p302
 606 | aS'ground beetle, carabid beetle'
 607 | p303
 608 | aS'long-horned beetle, longicorn, longicorn beetle'
 609 | p304
 610 | aS'leaf beetle, chrysomelid'
 611 | p305
 612 | aS'dung beetle'
 613 | p306
 614 | aS'rhinoceros beetle'
 615 | p307
 616 | aS'weevil'
 617 | p308
 618 | aS'fly'
 619 | p309
 620 | aS'bee'
 621 | p310
 622 | aS'ant, emmet, pismire'
 623 | p311
 624 | aS'grasshopper, hopper'
 625 | p312
 626 | aS'cricket'
 627 | p313
 628 | aS'walking stick, walkingstick, stick insect'
 629 | p314
 630 | aS'cockroach, roach'
 631 | p315
 632 | aS'mantis, mantid'
 633 | p316
 634 | aS'cicada, cicala'
 635 | p317
 636 | aS'leafhopper'
 637 | p318
 638 | aS'lacewing, lacewing fly'
 639 | p319
 640 | aS"dragonfly, darning needle, devil's darning needle, sewing needle, snake feeder, snake doctor, mosquito hawk, skeeter hawk"
 641 | p320
 642 | aS'damselfly'
 643 | p321
 644 | aS'admiral'
 645 | p322
 646 | aS'ringlet, ringlet butterfly'
 647 | p323
 648 | aS'monarch, monarch butterfly, milkweed butterfly, Danaus plexippus'
 649 | p324
 650 | aS'cabbage butterfly'
 651 | p325
 652 | aS'sulphur butterfly, sulfur butterfly'
 653 | p326
 654 | aS'lycaenid, lycaenid butterfly'
 655 | p327
 656 | aS'starfish, sea star'
 657 | p328
 658 | aS'sea urchin'
 659 | p329
 660 | aS'sea cucumber, holothurian'
 661 | p330
 662 | aS'wood rabbit, cottontail, cottontail rabbit'
 663 | p331
 664 | aS'hare'
 665 | p332
 666 | aS'Angora, Angora rabbit'
 667 | p333
 668 | aS'hamster'
 669 | p334
 670 | aS'porcupine, hedgehog'
 671 | p335
 672 | aS'fox squirrel, eastern fox squirrel, Sciurus niger'
 673 | p336
 674 | aS'marmot'
 675 | p337
 676 | aS'beaver'
 677 | p338
 678 | aS'guinea pig, Cavia cobaya'
 679 | p339
 680 | aS'sorrel'
 681 | p340
 682 | aS'zebra'
 683 | p341
 684 | aS'hog, pig, grunter, squealer, Sus scrofa'
 685 | p342
 686 | aS'wild boar, boar, Sus scrofa'
 687 | p343
 688 | aS'warthog'
 689 | p344
 690 | aS'hippopotamus, hippo, river horse, Hippopotamus amphibius'
 691 | p345
 692 | aS'ox'
 693 | p346
 694 | aS'water buffalo, water ox, Asiatic buffalo, Bubalus bubalis'
 695 | p347
 696 | aS'bison'
 697 | p348
 698 | aS'ram, tup'
 699 | p349
 700 | aS'bighorn, bighorn sheep, cimarron, Rocky Mountain bighorn, Rocky Mountain sheep, Ovis canadensis'
 701 | p350
 702 | aS'ibex, Capra ibex'
 703 | p351
 704 | aS'hartebeest'
 705 | p352
 706 | aS'impala, Aepyceros melampus'
 707 | p353
 708 | aS'gazelle'
 709 | p354
 710 | aS'Arabian camel, dromedary, Camelus dromedarius'
 711 | p355
 712 | aS'llama'
 713 | p356
 714 | aS'weasel'
 715 | p357
 716 | aS'mink'
 717 | p358
 718 | aS'polecat, fitch, foulmart, foumart, Mustela putorius'
 719 | p359
 720 | aS'black-footed ferret, ferret, Mustela nigripes'
 721 | p360
 722 | aS'otter'
 723 | p361
 724 | aS'skunk, polecat, wood pussy'
 725 | p362
 726 | aS'badger'
 727 | p363
 728 | aS'armadillo'
 729 | p364
 730 | aS'three-toed sloth, ai, Bradypus tridactylus'
 731 | p365
 732 | aS'orangutan, orang, orangutang, Pongo pygmaeus'
 733 | p366
 734 | aS'gorilla, Gorilla gorilla'
 735 | p367
 736 | aS'chimpanzee, chimp, Pan troglodytes'
 737 | p368
 738 | aS'gibbon, Hylobates lar'
 739 | p369
 740 | aS'siamang, Hylobates syndactylus, Symphalangus syndactylus'
 741 | p370
 742 | aS'guenon, guenon monkey'
 743 | p371
 744 | aS'patas, hussar monkey, Erythrocebus patas'
 745 | p372
 746 | aS'baboon'
 747 | p373
 748 | aS'macaque'
 749 | p374
 750 | aS'langur'
 751 | p375
 752 | aS'colobus, colobus monkey'
 753 | p376
 754 | aS'proboscis monkey, Nasalis larvatus'
 755 | p377
 756 | aS'marmoset'
 757 | p378
 758 | aS'capuchin, ringtail, Cebus capucinus'
 759 | p379
 760 | aS'howler monkey, howler'
 761 | p380
 762 | aS'titi, titi monkey'
 763 | p381
 764 | aS'spider monkey, Ateles geoffroyi'
 765 | p382
 766 | aS'squirrel monkey, Saimiri sciureus'
 767 | p383
 768 | aS'Madagascar cat, ring-tailed lemur, Lemur catta'
 769 | p384
 770 | aS'indri, indris, Indri indri, Indri brevicaudatus'
 771 | p385
 772 | aS'Indian elephant, Elephas maximus'
 773 | p386
 774 | aS'African elephant, Loxodonta africana'
 775 | p387
 776 | aS'lesser panda, red panda, panda, bear cat, cat bear, Ailurus fulgens'
 777 | p388
 778 | aS'giant panda, panda, panda bear, coon bear, Ailuropoda melanoleuca'
 779 | p389
 780 | aS'barracouta, snoek'
 781 | p390
 782 | aS'eel'
 783 | p391
 784 | aS'coho, cohoe, coho salmon, blue jack, silver salmon, Oncorhynchus kisutch'
 785 | p392
 786 | aS'rock beauty, Holocanthus tricolor'
 787 | p393
 788 | aS'anemone fish'
 789 | p394
 790 | aS'sturgeon'
 791 | p395
 792 | aS'gar, garfish, garpike, billfish, Lepisosteus osseus'
 793 | p396
 794 | aS'lionfish'
 795 | p397
 796 | aS'puffer, pufferfish, blowfish, globefish'
 797 | p398
 798 | aS'abacus'
 799 | p399
 800 | aS'abaya'
 801 | p400
 802 | aS"academic gown, academic robe, judge's robe"
 803 | p401
 804 | aS'accordion, piano accordion, squeeze box'
 805 | p402
 806 | aS'acoustic guitar'
 807 | p403
 808 | aS'aircraft carrier, carrier, flattop, attack aircraft carrier'
 809 | p404
 810 | aS'airliner'
 811 | p405
 812 | aS'airship, dirigible'
 813 | p406
 814 | aS'altar'
 815 | p407
 816 | aS'ambulance'
 817 | p408
 818 | aS'amphibian, amphibious vehicle'
 819 | p409
 820 | aS'analog clock'
 821 | p410
 822 | aS'apiary, bee house'
 823 | p411
 824 | aS'apron'
 825 | p412
 826 | aS'ashcan, trash can, garbage can, wastebin, ash bin, ash-bin, ashbin, dustbin, trash barrel, trash bin'
 827 | p413
 828 | aS'assault rifle, assault gun'
 829 | p414
 830 | aS'backpack, back pack, knapsack, packsack, rucksack, haversack'
 831 | p415
 832 | aS'bakery, bakeshop, bakehouse'
 833 | p416
 834 | aS'balance beam, beam'
 835 | p417
 836 | aS'balloon'
 837 | p418
 838 | aS'ballpoint, ballpoint pen, ballpen, Biro'
 839 | p419
 840 | aS'Band Aid'
 841 | p420
 842 | aS'banjo'
 843 | p421
 844 | aS'bannister, banister, balustrade, balusters, handrail'
 845 | p422
 846 | aS'barbell'
 847 | p423
 848 | aS'barber chair'
 849 | p424
 850 | aS'barbershop'
 851 | p425
 852 | aS'barn'
 853 | p426
 854 | aS'barometer'
 855 | p427
 856 | aS'barrel, cask'
 857 | p428
 858 | aS'barrow, garden cart, lawn cart, wheelbarrow'
 859 | p429
 860 | aS'baseball'
 861 | p430
 862 | aS'basketball'
 863 | p431
 864 | aS'bassinet'
 865 | p432
 866 | aS'bassoon'
 867 | p433
 868 | aS'bathing cap, swimming cap'
 869 | p434
 870 | aS'bath towel'
 871 | p435
 872 | aS'bathtub, bathing tub, bath, tub'
 873 | p436
 874 | aS'beach wagon, station wagon, wagon, estate car, beach waggon, station waggon, waggon'
 875 | p437
 876 | aS'beacon, lighthouse, beacon light, pharos'
 877 | p438
 878 | aS'beaker'
 879 | p439
 880 | aS'bearskin, busby, shako'
 881 | p440
 882 | aS'beer bottle'
 883 | p441
 884 | aS'beer glass'
 885 | p442
 886 | aS'bell cote, bell cot'
 887 | p443
 888 | aS'bib'
 889 | p444
 890 | aS'bicycle-built-for-two, tandem bicycle, tandem'
 891 | p445
 892 | aS'bikini, two-piece'
 893 | p446
 894 | aS'binder, ring-binder'
 895 | p447
 896 | aS'binoculars, field glasses, opera glasses'
 897 | p448
 898 | aS'birdhouse'
 899 | p449
 900 | aS'boathouse'
 901 | p450
 902 | aS'bobsled, bobsleigh, bob'
 903 | p451
 904 | aS'bolo tie, bolo, bola tie, bola'
 905 | p452
 906 | aS'bonnet, poke bonnet'
 907 | p453
 908 | aS'bookcase'
 909 | p454
 910 | aS'bookshop, bookstore, bookstall'
 911 | p455
 912 | aS'bottlecap'
 913 | p456
 914 | aS'bow'
 915 | p457
 916 | aS'bow tie, bow-tie, bowtie'
 917 | p458
 918 | aS'brass, memorial tablet, plaque'
 919 | p459
 920 | aS'brassiere, bra, bandeau'
 921 | p460
 922 | aS'breakwater, groin, groyne, mole, bulwark, seawall, jetty'
 923 | p461
 924 | aS'breastplate, aegis, egis'
 925 | p462
 926 | aS'broom'
 927 | p463
 928 | aS'bucket, pail'
 929 | p464
 930 | aS'buckle'
 931 | p465
 932 | aS'bulletproof vest'
 933 | p466
 934 | aS'bullet train, bullet'
 935 | p467
 936 | aS'butcher shop, meat market'
 937 | p468
 938 | aS'cab, hack, taxi, taxicab'
 939 | p469
 940 | aS'caldron, cauldron'
 941 | p470
 942 | aS'candle, taper, wax light'
 943 | p471
 944 | aS'cannon'
 945 | p472
 946 | aS'canoe'
 947 | p473
 948 | aS'can opener, tin opener'
 949 | p474
 950 | aS'cardigan'
 951 | p475
 952 | aS'car mirror'
 953 | p476
 954 | aS'carousel, carrousel, merry-go-round, roundabout, whirligig'
 955 | p477
 956 | aS"carpenter's kit, tool kit"
 957 | p478
 958 | aS'carton'
 959 | p479
 960 | aS'car wheel'
 961 | p480
 962 | aS'cash machine, cash dispenser, automated teller machine, automatic teller machine, automated teller, automatic teller, ATM'
 963 | p481
 964 | aS'cassette'
 965 | p482
 966 | aS'cassette player'
 967 | p483
 968 | aS'castle'
 969 | p484
 970 | aS'catamaran'
 971 | p485
 972 | aS'CD player'
 973 | p486
 974 | aS'cello, violoncello'
 975 | p487
 976 | aS'cellular telephone, cellular phone, cellphone, cell, mobile phone'
 977 | p488
 978 | aS'chain'
 979 | p489
 980 | aS'chainlink fence'
 981 | p490
 982 | aS'chain mail, ring mail, mail, chain armor, chain armour, ring armor, ring armour'
 983 | p491
 984 | aS'chain saw, chainsaw'
 985 | p492
 986 | aS'chest'
 987 | p493
 988 | aS'chiffonier, commode'
 989 | p494
 990 | aS'chime, bell, gong'
 991 | p495
 992 | aS'china cabinet, china closet'
 993 | p496
 994 | aS'Christmas stocking'
 995 | p497
 996 | aS'church, church building'
 997 | p498
 998 | aS'cinema, movie theater, movie theatre, movie house, picture palace'
 999 | p499
1000 | aS'cleaver, meat cleaver, chopper'
1001 | p500
1002 | aS'cliff dwelling'
1003 | p501
1004 | aS'cloak'
1005 | p502
1006 | aS'clog, geta, patten, sabot'
1007 | p503
1008 | aS'cocktail shaker'
1009 | p504
1010 | aS'coffee mug'
1011 | p505
1012 | aS'coffeepot'
1013 | p506
1014 | aS'coil, spiral, volute, whorl, helix'
1015 | p507
1016 | aS'combination lock'
1017 | p508
1018 | aS'computer keyboard, keypad'
1019 | p509
1020 | aS'confectionery, confectionary, candy store'
1021 | p510
1022 | aS'container ship, containership, container vessel'
1023 | p511
1024 | aS'convertible'
1025 | p512
1026 | aS'corkscrew, bottle screw'
1027 | p513
1028 | aS'cornet, horn, trumpet, trump'
1029 | p514
1030 | aS'cowboy boot'
1031 | p515
1032 | aS'cowboy hat, ten-gallon hat'
1033 | p516
1034 | aS'cradle'
1035 | p517
1036 | aS'crane'
1037 | p518
1038 | aS'crash helmet'
1039 | p519
1040 | aS'crate'
1041 | p520
1042 | aS'crib, cot'
1043 | p521
1044 | aS'Crock Pot'
1045 | p522
1046 | aS'croquet ball'
1047 | p523
1048 | aS'crutch'
1049 | p524
1050 | aS'cuirass'
1051 | p525
1052 | aS'dam, dike, dyke'
1053 | p526
1054 | aS'desk'
1055 | p527
1056 | aS'desktop computer'
1057 | p528
1058 | aS'dial telephone, dial phone'
1059 | p529
1060 | aS'diaper, nappy, napkin'
1061 | p530
1062 | aS'digital clock'
1063 | p531
1064 | aS'digital watch'
1065 | p532
1066 | aS'dining table, board'
1067 | p533
1068 | aS'dishrag, dishcloth'
1069 | p534
1070 | aS'dishwasher, dish washer, dishwashing machine'
1071 | p535
1072 | aS'disk brake, disc brake'
1073 | p536
1074 | aS'dock, dockage, docking facility'
1075 | p537
1076 | aS'dogsled, dog sled, dog sleigh'
1077 | p538
1078 | aS'dome'
1079 | p539
1080 | aS'doormat, welcome mat'
1081 | p540
1082 | aS'drilling platform, offshore rig'
1083 | p541
1084 | aS'drum, membranophone, tympan'
1085 | p542
1086 | aS'drumstick'
1087 | p543
1088 | aS'dumbbell'
1089 | p544
1090 | aS'Dutch oven'
1091 | p545
1092 | aS'electric fan, blower'
1093 | p546
1094 | aS'electric guitar'
1095 | p547
1096 | aS'electric locomotive'
1097 | p548
1098 | aS'entertainment center'
1099 | p549
1100 | aS'envelope'
1101 | p550
1102 | aS'espresso maker'
1103 | p551
1104 | aS'face powder'
1105 | p552
1106 | aS'feather boa, boa'
1107 | p553
1108 | aS'file, file cabinet, filing cabinet'
1109 | p554
1110 | aS'fireboat'
1111 | p555
1112 | aS'fire engine, fire truck'
1113 | p556
1114 | aS'fire screen, fireguard'
1115 | p557
1116 | aS'flagpole, flagstaff'
1117 | p558
1118 | aS'flute, transverse flute'
1119 | p559
1120 | aS'folding chair'
1121 | p560
1122 | aS'football helmet'
1123 | p561
1124 | aS'forklift'
1125 | p562
1126 | aS'fountain'
1127 | p563
1128 | aS'fountain pen'
1129 | p564
1130 | aS'four-poster'
1131 | p565
1132 | aS'freight car'
1133 | p566
1134 | aS'French horn, horn'
1135 | p567
1136 | aS'frying pan, frypan, skillet'
1137 | p568
1138 | aS'fur coat'
1139 | p569
1140 | aS'garbage truck, dustcart'
1141 | p570
1142 | aS'gasmask, respirator, gas helmet'
1143 | p571
1144 | aS'gas pump, gasoline pump, petrol pump, island dispenser'
1145 | p572
1146 | aS'goblet'
1147 | p573
1148 | aS'go-kart'
1149 | p574
1150 | aS'golf ball'
1151 | p575
1152 | aS'golfcart, golf cart'
1153 | p576
1154 | aS'gondola'
1155 | p577
1156 | aS'gong, tam-tam'
1157 | p578
1158 | aS'gown'
1159 | p579
1160 | aS'grand piano, grand'
1161 | p580
1162 | aS'greenhouse, nursery, glasshouse'
1163 | p581
1164 | aS'grille, radiator grille'
1165 | p582
1166 | aS'grocery store, grocery, food market, market'
1167 | p583
1168 | aS'guillotine'
1169 | p584
1170 | aS'hair slide'
1171 | p585
1172 | aS'hair spray'
1173 | p586
1174 | aS'half track'
1175 | p587
1176 | aS'hammer'
1177 | p588
1178 | aS'hamper'
1179 | p589
1180 | aS'hand blower, blow dryer, blow drier, hair dryer, hair drier'
1181 | p590
1182 | aS'hand-held computer, hand-held microcomputer'
1183 | p591
1184 | aS'handkerchief, hankie, hanky, hankey'
1185 | p592
1186 | aS'hard disc, hard disk, fixed disk'
1187 | p593
1188 | aS'harmonica, mouth organ, harp, mouth harp'
1189 | p594
1190 | aS'harp'
1191 | p595
1192 | aS'harvester, reaper'
1193 | p596
1194 | aS'hatchet'
1195 | p597
1196 | aS'holster'
1197 | p598
1198 | aS'home theater, home theatre'
1199 | p599
1200 | aS'honeycomb'
1201 | p600
1202 | aS'hook, claw'
1203 | p601
1204 | aS'hoopskirt, crinoline'
1205 | p602
1206 | aS'horizontal bar, high bar'
1207 | p603
1208 | aS'horse cart, horse-cart'
1209 | p604
1210 | aS'hourglass'
1211 | p605
1212 | aS'iPod'
1213 | p606
1214 | aS'iron, smoothing iron'
1215 | p607
1216 | aS"jack-o'-lantern"
1217 | p608
1218 | aS'jean, blue jean, denim'
1219 | p609
1220 | aS'jeep, landrover'
1221 | p610
1222 | aS'jersey, T-shirt, tee shirt'
1223 | p611
1224 | aS'jigsaw puzzle'
1225 | p612
1226 | aS'jinrikisha, ricksha, rickshaw'
1227 | p613
1228 | aS'joystick'
1229 | p614
1230 | aS'kimono'
1231 | p615
1232 | aS'knee pad'
1233 | p616
1234 | aS'knot'
1235 | p617
1236 | aS'lab coat, laboratory coat'
1237 | p618
1238 | aS'ladle'
1239 | p619
1240 | aS'lampshade, lamp shade'
1241 | p620
1242 | aS'laptop, laptop computer'
1243 | p621
1244 | aS'lawn mower, mower'
1245 | p622
1246 | aS'lens cap, lens cover'
1247 | p623
1248 | aS'letter opener, paper knife, paperknife'
1249 | p624
1250 | aS'library'
1251 | p625
1252 | aS'lifeboat'
1253 | p626
1254 | aS'lighter, light, igniter, ignitor'
1255 | p627
1256 | aS'limousine, limo'
1257 | p628
1258 | aS'liner, ocean liner'
1259 | p629
1260 | aS'lipstick, lip rouge'
1261 | p630
1262 | aS'Loafer'
1263 | p631
1264 | aS'lotion'
1265 | p632
1266 | aS'loudspeaker, speaker, speaker unit, loudspeaker system, speaker system'
1267 | p633
1268 | aS"loupe, jeweler's loupe"
1269 | p634
1270 | aS'lumbermill, sawmill'
1271 | p635
1272 | aS'magnetic compass'
1273 | p636
1274 | aS'mailbag, postbag'
1275 | p637
1276 | aS'mailbox, letter box'
1277 | p638
1278 | aS'maillot'
1279 | p639
1280 | aS'maillot, tank suit'
1281 | p640
1282 | aS'manhole cover'
1283 | p641
1284 | aS'maraca'
1285 | p642
1286 | aS'marimba, xylophone'
1287 | p643
1288 | aS'mask'
1289 | p644
1290 | aS'matchstick'
1291 | p645
1292 | aS'maypole'
1293 | p646
1294 | aS'maze, labyrinth'
1295 | p647
1296 | aS'measuring cup'
1297 | p648
1298 | aS'medicine chest, medicine cabinet'
1299 | p649
1300 | aS'megalith, megalithic structure'
1301 | p650
1302 | aS'microphone, mike'
1303 | p651
1304 | aS'microwave, microwave oven'
1305 | p652
1306 | aS'military uniform'
1307 | p653
1308 | aS'milk can'
1309 | p654
1310 | aS'minibus'
1311 | p655
1312 | aS'miniskirt, mini'
1313 | p656
1314 | aS'minivan'
1315 | p657
1316 | aS'missile'
1317 | p658
1318 | aS'mitten'
1319 | p659
1320 | aS'mixing bowl'
1321 | p660
1322 | aS'mobile home, manufactured home'
1323 | p661
1324 | aS'Model T'
1325 | p662
1326 | aS'modem'
1327 | p663
1328 | aS'monastery'
1329 | p664
1330 | aS'monitor'
1331 | p665
1332 | aS'moped'
1333 | p666
1334 | aS'mortar'
1335 | p667
1336 | aS'mortarboard'
1337 | p668
1338 | aS'mosque'
1339 | p669
1340 | aS'mosquito net'
1341 | p670
1342 | aS'motor scooter, scooter'
1343 | p671
1344 | aS'mountain bike, all-terrain bike, off-roader'
1345 | p672
1346 | aS'mountain tent'
1347 | p673
1348 | aS'mouse, computer mouse'
1349 | p674
1350 | aS'mousetrap'
1351 | p675
1352 | aS'moving van'
1353 | p676
1354 | aS'muzzle'
1355 | p677
1356 | aS'nail'
1357 | p678
1358 | aS'neck brace'
1359 | p679
1360 | aS'necklace'
1361 | p680
1362 | aS'nipple'
1363 | p681
1364 | aS'notebook, notebook computer'
1365 | p682
1366 | aS'obelisk'
1367 | p683
1368 | aS'oboe, hautboy, hautbois'
1369 | p684
1370 | aS'ocarina, sweet potato'
1371 | p685
1372 | aS'odometer, hodometer, mileometer, milometer'
1373 | p686
1374 | aS'oil filter'
1375 | p687
1376 | aS'organ, pipe organ'
1377 | p688
1378 | aS'oscilloscope, scope, cathode-ray oscilloscope, CRO'
1379 | p689
1380 | aS'overskirt'
1381 | p690
1382 | aS'oxcart'
1383 | p691
1384 | aS'oxygen mask'
1385 | p692
1386 | aS'packet'
1387 | p693
1388 | aS'paddle, boat paddle'
1389 | p694
1390 | aS'paddlewheel, paddle wheel'
1391 | p695
1392 | aS'padlock'
1393 | p696
1394 | aS'paintbrush'
1395 | p697
1396 | aS"pajama, pyjama, pj's, jammies"
1397 | p698
1398 | aS'palace'
1399 | p699
1400 | aS'panpipe, pandean pipe, syrinx'
1401 | p700
1402 | aS'paper towel'
1403 | p701
1404 | aS'parachute, chute'
1405 | p702
1406 | aS'parallel bars, bars'
1407 | p703
1408 | aS'park bench'
1409 | p704
1410 | aS'parking meter'
1411 | p705
1412 | aS'passenger car, coach, carriage'
1413 | p706
1414 | aS'patio, terrace'
1415 | p707
1416 | aS'pay-phone, pay-station'
1417 | p708
1418 | aS'pedestal, plinth, footstall'
1419 | p709
1420 | aS'pencil box, pencil case'
1421 | p710
1422 | aS'pencil sharpener'
1423 | p711
1424 | aS'perfume, essence'
1425 | p712
1426 | aS'Petri dish'
1427 | p713
1428 | aS'photocopier'
1429 | p714
1430 | aS'pick, plectrum, plectron'
1431 | p715
1432 | aS'pickelhaube'
1433 | p716
1434 | aS'picket fence, paling'
1435 | p717
1436 | aS'pickup, pickup truck'
1437 | p718
1438 | aS'pier'
1439 | p719
1440 | aS'piggy bank, penny bank'
1441 | p720
1442 | aS'pill bottle'
1443 | p721
1444 | aS'pillow'
1445 | p722
1446 | aS'ping-pong ball'
1447 | p723
1448 | aS'pinwheel'
1449 | p724
1450 | aS'pirate, pirate ship'
1451 | p725
1452 | aS'pitcher, ewer'
1453 | p726
1454 | aS"plane, carpenter's plane, woodworking plane"
1455 | p727
1456 | aS'planetarium'
1457 | p728
1458 | aS'plastic bag'
1459 | p729
1460 | aS'plate rack'
1461 | p730
1462 | aS'plow, plough'
1463 | p731
1464 | aS"plunger, plumber's helper"
1465 | p732
1466 | aS'Polaroid camera, Polaroid Land camera'
1467 | p733
1468 | aS'pole'
1469 | p734
1470 | aS'police van, police wagon, paddy wagon, patrol wagon, wagon, black Maria'
1471 | p735
1472 | aS'poncho'
1473 | p736
1474 | aS'pool table, billiard table, snooker table'
1475 | p737
1476 | aS'pop bottle, soda bottle'
1477 | p738
1478 | aS'pot, flowerpot'
1479 | p739
1480 | aS"potter's wheel"
1481 | p740
1482 | aS'power drill'
1483 | p741
1484 | aS'prayer rug, prayer mat'
1485 | p742
1486 | aS'printer'
1487 | p743
1488 | aS'prison, prison house'
1489 | p744
1490 | aS'projectile, missile'
1491 | p745
1492 | aS'projector'
1493 | p746
1494 | aS'puck, hockey puck'
1495 | p747
1496 | aS'punching bag, punch bag, punching ball, punchball'
1497 | p748
1498 | aS'purse'
1499 | p749
1500 | aS'quill, quill pen'
1501 | p750
1502 | aS'quilt, comforter, comfort, puff'
1503 | p751
1504 | aS'racer, race car, racing car'
1505 | p752
1506 | aS'racket, racquet'
1507 | p753
1508 | aS'radiator'
1509 | p754
1510 | aS'radio, wireless'
1511 | p755
1512 | aS'radio telescope, radio reflector'
1513 | p756
1514 | aS'rain barrel'
1515 | p757
1516 | aS'recreational vehicle, RV, R.V.'
1517 | p758
1518 | aS'reel'
1519 | p759
1520 | aS'reflex camera'
1521 | p760
1522 | aS'refrigerator, icebox'
1523 | p761
1524 | aS'remote control, remote'
1525 | p762
1526 | aS'restaurant, eating house, eating place, eatery'
1527 | p763
1528 | aS'revolver, six-gun, six-shooter'
1529 | p764
1530 | aS'rifle'
1531 | p765
1532 | aS'rocking chair, rocker'
1533 | p766
1534 | aS'rotisserie'
1535 | p767
1536 | aS'rubber eraser, rubber, pencil eraser'
1537 | p768
1538 | aS'rugby ball'
1539 | p769
1540 | aS'rule, ruler'
1541 | p770
1542 | aS'running shoe'
1543 | p771
1544 | aS'safe'
1545 | p772
1546 | aS'safety pin'
1547 | p773
1548 | aS'saltshaker, salt shaker'
1549 | p774
1550 | aS'sandal'
1551 | p775
1552 | aS'sarong'
1553 | p776
1554 | aS'sax, saxophone'
1555 | p777
1556 | aS'scabbard'
1557 | p778
1558 | aS'scale, weighing machine'
1559 | p779
1560 | aS'school bus'
1561 | p780
1562 | aS'schooner'
1563 | p781
1564 | aS'scoreboard'
1565 | p782
1566 | aS'screen, CRT screen'
1567 | p783
1568 | aS'screw'
1569 | p784
1570 | aS'screwdriver'
1571 | p785
1572 | aS'seat belt, seatbelt'
1573 | p786
1574 | aS'sewing machine'
1575 | p787
1576 | aS'shield, buckler'
1577 | p788
1578 | aS'shoe shop, shoe-shop, shoe store'
1579 | p789
1580 | aS'shoji'
1581 | p790
1582 | aS'shopping basket'
1583 | p791
1584 | aS'shopping cart'
1585 | p792
1586 | aS'shovel'
1587 | p793
1588 | aS'shower cap'
1589 | p794
1590 | aS'shower curtain'
1591 | p795
1592 | aS'ski'
1593 | p796
1594 | aS'ski mask'
1595 | p797
1596 | aS'sleeping bag'
1597 | p798
1598 | aS'slide rule, slipstick'
1599 | p799
1600 | aS'sliding door'
1601 | p800
1602 | aS'slot, one-armed bandit'
1603 | p801
1604 | aS'snorkel'
1605 | p802
1606 | aS'snowmobile'
1607 | p803
1608 | aS'snowplow, snowplough'
1609 | p804
1610 | aS'soap dispenser'
1611 | p805
1612 | aS'soccer ball'
1613 | p806
1614 | aS'sock'
1615 | p807
1616 | aS'solar dish, solar collector, solar furnace'
1617 | p808
1618 | aS'sombrero'
1619 | p809
1620 | aS'soup bowl'
1621 | p810
1622 | aS'space bar'
1623 | p811
1624 | aS'space heater'
1625 | p812
1626 | aS'space shuttle'
1627 | p813
1628 | aS'spatula'
1629 | p814
1630 | aS'speedboat'
1631 | p815
1632 | aS"spider web, spider's web"
1633 | p816
1634 | aS'spindle'
1635 | p817
1636 | aS'sports car, sport car'
1637 | p818
1638 | aS'spotlight, spot'
1639 | p819
1640 | aS'stage'
1641 | p820
1642 | aS'steam locomotive'
1643 | p821
1644 | aS'steel arch bridge'
1645 | p822
1646 | aS'steel drum'
1647 | p823
1648 | aS'stethoscope'
1649 | p824
1650 | aS'stole'
1651 | p825
1652 | aS'stone wall'
1653 | p826
1654 | aS'stopwatch, stop watch'
1655 | p827
1656 | aS'stove'
1657 | p828
1658 | aS'strainer'
1659 | p829
1660 | aS'streetcar, tram, tramcar, trolley, trolley car'
1661 | p830
1662 | aS'stretcher'
1663 | p831
1664 | aS'studio couch, day bed'
1665 | p832
1666 | aS'stupa, tope'
1667 | p833
1668 | aS'submarine, pigboat, sub, U-boat'
1669 | p834
1670 | aS'suit, suit of clothes'
1671 | p835
1672 | aS'sundial'
1673 | p836
1674 | aS'sunglass'
1675 | p837
1676 | aS'sunglasses, dark glasses, shades'
1677 | p838
1678 | aS'sunscreen, sunblock, sun blocker'
1679 | p839
1680 | aS'suspension bridge'
1681 | p840
1682 | aS'swab, swob, mop'
1683 | p841
1684 | aS'sweatshirt'
1685 | p842
1686 | aS'swimming trunks, bathing trunks'
1687 | p843
1688 | aS'swing'
1689 | p844
1690 | aS'switch, electric switch, electrical switch'
1691 | p845
1692 | aS'syringe'
1693 | p846
1694 | aS'table lamp'
1695 | p847
1696 | aS'tank, army tank, armored combat vehicle, armoured combat vehicle'
1697 | p848
1698 | aS'tape player'
1699 | p849
1700 | aS'teapot'
1701 | p850
1702 | aS'teddy, teddy bear'
1703 | p851
1704 | aS'television, television system'
1705 | p852
1706 | aS'tennis ball'
1707 | p853
1708 | aS'thatch, thatched roof'
1709 | p854
1710 | aS'theater curtain, theatre curtain'
1711 | p855
1712 | aS'thimble'
1713 | p856
1714 | aS'thresher, thrasher, threshing machine'
1715 | p857
1716 | aS'throne'
1717 | p858
1718 | aS'tile roof'
1719 | p859
1720 | aS'toaster'
1721 | p860
1722 | aS'tobacco shop, tobacconist shop, tobacconist'
1723 | p861
1724 | aS'toilet seat'
1725 | p862
1726 | aS'torch'
1727 | p863
1728 | aS'totem pole'
1729 | p864
1730 | aS'tow truck, tow car, wrecker'
1731 | p865
1732 | aS'toyshop'
1733 | p866
1734 | aS'tractor'
1735 | p867
1736 | aS'trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi'
1737 | p868
1738 | aS'tray'
1739 | p869
1740 | aS'trench coat'
1741 | p870
1742 | aS'tricycle, trike, velocipede'
1743 | p871
1744 | aS'trimaran'
1745 | p872
1746 | aS'tripod'
1747 | p873
1748 | aS'triumphal arch'
1749 | p874
1750 | aS'trolleybus, trolley coach, trackless trolley'
1751 | p875
1752 | aS'trombone'
1753 | p876
1754 | aS'tub, vat'
1755 | p877
1756 | aS'turnstile'
1757 | p878
1758 | aS'typewriter keyboard'
1759 | p879
1760 | aS'umbrella'
1761 | p880
1762 | aS'unicycle, monocycle'
1763 | p881
1764 | aS'upright, upright piano'
1765 | p882
1766 | aS'vacuum, vacuum cleaner'
1767 | p883
1768 | aS'vase'
1769 | p884
1770 | aS'vault'
1771 | p885
1772 | aS'velvet'
1773 | p886
1774 | aS'vending machine'
1775 | p887
1776 | aS'vestment'
1777 | p888
1778 | aS'viaduct'
1779 | p889
1780 | aS'violin, fiddle'
1781 | p890
1782 | aS'volleyball'
1783 | p891
1784 | aS'waffle iron'
1785 | p892
1786 | aS'wall clock'
1787 | p893
1788 | aS'wallet, billfold, notecase, pocketbook'
1789 | p894
1790 | aS'wardrobe, closet, press'
1791 | p895
1792 | aS'warplane, military plane'
1793 | p896
1794 | aS'washbasin, handbasin, washbowl, lavabo, wash-hand basin'
1795 | p897
1796 | aS'washer, automatic washer, washing machine'
1797 | p898
1798 | aS'water bottle'
1799 | p899
1800 | aS'water jug'
1801 | p900
1802 | aS'water tower'
1803 | p901
1804 | aS'whiskey jug'
1805 | p902
1806 | aS'whistle'
1807 | p903
1808 | aS'wig'
1809 | p904
1810 | aS'window screen'
1811 | p905
1812 | aS'window shade'
1813 | p906
1814 | aS'Windsor tie'
1815 | p907
1816 | aS'wine bottle'
1817 | p908
1818 | aS'wing'
1819 | p909
1820 | aS'wok'
1821 | p910
1822 | aS'wooden spoon'
1823 | p911
1824 | aS'wool, woolen, woollen'
1825 | p912
1826 | aS'worm fence, snake fence, snake-rail fence, Virginia fence'
1827 | p913
1828 | aS'wreck'
1829 | p914
1830 | aS'yawl'
1831 | p915
1832 | aS'yurt'
1833 | p916
1834 | aS'web site, website, internet site, site'
1835 | p917
1836 | aS'comic book'
1837 | p918
1838 | aS'crossword puzzle, crossword'
1839 | p919
1840 | aS'street sign'
1841 | p920
1842 | aS'traffic light, traffic signal, stoplight'
1843 | p921
1844 | aS'book jacket, dust cover, dust jacket, dust wrapper'
1845 | p922
1846 | aS'menu'
1847 | p923
1848 | aS'plate'
1849 | p924
1850 | aS'guacamole'
1851 | p925
1852 | aS'consomme'
1853 | p926
1854 | aS'hot pot, hotpot'
1855 | p927
1856 | aS'trifle'
1857 | p928
1858 | aS'ice cream, icecream'
1859 | p929
1860 | aS'ice lolly, lolly, lollipop, popsicle'
1861 | p930
1862 | aS'French loaf'
1863 | p931
1864 | aS'bagel, beigel'
1865 | p932
1866 | aS'pretzel'
1867 | p933
1868 | aS'cheeseburger'
1869 | p934
1870 | aS'hotdog, hot dog, red hot'
1871 | p935
1872 | aS'mashed potato'
1873 | p936
1874 | aS'head cabbage'
1875 | p937
1876 | aS'broccoli'
1877 | p938
1878 | aS'cauliflower'
1879 | p939
1880 | aS'zucchini, courgette'
1881 | p940
1882 | aS'spaghetti squash'
1883 | p941
1884 | aS'acorn squash'
1885 | p942
1886 | aS'butternut squash'
1887 | p943
1888 | aS'cucumber, cuke'
1889 | p944
1890 | aS'artichoke, globe artichoke'
1891 | p945
1892 | aS'bell pepper'
1893 | p946
1894 | aS'cardoon'
1895 | p947
1896 | aS'mushroom'
1897 | p948
1898 | aS'Granny Smith'
1899 | p949
1900 | aS'strawberry'
1901 | p950
1902 | aS'orange'
1903 | p951
1904 | aS'lemon'
1905 | p952
1906 | aS'fig'
1907 | p953
1908 | aS'pineapple, ananas'
1909 | p954
1910 | aS'banana'
1911 | p955
1912 | aS'jackfruit, jak, jack'
1913 | p956
1914 | aS'custard apple'
1915 | p957
1916 | aS'pomegranate'
1917 | p958
1918 | aS'hay'
1919 | p959
1920 | aS'carbonara'
1921 | p960
1922 | aS'chocolate sauce, chocolate syrup'
1923 | p961
1924 | aS'dough'
1925 | p962
1926 | aS'meat loaf, meatloaf'
1927 | p963
1928 | aS'pizza, pizza pie'
1929 | p964
1930 | aS'potpie'
1931 | p965
1932 | aS'burrito'
1933 | p966
1934 | aS'red wine'
1935 | p967
1936 | aS'espresso'
1937 | p968
1938 | aS'cup'
1939 | p969
1940 | aS'eggnog'
1941 | p970
1942 | aS'alp'
1943 | p971
1944 | aS'bubble'
1945 | p972
1946 | aS'cliff, drop, drop-off'
1947 | p973
1948 | aS'coral reef'
1949 | p974
1950 | aS'geyser'
1951 | p975
1952 | aS'lakeside, lakeshore'
1953 | p976
1954 | aS'promontory, headland, head, foreland'
1955 | p977
1956 | aS'sandbar, sand bar'
1957 | p978
1958 | aS'seashore, coast, seacoast, sea-coast'
1959 | p979
1960 | aS'valley, vale'
1961 | p980
1962 | aS'volcano'
1963 | p981
1964 | aS'ballplayer, baseball player'
1965 | p982
1966 | aS'groom, bridegroom'
1967 | p983
1968 | aS'scuba diver'
1969 | p984
1970 | aS'rapeseed'
1971 | p985
1972 | aS'daisy'
1973 | p986
1974 | aS"yellow lady's slipper, yellow lady-slipper, Cypripedium calceolus, Cypripedium parviflorum"
1975 | p987
1976 | aS'corn'
1977 | p988
1978 | aS'acorn'
1979 | p989
1980 | aS'hip, rose hip, rosehip'
1981 | p990
1982 | aS'buckeye, horse chestnut, conker'
1983 | p991
1984 | aS'coral fungus'
1985 | p992
1986 | aS'agaric'
1987 | p993
1988 | aS'gyromitra'
1989 | p994
1990 | aS'stinkhorn, carrion fungus'
1991 | p995
1992 | aS'earthstar'
1993 | p996
1994 | aS'hen-of-the-woods, hen of the woods, Polyporus frondosus, Grifola frondosa'
1995 | p997
1996 | aS'bolete'
1997 | p998
1998 | aS'ear, spike, capitulum'
1999 | p999
2000 | aS'toilet tissue, toilet paper, bathroom tissue'
2001 | p1000
2002 | a.


--------------------------------------------------------------------------------
/Data/classid_to_wnid.pkl:
--------------------------------------------------------------------------------
   1 | (lp0
   2 | S'n01440764'
   3 | p1
   4 | aS'n01443537'
   5 | p2
   6 | aS'n01484850'
   7 | p3
   8 | aS'n01491361'
   9 | p4
  10 | aS'n01494475'
  11 | p5
  12 | aS'n01496331'
  13 | p6
  14 | aS'n01498041'
  15 | p7
  16 | aS'n01514668'
  17 | p8
  18 | aS'n01514859'
  19 | p9
  20 | aS'n01518878'
  21 | p10
  22 | aS'n01530575'
  23 | p11
  24 | aS'n01531178'
  25 | p12
  26 | aS'n01532829'
  27 | p13
  28 | aS'n01534433'
  29 | p14
  30 | aS'n01537544'
  31 | p15
  32 | aS'n01558993'
  33 | p16
  34 | aS'n01560419'
  35 | p17
  36 | aS'n01580077'
  37 | p18
  38 | aS'n01582220'
  39 | p19
  40 | aS'n01592084'
  41 | p20
  42 | aS'n01601694'
  43 | p21
  44 | aS'n01608432'
  45 | p22
  46 | aS'n01614925'
  47 | p23
  48 | aS'n01616318'
  49 | p24
  50 | aS'n01622779'
  51 | p25
  52 | aS'n01629819'
  53 | p26
  54 | aS'n01630670'
  55 | p27
  56 | aS'n01631663'
  57 | p28
  58 | aS'n01632458'
  59 | p29
  60 | aS'n01632777'
  61 | p30
  62 | aS'n01641577'
  63 | p31
  64 | aS'n01644373'
  65 | p32
  66 | aS'n01644900'
  67 | p33
  68 | aS'n01664065'
  69 | p34
  70 | aS'n01665541'
  71 | p35
  72 | aS'n01667114'
  73 | p36
  74 | aS'n01667778'
  75 | p37
  76 | aS'n01669191'
  77 | p38
  78 | aS'n01675722'
  79 | p39
  80 | aS'n01677366'
  81 | p40
  82 | aS'n01682714'
  83 | p41
  84 | aS'n01685808'
  85 | p42
  86 | aS'n01687978'
  87 | p43
  88 | aS'n01688243'
  89 | p44
  90 | aS'n01689811'
  91 | p45
  92 | aS'n01692333'
  93 | p46
  94 | aS'n01693334'
  95 | p47
  96 | aS'n01694178'
  97 | p48
  98 | aS'n01695060'
  99 | p49
 100 | aS'n01697457'
 101 | p50
 102 | aS'n01698640'
 103 | p51
 104 | aS'n01704323'
 105 | p52
 106 | aS'n01728572'
 107 | p53
 108 | aS'n01728920'
 109 | p54
 110 | aS'n01729322'
 111 | p55
 112 | aS'n01729977'
 113 | p56
 114 | aS'n01734418'
 115 | p57
 116 | aS'n01735189'
 117 | p58
 118 | aS'n01737021'
 119 | p59
 120 | aS'n01739381'
 121 | p60
 122 | aS'n01740131'
 123 | p61
 124 | aS'n01742172'
 125 | p62
 126 | aS'n01744401'
 127 | p63
 128 | aS'n01748264'
 129 | p64
 130 | aS'n01749939'
 131 | p65
 132 | aS'n01751748'
 133 | p66
 134 | aS'n01753488'
 135 | p67
 136 | aS'n01755581'
 137 | p68
 138 | aS'n01756291'
 139 | p69
 140 | aS'n01768244'
 141 | p70
 142 | aS'n01770081'
 143 | p71
 144 | aS'n01770393'
 145 | p72
 146 | aS'n01773157'
 147 | p73
 148 | aS'n01773549'
 149 | p74
 150 | aS'n01773797'
 151 | p75
 152 | aS'n01774384'
 153 | p76
 154 | aS'n01774750'
 155 | p77
 156 | aS'n01775062'
 157 | p78
 158 | aS'n01776313'
 159 | p79
 160 | aS'n01784675'
 161 | p80
 162 | aS'n01795545'
 163 | p81
 164 | aS'n01796340'
 165 | p82
 166 | aS'n01797886'
 167 | p83
 168 | aS'n01798484'
 169 | p84
 170 | aS'n01806143'
 171 | p85
 172 | aS'n01806567'
 173 | p86
 174 | aS'n01807496'
 175 | p87
 176 | aS'n01817953'
 177 | p88
 178 | aS'n01818515'
 179 | p89
 180 | aS'n01819313'
 181 | p90
 182 | aS'n01820546'
 183 | p91
 184 | aS'n01824575'
 185 | p92
 186 | aS'n01828970'
 187 | p93
 188 | aS'n01829413'
 189 | p94
 190 | aS'n01833805'
 191 | p95
 192 | aS'n01843065'
 193 | p96
 194 | aS'n01843383'
 195 | p97
 196 | aS'n01847000'
 197 | p98
 198 | aS'n01855032'
 199 | p99
 200 | aS'n01855672'
 201 | p100
 202 | aS'n01860187'
 203 | p101
 204 | aS'n01871265'
 205 | p102
 206 | aS'n01872401'
 207 | p103
 208 | aS'n01873310'
 209 | p104
 210 | aS'n01877812'
 211 | p105
 212 | aS'n01882714'
 213 | p106
 214 | aS'n01883070'
 215 | p107
 216 | aS'n01910747'
 217 | p108
 218 | aS'n01914609'
 219 | p109
 220 | aS'n01917289'
 221 | p110
 222 | aS'n01924916'
 223 | p111
 224 | aS'n01930112'
 225 | p112
 226 | aS'n01943899'
 227 | p113
 228 | aS'n01944390'
 229 | p114
 230 | aS'n01945685'
 231 | p115
 232 | aS'n01950731'
 233 | p116
 234 | aS'n01955084'
 235 | p117
 236 | aS'n01968897'
 237 | p118
 238 | aS'n01978287'
 239 | p119
 240 | aS'n01978455'
 241 | p120
 242 | aS'n01980166'
 243 | p121
 244 | aS'n01981276'
 245 | p122
 246 | aS'n01983481'
 247 | p123
 248 | aS'n01984695'
 249 | p124
 250 | aS'n01985128'
 251 | p125
 252 | aS'n01986214'
 253 | p126
 254 | aS'n01990800'
 255 | p127
 256 | aS'n02002556'
 257 | p128
 258 | aS'n02002724'
 259 | p129
 260 | aS'n02006656'
 261 | p130
 262 | aS'n02007558'
 263 | p131
 264 | aS'n02009229'
 265 | p132
 266 | aS'n02009912'
 267 | p133
 268 | aS'n02011460'
 269 | p134
 270 | aS'n03126707'
 271 | p135
 272 | aS'n02013706'
 273 | p136
 274 | aS'n02017213'
 275 | p137
 276 | aS'n02018207'
 277 | p138
 278 | aS'n02018795'
 279 | p139
 280 | aS'n02025239'
 281 | p140
 282 | aS'n02027492'
 283 | p141
 284 | aS'n02028035'
 285 | p142
 286 | aS'n02033041'
 287 | p143
 288 | aS'n02037110'
 289 | p144
 290 | aS'n02051845'
 291 | p145
 292 | aS'n02056570'
 293 | p146
 294 | aS'n02058221'
 295 | p147
 296 | aS'n02066245'
 297 | p148
 298 | aS'n02071294'
 299 | p149
 300 | aS'n02074367'
 301 | p150
 302 | aS'n02077923'
 303 | p151
 304 | aS'n02085620'
 305 | p152
 306 | aS'n02085782'
 307 | p153
 308 | aS'n02085936'
 309 | p154
 310 | aS'n02086079'
 311 | p155
 312 | aS'n02086240'
 313 | p156
 314 | aS'n02086646'
 315 | p157
 316 | aS'n02086910'
 317 | p158
 318 | aS'n02087046'
 319 | p159
 320 | aS'n02087394'
 321 | p160
 322 | aS'n02088094'
 323 | p161
 324 | aS'n02088238'
 325 | p162
 326 | aS'n02088364'
 327 | p163
 328 | aS'n02088466'
 329 | p164
 330 | aS'n02088632'
 331 | p165
 332 | aS'n02089078'
 333 | p166
 334 | aS'n02089867'
 335 | p167
 336 | aS'n02089973'
 337 | p168
 338 | aS'n02090379'
 339 | p169
 340 | aS'n02090622'
 341 | p170
 342 | aS'n02090721'
 343 | p171
 344 | aS'n02091032'
 345 | p172
 346 | aS'n02091134'
 347 | p173
 348 | aS'n02091244'
 349 | p174
 350 | aS'n02091467'
 351 | p175
 352 | aS'n02091635'
 353 | p176
 354 | aS'n02091831'
 355 | p177
 356 | aS'n02092002'
 357 | p178
 358 | aS'n02092339'
 359 | p179
 360 | aS'n02093256'
 361 | p180
 362 | aS'n02093428'
 363 | p181
 364 | aS'n02093647'
 365 | p182
 366 | aS'n02093754'
 367 | p183
 368 | aS'n02093859'
 369 | p184
 370 | aS'n02093991'
 371 | p185
 372 | aS'n02094114'
 373 | p186
 374 | aS'n02094258'
 375 | p187
 376 | aS'n02094433'
 377 | p188
 378 | aS'n02095314'
 379 | p189
 380 | aS'n02095570'
 381 | p190
 382 | aS'n02095889'
 383 | p191
 384 | aS'n02096051'
 385 | p192
 386 | aS'n02096177'
 387 | p193
 388 | aS'n02096294'
 389 | p194
 390 | aS'n02096437'
 391 | p195
 392 | aS'n02096585'
 393 | p196
 394 | aS'n02097047'
 395 | p197
 396 | aS'n02097130'
 397 | p198
 398 | aS'n02097209'
 399 | p199
 400 | aS'n02097298'
 401 | p200
 402 | aS'n02097474'
 403 | p201
 404 | aS'n02097658'
 405 | p202
 406 | aS'n02098105'
 407 | p203
 408 | aS'n02098286'
 409 | p204
 410 | aS'n02098413'
 411 | p205
 412 | aS'n02099267'
 413 | p206
 414 | aS'n02099429'
 415 | p207
 416 | aS'n02099601'
 417 | p208
 418 | aS'n02099712'
 419 | p209
 420 | aS'n02099849'
 421 | p210
 422 | aS'n02100236'
 423 | p211
 424 | aS'n02100583'
 425 | p212
 426 | aS'n02100735'
 427 | p213
 428 | aS'n02100877'
 429 | p214
 430 | aS'n02101006'
 431 | p215
 432 | aS'n02101388'
 433 | p216
 434 | aS'n02101556'
 435 | p217
 436 | aS'n02102040'
 437 | p218
 438 | aS'n02102177'
 439 | p219
 440 | aS'n02102318'
 441 | p220
 442 | aS'n02102480'
 443 | p221
 444 | aS'n02102973'
 445 | p222
 446 | aS'n02104029'
 447 | p223
 448 | aS'n02104365'
 449 | p224
 450 | aS'n02105056'
 451 | p225
 452 | aS'n02105162'
 453 | p226
 454 | aS'n02105251'
 455 | p227
 456 | aS'n02105412'
 457 | p228
 458 | aS'n02105505'
 459 | p229
 460 | aS'n02105641'
 461 | p230
 462 | aS'n02105855'
 463 | p231
 464 | aS'n02106030'
 465 | p232
 466 | aS'n02106166'
 467 | p233
 468 | aS'n02106382'
 469 | p234
 470 | aS'n02106550'
 471 | p235
 472 | aS'n02106662'
 473 | p236
 474 | aS'n02107142'
 475 | p237
 476 | aS'n02107312'
 477 | p238
 478 | aS'n02107574'
 479 | p239
 480 | aS'n02107683'
 481 | p240
 482 | aS'n02107908'
 483 | p241
 484 | aS'n02108000'
 485 | p242
 486 | aS'n02108089'
 487 | p243
 488 | aS'n02108422'
 489 | p244
 490 | aS'n02108551'
 491 | p245
 492 | aS'n02108915'
 493 | p246
 494 | aS'n02109047'
 495 | p247
 496 | aS'n02109525'
 497 | p248
 498 | aS'n02109961'
 499 | p249
 500 | aS'n02110063'
 501 | p250
 502 | aS'n02110185'
 503 | p251
 504 | aS'n02110341'
 505 | p252
 506 | aS'n02110627'
 507 | p253
 508 | aS'n02110806'
 509 | p254
 510 | aS'n02110958'
 511 | p255
 512 | aS'n02111129'
 513 | p256
 514 | aS'n02111277'
 515 | p257
 516 | aS'n02111500'
 517 | p258
 518 | aS'n02111889'
 519 | p259
 520 | aS'n02112018'
 521 | p260
 522 | aS'n02112137'
 523 | p261
 524 | aS'n02112350'
 525 | p262
 526 | aS'n02112706'
 527 | p263
 528 | aS'n02113023'
 529 | p264
 530 | aS'n02113186'
 531 | p265
 532 | aS'n02113624'
 533 | p266
 534 | aS'n02113712'
 535 | p267
 536 | aS'n02113799'
 537 | p268
 538 | aS'n02113978'
 539 | p269
 540 | aS'n02114367'
 541 | p270
 542 | aS'n02114548'
 543 | p271
 544 | aS'n02114712'
 545 | p272
 546 | aS'n02114855'
 547 | p273
 548 | aS'n02115641'
 549 | p274
 550 | aS'n02115913'
 551 | p275
 552 | aS'n02116738'
 553 | p276
 554 | aS'n02117135'
 555 | p277
 556 | aS'n02119022'
 557 | p278
 558 | aS'n02119789'
 559 | p279
 560 | aS'n02120079'
 561 | p280
 562 | aS'n02120505'
 563 | p281
 564 | aS'n02123045'
 565 | p282
 566 | aS'n02123159'
 567 | p283
 568 | aS'n02123394'
 569 | p284
 570 | aS'n02123597'
 571 | p285
 572 | aS'n02124075'
 573 | p286
 574 | aS'n02125311'
 575 | p287
 576 | aS'n02127052'
 577 | p288
 578 | aS'n02128385'
 579 | p289
 580 | aS'n02128757'
 581 | p290
 582 | aS'n02128925'
 583 | p291
 584 | aS'n02129165'
 585 | p292
 586 | aS'n02129604'
 587 | p293
 588 | aS'n02130308'
 589 | p294
 590 | aS'n02132136'
 591 | p295
 592 | aS'n02133161'
 593 | p296
 594 | aS'n02134084'
 595 | p297
 596 | aS'n02134418'
 597 | p298
 598 | aS'n02137549'
 599 | p299
 600 | aS'n02138441'
 601 | p300
 602 | aS'n02165105'
 603 | p301
 604 | aS'n02165456'
 605 | p302
 606 | aS'n02167151'
 607 | p303
 608 | aS'n02168699'
 609 | p304
 610 | aS'n02169497'
 611 | p305
 612 | aS'n02172182'
 613 | p306
 614 | aS'n02174001'
 615 | p307
 616 | aS'n02177972'
 617 | p308
 618 | aS'n02190166'
 619 | p309
 620 | aS'n02206856'
 621 | p310
 622 | aS'n02219486'
 623 | p311
 624 | aS'n02226429'
 625 | p312
 626 | aS'n02229544'
 627 | p313
 628 | aS'n02231487'
 629 | p314
 630 | aS'n02233338'
 631 | p315
 632 | aS'n02236044'
 633 | p316
 634 | aS'n02256656'
 635 | p317
 636 | aS'n02259212'
 637 | p318
 638 | aS'n02264363'
 639 | p319
 640 | aS'n02268443'
 641 | p320
 642 | aS'n02268853'
 643 | p321
 644 | aS'n02276258'
 645 | p322
 646 | aS'n02277742'
 647 | p323
 648 | aS'n02279972'
 649 | p324
 650 | aS'n02280649'
 651 | p325
 652 | aS'n02281406'
 653 | p326
 654 | aS'n02281787'
 655 | p327
 656 | aS'n02317335'
 657 | p328
 658 | aS'n02319095'
 659 | p329
 660 | aS'n02321529'
 661 | p330
 662 | aS'n02325366'
 663 | p331
 664 | aS'n02326432'
 665 | p332
 666 | aS'n02328150'
 667 | p333
 668 | aS'n02342885'
 669 | p334
 670 | aS'n02346627'
 671 | p335
 672 | aS'n02356798'
 673 | p336
 674 | aS'n02361337'
 675 | p337
 676 | aS'n02363005'
 677 | p338
 678 | aS'n02364673'
 679 | p339
 680 | aS'n02389026'
 681 | p340
 682 | aS'n02391049'
 683 | p341
 684 | aS'n02395406'
 685 | p342
 686 | aS'n02396427'
 687 | p343
 688 | aS'n02397096'
 689 | p344
 690 | aS'n02398521'
 691 | p345
 692 | aS'n02403003'
 693 | p346
 694 | aS'n02408429'
 695 | p347
 696 | aS'n02410509'
 697 | p348
 698 | aS'n02412080'
 699 | p349
 700 | aS'n02415577'
 701 | p350
 702 | aS'n02417914'
 703 | p351
 704 | aS'n02422106'
 705 | p352
 706 | aS'n02422699'
 707 | p353
 708 | aS'n02423022'
 709 | p354
 710 | aS'n02437312'
 711 | p355
 712 | aS'n02437616'
 713 | p356
 714 | aS'n02441942'
 715 | p357
 716 | aS'n02442845'
 717 | p358
 718 | aS'n02443114'
 719 | p359
 720 | aS'n02443484'
 721 | p360
 722 | aS'n02444819'
 723 | p361
 724 | aS'n02445715'
 725 | p362
 726 | aS'n02447366'
 727 | p363
 728 | aS'n02454379'
 729 | p364
 730 | aS'n02457408'
 731 | p365
 732 | aS'n02480495'
 733 | p366
 734 | aS'n02480855'
 735 | p367
 736 | aS'n02481823'
 737 | p368
 738 | aS'n02483362'
 739 | p369
 740 | aS'n02483708'
 741 | p370
 742 | aS'n02484975'
 743 | p371
 744 | aS'n02486261'
 745 | p372
 746 | aS'n02486410'
 747 | p373
 748 | aS'n02487347'
 749 | p374
 750 | aS'n02488291'
 751 | p375
 752 | aS'n02488702'
 753 | p376
 754 | aS'n02489166'
 755 | p377
 756 | aS'n02490219'
 757 | p378
 758 | aS'n02492035'
 759 | p379
 760 | aS'n02492660'
 761 | p380
 762 | aS'n02493509'
 763 | p381
 764 | aS'n02493793'
 765 | p382
 766 | aS'n02494079'
 767 | p383
 768 | aS'n02497673'
 769 | p384
 770 | aS'n02500267'
 771 | p385
 772 | aS'n02504013'
 773 | p386
 774 | aS'n02504458'
 775 | p387
 776 | aS'n02509815'
 777 | p388
 778 | aS'n02510455'
 779 | p389
 780 | aS'n02514041'
 781 | p390
 782 | aS'n02526121'
 783 | p391
 784 | aS'n02536864'
 785 | p392
 786 | aS'n02606052'
 787 | p393
 788 | aS'n02607072'
 789 | p394
 790 | aS'n02640242'
 791 | p395
 792 | aS'n02641379'
 793 | p396
 794 | aS'n02643566'
 795 | p397
 796 | aS'n02655020'
 797 | p398
 798 | aS'n02666196'
 799 | p399
 800 | aS'n02667093'
 801 | p400
 802 | aS'n02669723'
 803 | p401
 804 | aS'n02672831'
 805 | p402
 806 | aS'n02676566'
 807 | p403
 808 | aS'n02687172'
 809 | p404
 810 | aS'n02690373'
 811 | p405
 812 | aS'n02692877'
 813 | p406
 814 | aS'n02699494'
 815 | p407
 816 | aS'n02701002'
 817 | p408
 818 | aS'n02704792'
 819 | p409
 820 | aS'n02708093'
 821 | p410
 822 | aS'n02727426'
 823 | p411
 824 | aS'n02730930'
 825 | p412
 826 | aS'n02747177'
 827 | p413
 828 | aS'n02749479'
 829 | p414
 830 | aS'n02769748'
 831 | p415
 832 | aS'n02776631'
 833 | p416
 834 | aS'n02777292'
 835 | p417
 836 | aS'n02782093'
 837 | p418
 838 | aS'n02783161'
 839 | p419
 840 | aS'n02786058'
 841 | p420
 842 | aS'n02787622'
 843 | p421
 844 | aS'n02788148'
 845 | p422
 846 | aS'n02790996'
 847 | p423
 848 | aS'n02791124'
 849 | p424
 850 | aS'n02791270'
 851 | p425
 852 | aS'n02793495'
 853 | p426
 854 | aS'n02794156'
 855 | p427
 856 | aS'n02795169'
 857 | p428
 858 | aS'n02797295'
 859 | p429
 860 | aS'n02799071'
 861 | p430
 862 | aS'n02802426'
 863 | p431
 864 | aS'n02804414'
 865 | p432
 866 | aS'n02804610'
 867 | p433
 868 | aS'n02807133'
 869 | p434
 870 | aS'n02808304'
 871 | p435
 872 | aS'n02808440'
 873 | p436
 874 | aS'n02814533'
 875 | p437
 876 | aS'n02814860'
 877 | p438
 878 | aS'n02815834'
 879 | p439
 880 | aS'n02817516'
 881 | p440
 882 | aS'n02823428'
 883 | p441
 884 | aS'n02823750'
 885 | p442
 886 | aS'n02825657'
 887 | p443
 888 | aS'n02834397'
 889 | p444
 890 | aS'n02835271'
 891 | p445
 892 | aS'n02837789'
 893 | p446
 894 | aS'n02840245'
 895 | p447
 896 | aS'n02841315'
 897 | p448
 898 | aS'n02843684'
 899 | p449
 900 | aS'n02859443'
 901 | p450
 902 | aS'n02860847'
 903 | p451
 904 | aS'n02865351'
 905 | p452
 906 | aS'n02869837'
 907 | p453
 908 | aS'n02870880'
 909 | p454
 910 | aS'n02871525'
 911 | p455
 912 | aS'n02877765'
 913 | p456
 914 | aS'n02879718'
 915 | p457
 916 | aS'n02883205'
 917 | p458
 918 | aS'n02892201'
 919 | p459
 920 | aS'n02892767'
 921 | p460
 922 | aS'n02894605'
 923 | p461
 924 | aS'n02895154'
 925 | p462
 926 | aS'n02906734'
 927 | p463
 928 | aS'n02909870'
 929 | p464
 930 | aS'n02910353'
 931 | p465
 932 | aS'n02916936'
 933 | p466
 934 | aS'n02917067'
 935 | p467
 936 | aS'n02927161'
 937 | p468
 938 | aS'n02930766'
 939 | p469
 940 | aS'n02939185'
 941 | p470
 942 | aS'n02948072'
 943 | p471
 944 | aS'n02950826'
 945 | p472
 946 | aS'n02951358'
 947 | p473
 948 | aS'n02951585'
 949 | p474
 950 | aS'n02963159'
 951 | p475
 952 | aS'n02965783'
 953 | p476
 954 | aS'n02966193'
 955 | p477
 956 | aS'n02966687'
 957 | p478
 958 | aS'n02971356'
 959 | p479
 960 | aS'n02974003'
 961 | p480
 962 | aS'n02977058'
 963 | p481
 964 | aS'n02978881'
 965 | p482
 966 | aS'n02979186'
 967 | p483
 968 | aS'n02980441'
 969 | p484
 970 | aS'n02981792'
 971 | p485
 972 | aS'n02988304'
 973 | p486
 974 | aS'n02992211'
 975 | p487
 976 | aS'n02992529'
 977 | p488
 978 | aS'n02999410'
 979 | p489
 980 | aS'n03000134'
 981 | p490
 982 | aS'n03000247'
 983 | p491
 984 | aS'n03000684'
 985 | p492
 986 | aS'n03014705'
 987 | p493
 988 | aS'n03016953'
 989 | p494
 990 | aS'n03017168'
 991 | p495
 992 | aS'n03018349'
 993 | p496
 994 | aS'n03026506'
 995 | p497
 996 | aS'n03028079'
 997 | p498
 998 | aS'n03032252'
 999 | p499
1000 | aS'n03041632'
1001 | p500
1002 | aS'n03042490'
1003 | p501
1004 | aS'n03045698'
1005 | p502
1006 | aS'n03047690'
1007 | p503
1008 | aS'n03062245'
1009 | p504
1010 | aS'n03063599'
1011 | p505
1012 | aS'n03063689'
1013 | p506
1014 | aS'n03065424'
1015 | p507
1016 | aS'n03075370'
1017 | p508
1018 | aS'n03085013'
1019 | p509
1020 | aS'n03089624'
1021 | p510
1022 | aS'n03095699'
1023 | p511
1024 | aS'n03100240'
1025 | p512
1026 | aS'n03109150'
1027 | p513
1028 | aS'n03110669'
1029 | p514
1030 | aS'n03124043'
1031 | p515
1032 | aS'n03124170'
1033 | p516
1034 | aS'n03125729'
1035 | p517
1036 | ag135
1037 | aS'n03127747'
1038 | p518
1039 | aS'n03127925'
1040 | p519
1041 | aS'n03131574'
1042 | p520
1043 | aS'n03133878'
1044 | p521
1045 | aS'n03134739'
1046 | p522
1047 | aS'n03141823'
1048 | p523
1049 | aS'n03146219'
1050 | p524
1051 | aS'n03160309'
1052 | p525
1053 | aS'n03179701'
1054 | p526
1055 | aS'n03180011'
1056 | p527
1057 | aS'n03187595'
1058 | p528
1059 | aS'n03188531'
1060 | p529
1061 | aS'n03196217'
1062 | p530
1063 | aS'n03197337'
1064 | p531
1065 | aS'n03201208'
1066 | p532
1067 | aS'n03207743'
1068 | p533
1069 | aS'n03207941'
1070 | p534
1071 | aS'n03208938'
1072 | p535
1073 | aS'n03216828'
1074 | p536
1075 | aS'n03218198'
1076 | p537
1077 | aS'n03220513'
1078 | p538
1079 | aS'n03223299'
1080 | p539
1081 | aS'n03240683'
1082 | p540
1083 | aS'n03249569'
1084 | p541
1085 | aS'n03250847'
1086 | p542
1087 | aS'n03255030'
1088 | p543
1089 | aS'n03259280'
1090 | p544
1091 | aS'n03271574'
1092 | p545
1093 | aS'n03272010'
1094 | p546
1095 | aS'n03272562'
1096 | p547
1097 | aS'n03290653'
1098 | p548
1099 | aS'n03291819'
1100 | p549
1101 | aS'n03297495'
1102 | p550
1103 | aS'n03314780'
1104 | p551
1105 | aS'n03325584'
1106 | p552
1107 | aS'n03337140'
1108 | p553
1109 | aS'n03344393'
1110 | p554
1111 | aS'n03345487'
1112 | p555
1113 | aS'n03347037'
1114 | p556
1115 | aS'n03355925'
1116 | p557
1117 | aS'n03372029'
1118 | p558
1119 | aS'n03376595'
1120 | p559
1121 | aS'n03379051'
1122 | p560
1123 | aS'n03384352'
1124 | p561
1125 | aS'n03388043'
1126 | p562
1127 | aS'n03388183'
1128 | p563
1129 | aS'n03388549'
1130 | p564
1131 | aS'n03393912'
1132 | p565
1133 | aS'n03394916'
1134 | p566
1135 | aS'n03400231'
1136 | p567
1137 | aS'n03404251'
1138 | p568
1139 | aS'n03417042'
1140 | p569
1141 | aS'n03424325'
1142 | p570
1143 | aS'n03425413'
1144 | p571
1145 | aS'n03443371'
1146 | p572
1147 | aS'n03444034'
1148 | p573
1149 | aS'n03445777'
1150 | p574
1151 | aS'n03445924'
1152 | p575
1153 | aS'n03447447'
1154 | p576
1155 | aS'n03447721'
1156 | p577
1157 | aS'n03450230'
1158 | p578
1159 | aS'n03452741'
1160 | p579
1161 | aS'n03457902'
1162 | p580
1163 | aS'n03459775'
1164 | p581
1165 | aS'n03461385'
1166 | p582
1167 | aS'n03467068'
1168 | p583
1169 | aS'n03476684'
1170 | p584
1171 | aS'n03476991'
1172 | p585
1173 | aS'n03478589'
1174 | p586
1175 | aS'n03481172'
1176 | p587
1177 | aS'n03482405'
1178 | p588
1179 | aS'n03483316'
1180 | p589
1181 | aS'n03485407'
1182 | p590
1183 | aS'n03485794'
1184 | p591
1185 | aS'n03492542'
1186 | p592
1187 | aS'n03494278'
1188 | p593
1189 | aS'n03495258'
1190 | p594
1191 | aS'n03496892'
1192 | p595
1193 | aS'n03498962'
1194 | p596
1195 | aS'n03527444'
1196 | p597
1197 | aS'n03529860'
1198 | p598
1199 | aS'n03530642'
1200 | p599
1201 | aS'n03532672'
1202 | p600
1203 | aS'n03534580'
1204 | p601
1205 | aS'n03535780'
1206 | p602
1207 | aS'n03538406'
1208 | p603
1209 | aS'n03544143'
1210 | p604
1211 | aS'n03584254'
1212 | p605
1213 | aS'n03584829'
1214 | p606
1215 | aS'n03590841'
1216 | p607
1217 | aS'n03594734'
1218 | p608
1219 | aS'n03594945'
1220 | p609
1221 | aS'n03595614'
1222 | p610
1223 | aS'n03598930'
1224 | p611
1225 | aS'n03599486'
1226 | p612
1227 | aS'n03602883'
1228 | p613
1229 | aS'n03617480'
1230 | p614
1231 | aS'n03623198'
1232 | p615
1233 | aS'n03627232'
1234 | p616
1235 | aS'n03630383'
1236 | p617
1237 | aS'n03633091'
1238 | p618
1239 | aS'n03637318'
1240 | p619
1241 | aS'n03642806'
1242 | p620
1243 | aS'n03649909'
1244 | p621
1245 | aS'n03657121'
1246 | p622
1247 | aS'n03658185'
1248 | p623
1249 | aS'n03661043'
1250 | p624
1251 | aS'n03662601'
1252 | p625
1253 | aS'n03666591'
1254 | p626
1255 | aS'n03670208'
1256 | p627
1257 | aS'n03673027'
1258 | p628
1259 | aS'n03676483'
1260 | p629
1261 | aS'n03680355'
1262 | p630
1263 | aS'n03690938'
1264 | p631
1265 | aS'n03691459'
1266 | p632
1267 | aS'n03692522'
1268 | p633
1269 | aS'n03697007'
1270 | p634
1271 | aS'n03706229'
1272 | p635
1273 | aS'n03709823'
1274 | p636
1275 | aS'n03710193'
1276 | p637
1277 | aS'n03710637'
1278 | p638
1279 | aS'n03710721'
1280 | p639
1281 | aS'n03717622'
1282 | p640
1283 | aS'n03720891'
1284 | p641
1285 | aS'n03721384'
1286 | p642
1287 | aS'n03724870'
1288 | p643
1289 | aS'n03729826'
1290 | p644
1291 | aS'n03733131'
1292 | p645
1293 | aS'n03733281'
1294 | p646
1295 | aS'n03733805'
1296 | p647
1297 | aS'n03742115'
1298 | p648
1299 | aS'n03743016'
1300 | p649
1301 | aS'n03759954'
1302 | p650
1303 | aS'n03761084'
1304 | p651
1305 | aS'n03763968'
1306 | p652
1307 | aS'n03764736'
1308 | p653
1309 | aS'n03769881'
1310 | p654
1311 | aS'n03770439'
1312 | p655
1313 | aS'n03770679'
1314 | p656
1315 | aS'n03773504'
1316 | p657
1317 | aS'n03775071'
1318 | p658
1319 | aS'n03775546'
1320 | p659
1321 | aS'n03776460'
1322 | p660
1323 | aS'n03777568'
1324 | p661
1325 | aS'n03777754'
1326 | p662
1327 | aS'n03781244'
1328 | p663
1329 | aS'n03782006'
1330 | p664
1331 | aS'n03785016'
1332 | p665
1333 | aS'n03786901'
1334 | p666
1335 | aS'n03787032'
1336 | p667
1337 | aS'n03788195'
1338 | p668
1339 | aS'n03788365'
1340 | p669
1341 | aS'n03791053'
1342 | p670
1343 | aS'n03792782'
1344 | p671
1345 | aS'n03792972'
1346 | p672
1347 | aS'n03793489'
1348 | p673
1349 | aS'n03794056'
1350 | p674
1351 | aS'n03796401'
1352 | p675
1353 | aS'n03803284'
1354 | p676
1355 | aS'n03804744'
1356 | p677
1357 | aS'n03814639'
1358 | p678
1359 | aS'n03814906'
1360 | p679
1361 | aS'n03825788'
1362 | p680
1363 | aS'n03832673'
1364 | p681
1365 | aS'n03837869'
1366 | p682
1367 | aS'n03838899'
1368 | p683
1369 | aS'n03840681'
1370 | p684
1371 | aS'n03841143'
1372 | p685
1373 | aS'n03843555'
1374 | p686
1375 | aS'n03854065'
1376 | p687
1377 | aS'n03857828'
1378 | p688
1379 | aS'n03866082'
1380 | p689
1381 | aS'n03868242'
1382 | p690
1383 | aS'n03868863'
1384 | p691
1385 | aS'n03871628'
1386 | p692
1387 | aS'n03873416'
1388 | p693
1389 | aS'n03874293'
1390 | p694
1391 | aS'n03874599'
1392 | p695
1393 | aS'n03876231'
1394 | p696
1395 | aS'n03877472'
1396 | p697
1397 | aS'n03877845'
1398 | p698
1399 | aS'n03884397'
1400 | p699
1401 | aS'n03887697'
1402 | p700
1403 | aS'n03888257'
1404 | p701
1405 | aS'n03888605'
1406 | p702
1407 | aS'n03891251'
1408 | p703
1409 | aS'n03891332'
1410 | p704
1411 | aS'n03895866'
1412 | p705
1413 | aS'n03899768'
1414 | p706
1415 | aS'n03902125'
1416 | p707
1417 | aS'n03903868'
1418 | p708
1419 | aS'n03908618'
1420 | p709
1421 | aS'n03908714'
1422 | p710
1423 | aS'n03916031'
1424 | p711
1425 | aS'n03920288'
1426 | p712
1427 | aS'n03924679'
1428 | p713
1429 | aS'n03929660'
1430 | p714
1431 | aS'n03929855'
1432 | p715
1433 | aS'n03930313'
1434 | p716
1435 | aS'n03930630'
1436 | p717
1437 | aS'n03933933'
1438 | p718
1439 | aS'n03935335'
1440 | p719
1441 | aS'n03937543'
1442 | p720
1443 | aS'n03938244'
1444 | p721
1445 | aS'n03942813'
1446 | p722
1447 | aS'n03944341'
1448 | p723
1449 | aS'n03947888'
1450 | p724
1451 | aS'n03950228'
1452 | p725
1453 | aS'n03954731'
1454 | p726
1455 | aS'n03956157'
1456 | p727
1457 | aS'n03958227'
1458 | p728
1459 | aS'n03961711'
1460 | p729
1461 | aS'n03967562'
1462 | p730
1463 | aS'n03970156'
1464 | p731
1465 | aS'n03976467'
1466 | p732
1467 | aS'n03976657'
1468 | p733
1469 | aS'n03977966'
1470 | p734
1471 | aS'n03980874'
1472 | p735
1473 | aS'n03982430'
1474 | p736
1475 | aS'n03983396'
1476 | p737
1477 | aS'n03991062'
1478 | p738
1479 | aS'n03992509'
1480 | p739
1481 | aS'n03995372'
1482 | p740
1483 | aS'n03998194'
1484 | p741
1485 | aS'n04004767'
1486 | p742
1487 | aS'n04005630'
1488 | p743
1489 | aS'n04008634'
1490 | p744
1491 | aS'n04009552'
1492 | p745
1493 | aS'n04019541'
1494 | p746
1495 | aS'n04023962'
1496 | p747
1497 | aS'n04026417'
1498 | p748
1499 | aS'n04033901'
1500 | p749
1501 | aS'n04033995'
1502 | p750
1503 | aS'n04037443'
1504 | p751
1505 | aS'n04039381'
1506 | p752
1507 | aS'n04040759'
1508 | p753
1509 | aS'n04041544'
1510 | p754
1511 | aS'n04044716'
1512 | p755
1513 | aS'n04049303'
1514 | p756
1515 | aS'n04065272'
1516 | p757
1517 | aS'n04067472'
1518 | p758
1519 | aS'n04069434'
1520 | p759
1521 | aS'n04070727'
1522 | p760
1523 | aS'n04074963'
1524 | p761
1525 | aS'n04081281'
1526 | p762
1527 | aS'n04086273'
1528 | p763
1529 | aS'n04090263'
1530 | p764
1531 | aS'n04099969'
1532 | p765
1533 | aS'n04111531'
1534 | p766
1535 | aS'n04116512'
1536 | p767
1537 | aS'n04118538'
1538 | p768
1539 | aS'n04118776'
1540 | p769
1541 | aS'n04120489'
1542 | p770
1543 | aS'n04125021'
1544 | p771
1545 | aS'n04127249'
1546 | p772
1547 | aS'n04131690'
1548 | p773
1549 | aS'n04133789'
1550 | p774
1551 | aS'n04136333'
1552 | p775
1553 | aS'n04141076'
1554 | p776
1555 | aS'n04141327'
1556 | p777
1557 | aS'n04141975'
1558 | p778
1559 | aS'n04146614'
1560 | p779
1561 | aS'n04147183'
1562 | p780
1563 | aS'n04149813'
1564 | p781
1565 | aS'n04152593'
1566 | p782
1567 | aS'n04153751'
1568 | p783
1569 | aS'n04154565'
1570 | p784
1571 | aS'n04162706'
1572 | p785
1573 | aS'n04179913'
1574 | p786
1575 | aS'n04192698'
1576 | p787
1577 | aS'n04200800'
1578 | p788
1579 | aS'n04201297'
1580 | p789
1581 | aS'n04204238'
1582 | p790
1583 | aS'n04204347'
1584 | p791
1585 | aS'n04208210'
1586 | p792
1587 | aS'n04209133'
1588 | p793
1589 | aS'n04209239'
1590 | p794
1591 | aS'n04228054'
1592 | p795
1593 | aS'n04229816'
1594 | p796
1595 | aS'n04235860'
1596 | p797
1597 | aS'n04238763'
1598 | p798
1599 | aS'n04239074'
1600 | p799
1601 | aS'n04243546'
1602 | p800
1603 | aS'n04251144'
1604 | p801
1605 | aS'n04252077'
1606 | p802
1607 | aS'n04252225'
1608 | p803
1609 | aS'n04254120'
1610 | p804
1611 | aS'n04254680'
1612 | p805
1613 | aS'n04254777'
1614 | p806
1615 | aS'n04258138'
1616 | p807
1617 | aS'n04259630'
1618 | p808
1619 | aS'n04263257'
1620 | p809
1621 | aS'n04264628'
1622 | p810
1623 | aS'n04265275'
1624 | p811
1625 | aS'n04266014'
1626 | p812
1627 | aS'n04270147'
1628 | p813
1629 | aS'n04273569'
1630 | p814
1631 | aS'n04275548'
1632 | p815
1633 | aS'n04277352'
1634 | p816
1635 | aS'n04285008'
1636 | p817
1637 | aS'n04286575'
1638 | p818
1639 | aS'n04296562'
1640 | p819
1641 | aS'n04310018'
1642 | p820
1643 | aS'n04311004'
1644 | p821
1645 | aS'n04311174'
1646 | p822
1647 | aS'n04317175'
1648 | p823
1649 | aS'n04325704'
1650 | p824
1651 | aS'n04326547'
1652 | p825
1653 | aS'n04328186'
1654 | p826
1655 | aS'n04330267'
1656 | p827
1657 | aS'n04332243'
1658 | p828
1659 | aS'n04335435'
1660 | p829
1661 | aS'n04336792'
1662 | p830
1663 | aS'n04344873'
1664 | p831
1665 | aS'n04346328'
1666 | p832
1667 | aS'n04347754'
1668 | p833
1669 | aS'n04350905'
1670 | p834
1671 | aS'n04355338'
1672 | p835
1673 | aS'n04355933'
1674 | p836
1675 | aS'n04356056'
1676 | p837
1677 | aS'n04357314'
1678 | p838
1679 | aS'n04366367'
1680 | p839
1681 | aS'n04367480'
1682 | p840
1683 | aS'n04370456'
1684 | p841
1685 | aS'n04371430'
1686 | p842
1687 | aS'n04371774'
1688 | p843
1689 | aS'n04372370'
1690 | p844
1691 | aS'n04376876'
1692 | p845
1693 | aS'n04380533'
1694 | p846
1695 | aS'n04389033'
1696 | p847
1697 | aS'n04392985'
1698 | p848
1699 | aS'n04398044'
1700 | p849
1701 | aS'n04399382'
1702 | p850
1703 | aS'n04404412'
1704 | p851
1705 | aS'n04409515'
1706 | p852
1707 | aS'n04417672'
1708 | p853
1709 | aS'n04418357'
1710 | p854
1711 | aS'n04423845'
1712 | p855
1713 | aS'n04428191'
1714 | p856
1715 | aS'n04429376'
1716 | p857
1717 | aS'n04435653'
1718 | p858
1719 | aS'n04442312'
1720 | p859
1721 | aS'n04443257'
1722 | p860
1723 | aS'n04447861'
1724 | p861
1725 | aS'n04456115'
1726 | p862
1727 | aS'n04458633'
1728 | p863
1729 | aS'n04461696'
1730 | p864
1731 | aS'n04462240'
1732 | p865
1733 | aS'n04465501'
1734 | p866
1735 | aS'n04467665'
1736 | p867
1737 | aS'n04476259'
1738 | p868
1739 | aS'n04479046'
1740 | p869
1741 | aS'n04482393'
1742 | p870
1743 | aS'n04483307'
1744 | p871
1745 | aS'n04485082'
1746 | p872
1747 | aS'n04486054'
1748 | p873
1749 | aS'n04487081'
1750 | p874
1751 | aS'n04487394'
1752 | p875
1753 | aS'n04493381'
1754 | p876
1755 | aS'n04501370'
1756 | p877
1757 | aS'n04505470'
1758 | p878
1759 | aS'n04507155'
1760 | p879
1761 | aS'n04509417'
1762 | p880
1763 | aS'n04515003'
1764 | p881
1765 | aS'n04517823'
1766 | p882
1767 | aS'n04522168'
1768 | p883
1769 | aS'n04523525'
1770 | p884
1771 | aS'n04525038'
1772 | p885
1773 | aS'n04525305'
1774 | p886
1775 | aS'n04532106'
1776 | p887
1777 | aS'n04532670'
1778 | p888
1779 | aS'n04536866'
1780 | p889
1781 | aS'n04540053'
1782 | p890
1783 | aS'n04542943'
1784 | p891
1785 | aS'n04548280'
1786 | p892
1787 | aS'n04548362'
1788 | p893
1789 | aS'n04550184'
1790 | p894
1791 | aS'n04552348'
1792 | p895
1793 | aS'n04553703'
1794 | p896
1795 | aS'n04554684'
1796 | p897
1797 | aS'n04557648'
1798 | p898
1799 | aS'n04560804'
1800 | p899
1801 | aS'n04562935'
1802 | p900
1803 | aS'n04579145'
1804 | p901
1805 | aS'n04579432'
1806 | p902
1807 | aS'n04584207'
1808 | p903
1809 | aS'n04589890'
1810 | p904
1811 | aS'n04590129'
1812 | p905
1813 | aS'n04591157'
1814 | p906
1815 | aS'n04591713'
1816 | p907
1817 | aS'n04592741'
1818 | p908
1819 | aS'n04596742'
1820 | p909
1821 | aS'n04597913'
1822 | p910
1823 | aS'n04599235'
1824 | p911
1825 | aS'n04604644'
1826 | p912
1827 | aS'n04606251'
1828 | p913
1829 | aS'n04612504'
1830 | p914
1831 | aS'n04613696'
1832 | p915
1833 | aS'n06359193'
1834 | p916
1835 | aS'n06596364'
1836 | p917
1837 | aS'n06785654'
1838 | p918
1839 | aS'n06794110'
1840 | p919
1841 | aS'n06874185'
1842 | p920
1843 | aS'n07248320'
1844 | p921
1845 | aS'n07565083'
1846 | p922
1847 | aS'n07579787'
1848 | p923
1849 | aS'n07583066'
1850 | p924
1851 | aS'n07584110'
1852 | p925
1853 | aS'n07590611'
1854 | p926
1855 | aS'n07613480'
1856 | p927
1857 | aS'n07614500'
1858 | p928
1859 | aS'n07615774'
1860 | p929
1861 | aS'n07684084'
1862 | p930
1863 | aS'n07693725'
1864 | p931
1865 | aS'n07695742'
1866 | p932
1867 | aS'n07697313'
1868 | p933
1869 | aS'n07697537'
1870 | p934
1871 | aS'n07711569'
1872 | p935
1873 | aS'n07714571'
1874 | p936
1875 | aS'n07714990'
1876 | p937
1877 | aS'n07715103'
1878 | p938
1879 | aS'n07716358'
1880 | p939
1881 | aS'n07716906'
1882 | p940
1883 | aS'n07717410'
1884 | p941
1885 | aS'n07717556'
1886 | p942
1887 | aS'n07718472'
1888 | p943
1889 | aS'n07718747'
1890 | p944
1891 | aS'n07720875'
1892 | p945
1893 | aS'n07730033'
1894 | p946
1895 | aS'n07734744'
1896 | p947
1897 | aS'n07742313'
1898 | p948
1899 | aS'n07745940'
1900 | p949
1901 | aS'n07747607'
1902 | p950
1903 | aS'n07749582'
1904 | p951
1905 | aS'n07753113'
1906 | p952
1907 | aS'n07753275'
1908 | p953
1909 | aS'n07753592'
1910 | p954
1911 | aS'n07754684'
1912 | p955
1913 | aS'n07760859'
1914 | p956
1915 | aS'n07768694'
1916 | p957
1917 | aS'n07802026'
1918 | p958
1919 | aS'n07831146'
1920 | p959
1921 | aS'n07836838'
1922 | p960
1923 | aS'n07860988'
1924 | p961
1925 | aS'n07871810'
1926 | p962
1927 | aS'n07873807'
1928 | p963
1929 | aS'n07875152'
1930 | p964
1931 | aS'n07880968'
1932 | p965
1933 | aS'n07892512'
1934 | p966
1935 | aS'n07920052'
1936 | p967
1937 | aS'n07930864'
1938 | p968
1939 | aS'n07932039'
1940 | p969
1941 | aS'n09193705'
1942 | p970
1943 | aS'n09229709'
1944 | p971
1945 | aS'n09246464'
1946 | p972
1947 | aS'n09256479'
1948 | p973
1949 | aS'n09288635'
1950 | p974
1951 | aS'n09332890'
1952 | p975
1953 | aS'n09399592'
1954 | p976
1955 | aS'n09421951'
1956 | p977
1957 | aS'n09428293'
1958 | p978
1959 | aS'n09468604'
1960 | p979
1961 | aS'n09472597'
1962 | p980
1963 | aS'n09835506'
1964 | p981
1965 | aS'n10148035'
1966 | p982
1967 | aS'n10565667'
1968 | p983
1969 | aS'n11879895'
1970 | p984
1971 | aS'n11939491'
1972 | p985
1973 | aS'n12057211'
1974 | p986
1975 | aS'n12144580'
1976 | p987
1977 | aS'n12267677'
1978 | p988
1979 | aS'n12620546'
1980 | p989
1981 | aS'n12768682'
1982 | p990
1983 | aS'n12985857'
1984 | p991
1985 | aS'n12998815'
1986 | p992
1987 | aS'n13037406'
1988 | p993
1989 | aS'n13040303'
1990 | p994
1991 | aS'n13044778'
1992 | p995
1993 | aS'n13052670'
1994 | p996
1995 | aS'n13054560'
1996 | p997
1997 | aS'n13133613'
1998 | p998
1999 | aS'n15075141'
2000 | p999
2001 | a.


--------------------------------------------------------------------------------
/Images/apple.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/Images/apple.jpg


--------------------------------------------------------------------------------
/Images/beach.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/Images/beach.jpg


--------------------------------------------------------------------------------
/Images/beetroot.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/Images/beetroot.jpg


--------------------------------------------------------------------------------
/Images/cat.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/Images/cat.jpg


--------------------------------------------------------------------------------
/Images/dog.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/Images/dog.jpg


--------------------------------------------------------------------------------
/Images/dolph.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/Images/dolph.jpg


--------------------------------------------------------------------------------
/Images/einstein.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/Images/einstein.jpg


--------------------------------------------------------------------------------
/Images/kitten.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/Images/kitten.jpg


--------------------------------------------------------------------------------
/Images/orange.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/Images/orange.jpg


--------------------------------------------------------------------------------
/Images/pig.html:
--------------------------------------------------------------------------------
 1 | 
 2 | <table align="center" style="background-color:#FFFFFF; margin:5px auto; padding:3px 5px; width:100%;">
 3 | <tr>
 4 | 	<td style="vertical-align:top; padding:10px; width:680px;">
 5 | 
 6 | <div id="custom-content" class="white-popup-block" style="max-width:750px; margin: 20px auto;">
 7 | <style>
 8 |     #custom-content img {max-width: 100%;margin-bottom: 10px;}
 9 |     </style>
10 | <div style="margin-bottom: 8px;">
11 | </div>
12 | <a href="/" style="color: black;font-size: 16px;"><i>Home</i></a> / <a href="pig.html" style="color: black;font-size: 16px;"><i>Pig</i></a>
13 | <script async src="//pagead2.googlesyndication.com/pagead/js/adsbygoogle.js"></script>
14 | <!-- animal3 -->
15 | <ins class="adsbygoogle"
16 |      style="display:block"
17 |      data-ad-client="ca-pub-3808762131343667"
18 |      data-ad-slot="7278748839"
19 |      data-ad-format="auto"></ins>
20 | <script>
21 | (adsbygoogle = window.adsbygoogle || []).push({});
22 | </script>
23 | 				 <img src="/data_images/pig/pig1.jpg" alt="" style="padding: 0;width: 600px;height: 400px;">
24 | <script async src="//pagead2.googlesyndication.com/pagead/js/adsbygoogle.js"></script>
25 | <!-- animal3 -->
26 | <ins class="adsbygoogle"
27 |      style="display:block"
28 |      data-ad-client="ca-pub-3808762131343667"
29 |      data-ad-slot="7278748839"
30 |      data-ad-format="auto"></ins>
31 | <script>
32 | (adsbygoogle = window.adsbygoogle || []).push({});
33 | </script>
34 | 		  
35 | 
36 | 
37 | 	</td>
38 | 	<td style="padding:0px 0px 0px 10px; width:300px;">
39 | <script async src="//pagead2.googlesyndication.com/pagead/js/adsbygoogle.js"></script>
40 | <!-- animal600 -->
41 | <ins class="adsbygoogle"
42 |      style="display:inline-block;width:300px;height:600px"
43 |      data-ad-client="ca-pub-3808762131343667"
44 |      data-ad-slot="3382926034"></ins>
45 | <script>
46 | (adsbygoogle = window.adsbygoogle || []).push({});
47 | </script>	
48 | 	</td>
49 | </tr>
50 | </table>
51 | <div style="float:right;">
52 | <!--LiveInternet counter--><script type="text/javascript"><!--
53 | document.write("<a href='//www.liveinternet.ru/click' "+
54 | "target=_blank><img src='//counter.yadro.ru/hit?t38.1;r"+
55 | escape(document.referrer)+((typeof(screen)=="undefined")?"":
56 | ";s"+screen.width+"*"+screen.height+"*"+(screen.colorDepth?
57 | screen.colorDepth:screen.pixelDepth))+";u"+escape(document.URL)+
58 | ";"+Math.random()+
59 | "' alt='' title='LiveInternet' "+
60 | "border='0' width='31' height='31'></a>")
61 | //--></script><!--/LiveInternet-->	     		
62 | </div>
63 | 
64 | 


--------------------------------------------------------------------------------
/Images/pig.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/Images/pig.jpg


--------------------------------------------------------------------------------
/Images/pig_with_dog.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/Images/pig_with_dog.jpg


--------------------------------------------------------------------------------
/Images/rose.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/Images/rose.jpg


--------------------------------------------------------------------------------
/Images/sky.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/Images/sky.jpg


--------------------------------------------------------------------------------
/Poster.pdf:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/Poster.pdf


--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
 1 | Repository uses code from assignments of Stanford`s course - CS 231n Convolutional Networks
 2 | 
 3 | Setup:
 4 | 
 5 | 1. Inside the library/ directory, run 
 6 | python setup.py build_ext --inplace
 7 | 
 8 | 2. Download the file vgg16_weights.h5 from https://drive.google.com/file/d/0Bz7KyqmuGsilT0J5dmRCM0ROVHc/view?usp=sharing
 9 | Place the downloaded weight file (vgg16_weights.h5) in Data folder
10 | 


--------------------------------------------------------------------------------
/Validation.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |  "cells": [
  3 |   {
  4 |    "cell_type": "code",
  5 |    "execution_count": 37,
  6 |    "metadata": {
  7 |     "collapsed": false
  8 |    },
  9 |    "outputs": [
 10 |     {
 11 |      "name": "stdout",
 12 |      "output_type": "stream",
 13 |      "text": [
 14 |       "The autoreload extension is already loaded. To reload it, use:\n",
 15 |       "  %reload_ext autoreload\n",
 16 |       "Loaded Model\n",
 17 |       "Loaded Class File\n",
 18 |       "Loaded Input Data Set\n",
 19 |       "[8] 11,349,0.414949\n",
 20 |       "[9] Skip (208,65): http://farm2.static.flickr.com/1246/905331985_1f0065e586.jpg : URL is Empty(White) Image\n",
 21 |       "[10] 338,246,0.404768\n"
 22 |      ]
 23 |     },
 24 |     {
 25 |      "ename": "KeyboardInterrupt",
 26 |      "evalue": "",
 27 |      "output_type": "error",
 28 |      "traceback": [
 29 |       "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
 30 |       "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
 31 |       "\u001b[0;32m<ipython-input-37-45fb8615163d>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     57\u001b[0m         \u001b[0;32mprint\u001b[0m \u001b[0;34m\"[%d] Skip (%d,%d): %s : URL is Empty(White) Image\"\u001b[0m\u001b[0;34m%\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mindex\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mclass_idx\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mxmlidx\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0murl\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     58\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 59\u001b[0;31m         \u001b[0mbbox_coords\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mbbox\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mim\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mclass_no\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mclass_idx\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mn_neurons\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m5\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mkmax\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m10\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     60\u001b[0m         \u001b[0mxmin_out\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mxmax_out\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mymin_out\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mymax_out\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mbbox_coords\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     61\u001b[0m         \u001b[0mprecision\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0meval_precision\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mbbox_coords\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mxmin\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mxmax\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mymin\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mymax\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
 32 |       "\u001b[0;32m/Users/deepakmenghani/Acads/Stanford/Q5/CS231n/project/FastLocalization/library/localization.py\u001b[0m in \u001b[0;36mbbox\u001b[0;34m(im, model, layer, n_neurons, kmax, class_no)\u001b[0m\n\u001b[1;32m     22\u001b[0m     \u001b[0;31m# k_max = Maximum Number of neurons to evaluate\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     23\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 24\u001b[0;31m     \u001b[0mmask\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcache\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mget_localization_mask\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mim\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlayer\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mn_neurons\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkmax\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mclass_no\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     25\u001b[0m     \u001b[0mclass_no\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0moriginal_size\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcache\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     26\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0mget_box_from_mask\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmask\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moriginal_size\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
 33 |       "\u001b[0;32m/Users/deepakmenghani/Acads/Stanford/Q5/CS231n/project/FastLocalization/library/localization.py\u001b[0m in \u001b[0;36mget_localization_mask\u001b[0;34m(im, model, layer, n_neurons, kmax, class_no)\u001b[0m\n\u001b[1;32m     45\u001b[0m     \u001b[0;31m#Get the Filters of Interest in Sorted Order\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     46\u001b[0m     \u001b[0mamax\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfilter_of_intr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mactivs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mback_grad\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mkmax\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mlayer\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 47\u001b[0;31m     \u001b[0mfilter_scores\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mget_filter_scores\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mamax\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mim\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mactivs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mcaches\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlayer\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mclass_no\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpercentile_thresh\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m40\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0muse_blob\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mTrue\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     48\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     49\u001b[0m     \u001b[0;31m#Union Blobs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
 34 |       "\u001b[0;32m/Users/deepakmenghani/Acads/Stanford/Q5/CS231n/project/FastLocalization/deconv_utils.pyc\u001b[0m in \u001b[0;36mget_filter_scores\u001b[0;34m(amax, model, im, activs, caches, layer, class_no, percentile_thresh, use_blob, verbose)\u001b[0m\n\u001b[1;32m     28\u001b[0m   \u001b[0mfilter_scores\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     29\u001b[0m   \u001b[0;32mfor\u001b[0m \u001b[0mzero\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0my\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mamax\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 30\u001b[0;31m       \u001b[0mback_grad\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdeconv\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mactivs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mcaches\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mlayer\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     31\u001b[0m       \u001b[0;32mif\u001b[0m \u001b[0muse_blob\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     32\u001b[0m           \u001b[0mblob\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfind_blob\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mback_grad\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mpercentile_thresh\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m80\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
 35 |       "\u001b[0;32m/Users/deepakmenghani/Acads/Stanford/Q5/CS231n/project/FastLocalization/deconv_utils.pyc\u001b[0m in \u001b[0;36mdeconv\u001b[0;34m(model, activs, caches, layer, neuron)\u001b[0m\n\u001b[1;32m     83\u001b[0m     \u001b[0;32mfor\u001b[0m \u001b[0mi\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mreversed\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlayer\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     84\u001b[0m         \u001b[0mback_grad\u001b[0m  \u001b[0;34m=\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mback_grad\u001b[0m\u001b[0;34m>\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0mback_grad\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 85\u001b[0;31m         \u001b[0mback_grad\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0m_\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mback_grad\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mcaches\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     86\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0mback_grad\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     87\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
 36 |       "\u001b[0;32m/Users/deepakmenghani/Acads/Stanford/Q5/CS231n/project/FastLocalization/library/classifiers/pretrained_vgg16.pyc\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(self, dout, cache)\u001b[0m\n\u001b[1;32m    290\u001b[0m         \u001b[0;31m# This is a conv layer\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    291\u001b[0m         \u001b[0;32mif\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mmax_pool_layers\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 292\u001b[0;31m             \u001b[0mtemp\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mconv_relu_pool_backward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdnext_a\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mlayer_caches\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mj\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    293\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    294\u001b[0m             \u001b[0mtemp\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mconv_relu_backward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdnext_a\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlayer_caches\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mj\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
 37 |       "\u001b[0;32m/Users/deepakmenghani/Acads/Stanford/Q5/CS231n/project/FastLocalization/library/layer_utils.pyc\u001b[0m in \u001b[0;36mconv_relu_pool_backward\u001b[0;34m(dout, cache)\u001b[0m\n\u001b[1;32m    138\u001b[0m   \u001b[0mds\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmax_pool_backward_fast\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdout\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpool_cache\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    139\u001b[0m   \u001b[0mda\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mrelu_backward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mds\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrelu_cache\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 140\u001b[0;31m   \u001b[0mdx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdw\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdb\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mconv_backward_fast\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mda\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mconv_cache\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    141\u001b[0m   \u001b[0;32mreturn\u001b[0m \u001b[0mdx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdw\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdb\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    142\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
 38 |       "\u001b[0;32m/Users/deepakmenghani/Acads/Stanford/Q5/CS231n/project/FastLocalization/library/fast_layers.pyc\u001b[0m in \u001b[0;36mconv_backward_strides\u001b[0;34m(dout, cache)\u001b[0m\n\u001b[1;32m     97\u001b[0m   \u001b[0mdw\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdout_reshaped\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx_cols\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mT\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mreshape\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mw\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshape\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     98\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 99\u001b[0;31m   \u001b[0mdx_cols\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mw\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mreshape\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mF\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mT\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdout_reshaped\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    100\u001b[0m   \u001b[0mdx_cols\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshape\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mC\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mHH\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mWW\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mN\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mout_h\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mout_w\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    101\u001b[0m   \u001b[0mdx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcol2im_6d_cython\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdx_cols\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mN\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mC\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mH\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mW\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mHH\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mWW\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpad\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mstride\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
 39 |       "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
 40 |      ]
 41 |     }
 42 |    ],
 43 |    "source": [
 44 |     "import types\n",
 45 |     "from library.localization import *\n",
 46 |     "from deconv_utils import *\n",
 47 |     "from library.image_utils import *\n",
 48 |     "import csv\n",
 49 |     "import sys\n",
 50 |     "\n",
 51 |     "# for auto-reloading external modules\n",
 52 |     "# see http://stackoverflow.com/questions/1907993/autoreload-of-modules-in-ipython\n",
 53 |     "%load_ext autoreload\n",
 54 |     "%autoreload 2\n",
 55 |     "\n",
 56 |     "#Validation & Model File Inputs\n",
 57 |     "data_file = \"Data/validation_1\"\n",
 58 |     "data_extn = \".pkl\"\n",
 59 |     "results_file = data_file + \"_results.csv\"\n",
 60 |     "model_file = 'Data/vgg16_weights.h5'\n",
 61 |     "class_file = 'Data/CLASSES.pkl'\n",
 62 |     "\n",
 63 |     "#Image Processing Inputs\n",
 64 |     "Param.num_dilation = 15\n",
 65 |     "\n",
 66 |     "start_index = 0\n",
 67 |     "try:\n",
 68 |     "    #Check if csv file exists and read the last index outputted\n",
 69 |     "    results_csv = open(results_file,'r')\n",
 70 |     "    lastline = results_csv.read().split('\\n')[-2]\n",
 71 |     "    start_index = int(lastline.split(',')[0]) + 1\n",
 72 |     "except IOError as e:\n",
 73 |     "    pass\n",
 74 |     "\n",
 75 |     "#Write to CSV file\n",
 76 |     "results_csv = open(results_file,'a')\n",
 77 |     "\n",
 78 |     "#Load the VGG Model and ImageNet Class Mappings\n",
 79 |     "from library.classifiers.pretrained_vgg16 import PretrainedVGG\n",
 80 |     "model = PretrainedVGG(h5_file = model_file)\n",
 81 |     "print \"Loaded Model\"\n",
 82 |     "CLASSES = pickle.load(open(class_file))\n",
 83 |     "print \"Loaded Class File\"\n",
 84 |     "\n",
 85 |     "#Read Input File for the candidates\n",
 86 |     "candidates = pickle.load(open(data_file+data_extn))\n",
 87 |     "print \"Loaded Input Data Set\"\n",
 88 |     "\n",
 89 |     "#Read Input File and start at the index higher than that of last line\n",
 90 |     "\n",
 91 |     "for index in range(start_index,len(candidates)):\n",
 92 |     "    ret = candidates[index]\n",
 93 |     "    class_wnid,imgid,class_idx,xmlidx,url,xmin,xmax,ymin,ymax = ret\n",
 94 |     "    im = image_from_url(url)\n",
 95 |     "    if type(im) == types.NoneType:\n",
 96 |     "        #Debug Print\n",
 97 |     "        print \"[%d] Skip (%d,%d): %s : URL is bad.\"%(index, class_idx,xmlidx, url)\n",
 98 |     "    elif np.mean(im)>=253:\n",
 99 |     "        #Debug Print\n",
100 |     "        print \"[%d] Skip (%d,%d): %s : URL is Empty(White) Image\"%(index, class_idx,xmlidx,url)\n",
101 |     "    else:  \n",
102 |     "        bbox_coords = bbox(im,model,class_no=class_idx,n_neurons = 5,kmax=10)\n",
103 |     "        xmin_out, xmax_out, ymin_out, ymax_out = bbox_coords\n",
104 |     "        precision = eval_precision(bbox_coords, (xmin,xmax,ymin,ymax))\n",
105 |     "        #Write Results to file\n",
106 |     "        result_row = \"%d, %s, %d, %d, %f,\"%(index, imgid, class_idx, xmlidx, precision)\n",
107 |     "        result_row += \"%d, %d, %d, %d,\"%(xmin, xmax, ymin, ymax,)\n",
108 |     "        result_row += \"%d, %d, %d, %d\"%(xmin_out, xmax_out, ymin_out, ymax_out)\n",
109 |     "        result_row += \"\\n\"\n",
110 |     "        results_csv.write(result_row)\n",
111 |     "        results_csv.flush()\n",
112 |     "        #Debug Print\n",
113 |     "        print \"[%d] %d,%d,%f\"%(index, class_idx,xmlidx,precision)"
114 |    ]
115 |   },
116 |   {
117 |    "cell_type": "code",
118 |    "execution_count": null,
119 |    "metadata": {
120 |     "collapsed": true
121 |    },
122 |    "outputs": [],
123 |    "source": []
124 |   }
125 |  ],
126 |  "metadata": {
127 |   "kernelspec": {
128 |    "display_name": "Python 2",
129 |    "language": "python",
130 |    "name": "python2"
131 |   },
132 |   "language_info": {
133 |    "codemirror_mode": {
134 |     "name": "ipython",
135 |     "version": 2
136 |    },
137 |    "file_extension": ".py",
138 |    "mimetype": "text/x-python",
139 |    "name": "python",
140 |    "nbconvert_exporter": "python",
141 |    "pygments_lexer": "ipython2",
142 |    "version": "2.7.8"
143 |   }
144 |  },
145 |  "nbformat": 4,
146 |  "nbformat_minor": 0
147 | }
148 | 


--------------------------------------------------------------------------------
/deconv_utils.py:
--------------------------------------------------------------------------------
  1 | import cv2
  2 | import numpy as np
  3 | import matplotlib.pyplot as plt
  4 | from scipy.misc import imread, imresize
  5 | from scipy import ndimage
  6 | 
  7 | #Global Variables
  8 | 
  9 | #Parameters to Process blob
 10 | class Param:
 11 |   process_conv_size = (4,4)
 12 |   num_dilation = 20
 13 |   num_erosion = 15
 14 | 
 15 | def get_filter_scores2(amax, model, im, activs,caches,layer,  class_no, percentile_thresh=80):
 16 |     filter_scores = []
 17 |     for zero,i,x,y in amax:
 18 |         #print i,(x,y)
 19 |         back_grad = deconv(model,activs,caches,layer,(0,i,x,y))
 20 |         xmin,xmax,ymin,ymax = find_box(back_grad,percentile_thresh=80)
 21 |         if (xmin,xmax,ymin,ymax) == (0,0,0,0):
 22 |             continue
 23 |         n_score = get_score(im,class_no,model, xmin,xmax,ymin,ymax)
 24 |         filter_scores += [[i,n_score,xmin,xmax,ymin,ymax]]
 25 |     return filter_scores
 26 | 
 27 | def get_filter_scores(amax, model, im, activs,caches,layer,  class_no, percentile_thresh=80,use_blob = False, verbose=False):
 28 |   filter_scores = []
 29 |   for zero,i,x,y in amax:
 30 |       back_grad = deconv(model,activs,caches,layer,(0,i,x,y))
 31 |       plt.figure()
 32 |       plt.imshow(back_grad[0].transpose(1,2,0))
 33 |       if use_blob:
 34 |           blob = find_blob(back_grad,percentile_thresh=80)
 35 |       else:
 36 |           xmin,xmax,ymin,ymax = find_box(back_grad,percentile_thresh=percentile_thresh)
 37 |           blob = np.zeros(im.shape)
 38 |           blob[0,:,xmin:(xmax+1),ymin:(ymax+1)] = 1
 39 |       if np.sum(blob) == 0:
 40 |           continue
 41 |       n_score = get_score(im,class_no,model, mask = blob)
 42 |       if (verbose):
 43 |             print "score =" ,n_score
 44 |       if use_blob:
 45 |         filter_scores += [[i,n_score,blob,0,0,0,0]]
 46 |       else:
 47 |         filter_scores += [[i,n_score,blob,xmin,xmax,ymin,ymax]]
 48 |   return filter_scores
 49 | 
 50 | def get_fast_filter_scores(amax, model, im, activs,caches,layer,  class_no, percentile_thresh=80,use_blob = False, verbose=False):
 51 |   filter_scores = []
 52 |   k=0
 53 |   for zero,i,x,y in amax:
 54 |       #if np.sum(activs[layer][0,i]>0)
 55 |       back_grad = deconv(model,activs,caches,layer,(0,i,x,y))
 56 |       if use_blob:
 57 |           blob = find_blob(back_grad,percentile_thresh=80)
 58 |       else:
 59 |           xmin,xmax,ymin,ymax = find_box(back_grad,percentile_thresh=percentile_thresh)
 60 |           blob = np.zeros(im.shape)
 61 |           blob[0,:,xmin:(xmax+1),ymin:(ymax+1)] = 1
 62 |       if np.sum(blob) == 0:
 63 |           continue
 64 |       n_score = -k
 65 |       if verbose:
 66 |           print "score =" ,n_score
 67 |       if use_blob:
 68 |         filter_scores += [[i,n_score,blob,0,0,0,0]]
 69 |       else:
 70 |         filter_scores += [[i,n_score,blob,xmin,xmax,ymin,ymax]]
 71 |       k=k+1
 72 |   return filter_scores
 73 | 
 74 | def get_backgrad (activs, model, class_no, layer, caches):
 75 |     back_grad= np.zeros(activs[15].shape)
 76 |     back_grad[0,class_no]=1
 77 |     for i in reversed(range(layer,16)):
 78 |         back_grad  = (back_grad>0)*back_grad
 79 |         back_grad, _ = model.backward(back_grad,caches[i])
 80 |     return back_grad
 81 | 
 82 | def deconv(model,activs,caches,layer,neuron):
 83 |     back_grad = np.zeros(activs[layer].shape)
 84 |     back_grad[neuron] = 1
 85 |     for i in reversed(range(layer+1)):
 86 |         back_grad  = (back_grad>0)*back_grad
 87 |         back_grad, _ = model.backward(back_grad,caches[i])
 88 |     return back_grad
 89 | 
 90 | #Define Function for deconv
 91 | def deconv_2(model,activs, caches,layer,neuron,slayer):
 92 |     back_grad = np.zeros(activs[slayer].shape)
 93 |     print back_grad.shape
 94 |     if(len(neuron)==3):
 95 |         back_grad[neuron[0],neuron[1]] = 1
 96 |     else:
 97 |         back_grad[neuron] = 1
 98 |     for i in reversed(range(layer,slayer+1)):
 99 |         back_grad  = (back_grad>0)*back_grad
100 |         back_grad, _ = model.backward(back_grad,caches[i])
101 |     return back_grad
102 | 
103 | #Takes the processed image to give activations and caches for entire forward pass
104 | def get_activs(model, im, num_layers=16):
105 |     activ = im
106 |     caches = []
107 |     activs = []
108 |     for i in range(num_layers):
109 |         out,cache = model.forward(activ,start = i, end=i)
110 |         activ = out;
111 |         activs += [activ]
112 |         caches += [cache]
113 |     return activs,caches
114 | 
115 | ###############TO-DO###################
116 | def deconv_batch(model, ims, layer=10):
117 |     #Function to get deconv of a batch of images
118 |     #ims : N X 3 X 224 X 224
119 |     pass
120 |     
121 | ##############TO DO###########
122 | #Function to get blob
123 | def get_blob():
124 |     pass
125 |     
126 | #Image Utilities
127 | def load_image(imgf):
128 |     im = imread(imgf)
129 |     im = resize_image(im)
130 |     return process_image(im)
131 | 
132 | def load_image_cv2(imgf):
133 |     im = cv2.imread(imgf)
134 |     im = resize_image(im)
135 |     return process_image(im)
136 | 
137 | def resize_image(im):
138 |     return cv2.resize(im, (224, 224)).astype(np.float32)
139 | 
140 | def process_image(im):
141 |     im[:,:,0] -= 103.939
142 |     im[:,:,1] -= 116.779
143 |     im[:,:,2] -= 123.68
144 |     im = im.transpose((2,0,1))
145 |     im = np.expand_dims(im, axis=0)
146 |     return im
147 | 
148 | def deprocess_image(img):
149 |     im = img[0].transpose(1,2,0)
150 |     im[:,:,0] += 103.939
151 |     im[:,:,1] += 116.779
152 |     im[:,:,2] += 123.68
153 |     im = im.astype(np.uint8)
154 |     return im
155 | 
156 | #Takes  input 
157 | def plot_image(im):
158 |     im = deprocess_image(im)
159 |     plt.figure()
160 |     plt.imshow(im)
161 | 
162 | def plot_image_cv2(im):
163 |     im = deprocess_image(im)
164 |     im = cv2.cvtColor(im, cv2.cv.CV_BGR2RGB)
165 |     plt.figure()
166 |     plt.imshow(im)
167 |     #cv.imshow("Image",im)
168 |     
169 | #import matplotlib.pyplot as plt
170 | def grid_plot_activs(act):
171 |     grid = visualize_grid((act).transpose(1,2,3,0))
172 |     plt.imshow(grid.transpose(2,0,1)[0])
173 |     #plt.axis('off')
174 |     plt.gcf().set_size_inches(10, 10)
175 |     plt.show()
176 |     
177 | def find_box(back_grad, percentile_thresh = 40):
178 |     meanimg = np.mean(back_grad[0],axis=0)
179 |     if np.sum(abs(meanimg)) < 1e-16:
180 |         return 0,0,0,0
181 |     thresh = np.percentile(meanimg[meanimg>0],[percentile_thresh])[0]
182 |     meanimg = np.mean(back_grad[0],axis=0)
183 |     threshimg = np.mean(back_grad[0],axis=0)>thresh
184 |     #plt.imshow(np.mean(back_grad[0],axis=0)>thresh)
185 |     idxs = np.where(np.sum(threshimg,axis = 1)>0)[0]
186 |     xmin,xmax = min(idxs),max(idxs)
187 |     idxs = np.where(np.sum(threshimg,axis = 0)>0)[0]
188 |     ymin,ymax = min(idxs),max(idxs)
189 |     return xmin,xmax,ymin,ymax
190 | 
191 | #Get Score for a given image and class number. If mask != 0, it uses the mask area, else defaults to the area specified by the rectangle specified by (xmin,ymin) (xmax,ymax)
192 | def get_score(im,class_no, model, xmin=0,xmax=0,ymin=0,ymax=0,mask=0):
193 |     #print xmin,xmax,ymin,ymax
194 |     if np.sum(mask )== 0:
195 |         mask = np.zeros(im.shape)
196 |         mask[0,:,xmin:(xmax+1),ymin:(ymax+1)] = 1
197 |     newim = im.copy()*mask
198 |     scores,_ = model.forward(newim)
199 |     return scores[0,class_no]
200 | 
201 | #Given Activations, Deconv, Layer Number
202 | #k = No of Neurons of Interest
203 | #Returns the filters of interest
204 | def filter_of_intr(activs,back_grad,kmax,layer):
205 |    cum =[]
206 |    tally = []
207 |    tmp1 = activs[layer] *back_grad
208 |    k=0
209 |    while k<kmax:
210 |        index = np.unravel_index(tmp1.argmax(), tmp1.shape)
211 |        #print index[1]
212 |        if index[1] not in tally:
213 |            tally += [index[1]]
214 |            cum+= [index]
215 |            k+=1
216 |        tmp1[index] = -1000
217 |    return cum
218 | 
219 | 
220 | 
221 | def grabCut(im2,xmin,xmax,ymin,ymax):
222 |     h,w = im2.shape[:2]
223 |     mask = np.zeros((h,w),dtype='uint8')
224 |     rect = (ymin,xmin,ymax-ymin,xmax-xmin)
225 |     #rect = (10,10,213,213)
226 |     tmp1 = np.zeros((1, 13 * 5))
227 |     tmp2 = np.zeros((1, 13 * 5))  
228 |     cv2.grabCut(im2,mask,rect,tmp1,tmp2,10,mode=cv2.GC_INIT_WITH_RECT)
229 |     mask[mask==2]=0
230 |     return mask
231 | 
232 | 
233 | def process_blob(cim):
234 |     #cim = ndimage.binary_erosion(cim>0)
235 |     for i in range(4):
236 |         cim = ndimage.binary_erosion(cim>0)
237 |         cim=ndimage.binary_dilation(cim>0)
238 | 
239 |     filterk = np.ones(Param.process_conv_size);
240 |     cim = ndimage.convolve(cim, filterk, mode='constant', cval=0.0)
241 | 
242 |     for i in range(Param.num_dilation):
243 |         cim = ndimage.binary_dilation(cim>0)
244 |     for i in range(Param.num_erosion):
245 |         cim = ndimage.binary_erosion(cim>0)
246 |     return cim
247 | 
248 | def find_blob(back_grad, percentile_thresh = 40):
249 |     meanimg = np.mean(back_grad[0],axis=0)
250 |     if np.sum(abs(meanimg)) < 1e-16:
251 |         return 0,0,0,0
252 |     thresh = np.percentile(meanimg[meanimg>0],[percentile_thresh])[0]
253 |     meanimg = np.mean(back_grad[0],axis=0)
254 |     threshimg = np.mean(back_grad[0],axis=0)>thresh
255 |     
256 |     return process_blob(threshimg)
257 | 


--------------------------------------------------------------------------------
/frameworkpython:
--------------------------------------------------------------------------------
 1 | #!/bin/bash
 2 | 
 3 | # what real Python executable to use
 4 | PYVER=2.7
 5 | PATHTOPYTHON=/usr/local/bin/
 6 | PYTHON=${PATHTOPYTHON}python${PYVER}
 7 | 
 8 | # find the root of the virtualenv, it should be the parent of the dir this script is in
 9 | ENV=`$PYTHON -c "import os; print os.path.abspath(os.path.join(os.path.dirname(\"$0\"), '..'))"`
10 | 
11 | # now run Python with the virtualenv set as Python's HOME
12 | export PYTHONHOME=$ENV
13 | exec $PYTHON "$@"
14 | 


--------------------------------------------------------------------------------
/library/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/library/__init__.py


--------------------------------------------------------------------------------
/library/classifiers/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/deepakrox/FastObjectLocalization/5dcb923600b0ca3bb7f855f31bc606d85c3b07e8/library/classifiers/__init__.py


--------------------------------------------------------------------------------
/library/classifiers/pretrained_vgg16.py:
--------------------------------------------------------------------------------
  1 | import numpy as np
  2 | import h5py
  3 | 
  4 | from library.layers import *
  5 | from library.fast_layers import *
  6 | from library.layer_utils import *
  7 | 
  8 | 
  9 | class PretrainedVGG(object):
 10 |   def __init__(self, dtype=np.float32, num_classes=1000, input_size=224, h5_file=None,verbose=False):
 11 |     #Input Size is 224X224
 12 |     #Num classes = 1000 for ImageNet
 13 |     
 14 |     self.dtype = dtype
 15 |     self.conv_params = []
 16 |     self.input_size = input_size
 17 |     self.num_classes = num_classes
 18 |     
 19 |     # TODO: In the future it would be nice if the architecture could be loaded from
 20 |     # the HDF5 file rather than being hardcoded. For now this will have to do.
 21 |     
 22 |     self.conv_params.append({'stride': 1, 'pad': 1})
 23 |     self.conv_params.append({'stride': 1, 'pad': 1})
 24 |     
 25 |     self.conv_params.append({'stride': 1, 'pad': 1})
 26 |     self.conv_params.append({'stride': 1, 'pad': 1})
 27 |     
 28 |     self.conv_params.append({'stride': 1, 'pad': 1})
 29 |     self.conv_params.append({'stride': 1, 'pad': 1})
 30 |     self.conv_params.append({'stride': 1, 'pad': 1})
 31 |     
 32 |     self.conv_params.append({'stride': 1, 'pad': 1})
 33 |     self.conv_params.append({'stride': 1, 'pad': 1})
 34 |     self.conv_params.append({'stride': 1, 'pad': 1})
 35 |     
 36 |     self.conv_params.append({'stride': 1, 'pad': 1})
 37 |     self.conv_params.append({'stride': 1, 'pad': 1})
 38 |     self.conv_params.append({'stride': 1, 'pad': 1})
 39 |     
 40 |     #Map Describing the layer type
 41 |     self.layer_map = {1:'conv', 2:'relu', 3:'conv', 4: 'relu', 5:'maxpool',
 42 |                       6:'conv', 7:'relu', 8:'conv', 9: 'relu', 10:'maxpool',
 43 |                       11:'conv', 12:'relu', 13:'conv', 14: 'relu', 15:'conv', 16:'relu' ,17:'maxpool',
 44 |                       18:'conv', 19:'relu', 20:'conv', 21: 'relu', 22:'conv', 23:'relu' ,24:'maxpool',
 45 |                       25:'conv', 26:'relu', 27:'conv', 28: 'relu', 29:'conv', 30:'relu' ,31:'maxpool',
 46 |                       32:'affine', 33:'relu', 34:'affine', 35:'relu', 36:'affine',37:'softmax'
 47 |                      }
 48 |     #self. weight_layers = {1:1,2:3,3:6,4:8,5:11,6:13,7:15,8:18,9:20,10:22,11:25,12:27,13:29,14:32,15:34,16:36}
 49 |     self.weight_layers_inv = {1:1,3:2,6:3,8:4,11:5,13:6,15:7,18:8,20:9,22:10,25:11,27:12,29:13,32:14,34:15,36:16}
 50 | 
 51 |     self.filter_sizes = [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3]
 52 |     self.num_filters = [64, 64, 128, 128, 256, 256, 256, 512, 512, 512, 512, 512, 512]
 53 | 
 54 |     hidden_dim_1 = 4096
 55 |     hidden_dim_2 = 4096
 56 |     
 57 |     cur_size = input_size
 58 |     prev_dim = 3
 59 |     self.params = {}
 60 |     
 61 |     for i, (f, next_dim) in enumerate(zip(self.filter_sizes, self.num_filters)):
 62 |       self.params['W%d' % (i + 1)] = np.random.randn(next_dim, prev_dim, f, f)
 63 |       self.params['b%d' % (i + 1)] = np.zeros(next_dim)
 64 |       prev_dim = next_dim
 65 |       
 66 |     cur_size = cur_size/32
 67 |     # Add a fully-connected layers
 68 |     fan_in = cur_size * cur_size * self.num_filters[-1]
 69 |     self.params['W%d' % (i + 2)] = np.zeros((fan_in,hidden_dim_1))
 70 |     self.params['b%d' % (i + 2)] = np.zeros(hidden_dim_1)
 71 |     
 72 |     self.params['W%d' % (i + 3)] = np.zeros((hidden_dim_1, hidden_dim_2))
 73 |     self.params['b%d' % (i + 3)] = np.zeros(hidden_dim_2)
 74 | 
 75 |     self.params['W%d' % (i + 4)] = np.zeros((hidden_dim_2, num_classes))
 76 |     self.params['b%d' % (i + 4)] = np.zeros(num_classes)
 77 | 
 78 |     for k, v in self.params.iteritems():
 79 |       self.params[k] = v.astype(dtype)
 80 | 
 81 |     if h5_file is not None:
 82 |       self.load_weights(h5_file, verbose)
 83 |       pass
 84 | 
 85 |   
 86 |   def load_weights(self, h5_file, verbose=False):
 87 |     """
 88 |     Load pretrained weights from an HDF5 file.
 89 | 
 90 |     Inputs:
 91 |     - h5_file: Path to the HDF5 file where pretrained weights are stored.
 92 |     - verbose: Whether to print debugging info
 93 |     """
 94 |     with h5py.File(h5_file, 'r') as f:
 95 |       for k, v in f.iteritems():
 96 |         layer_no = int(k.split('_')[1]) 
 97 |         if(layer_no in self.weight_layers_inv.keys() ):
 98 |             for key,value in v.iteritems():
 99 |                 wt_layer_no = self.weight_layers_inv[layer_no]
100 |                 if('param_0' in value.name):
101 |                    param_name = 'W%d' % wt_layer_no
102 |                    param_value = np.asarray(value.value)
103 |                    if self.layer_map[layer_no] == 'conv':
104 |                         NF, D,W,H = param_value.shape
105 |                         param_value[:,:,range(W-1,-1,-1),:] =param_value.copy()
106 |                         param_value[:,:,:,range(H-1,-1,-1)] =param_value.copy()
107 |                 elif ('param_1' in value.name):
108 |                    param_name = 'b%d' % wt_layer_no
109 |                    param_value = np.asarray(value.value)
110 |                 if verbose: print param_name, self.params[param_name].shape
111 |                 if param_value.shape == self.params[param_name].shape:
112 |                     self.params[param_name] = param_value
113 |                 elif param_value.T.shape == self.params[param_name].shape:
114 |                     asself.params[param_name] = param_value.T
115 |                 else:
116 |                     raise ValueError('shapes for %s do not match' % param_name)
117 |     for k, v in self.params.iteritems():
118 |       self.params[k] = v.astype(self.dtype)
119 | 
120 |   
121 |   def forward(self, X, start=None, end=None, mode='test'):
122 |     """
123 |     Run part of the model forward, starting and ending at an arbitrary layer,
124 |     in either training mode or testing mode.
125 | 
126 |     You can pass arbitrary input to the starting layer, and you will receive
127 |     output from the ending layer and a cache object that can be used to run
128 |     the model backward over the same set of layers.
129 | 
130 |     For the purposes of this function, a "layer" is one of the following blocks:
131 | 
132 |     [conv3-64 - relu] X 2
133 |     MaxPool
134 |     [conv3-128 - relu] X 2
135 |     MaxPool
136 |     [conv3-256 - relu] X 3
137 |     MaxPool
138 |     [conv3-512 - relu] X 2
139 |     [conv1-512 - relu] X 1
140 |     MaxPool
141 |     [conv3-512 - relu] X 2
142 |     [conv1-512 - relu] X 1
143 |     MaxPool
144 |     [affine - relu] X 2
145 |     [affine] 
146 |     Softmax
147 |     
148 |     Layer 0
149 |     Layer1 - Conv3-64
150 |     Layer2 - Relu
151 |     Layer3 - Conv3-64
152 |     Layer4 - Relu
153 |     Layer5 - Maxpool
154 |     
155 |     Layer6 - Conv3-128
156 |     Layer7 - Relu
157 |     Layer8 - Conv3-128
158 |         9  - Relu
159 |         10 - Maxpool
160 |         
161 |         11 - Conv3-256
162 |         12 - Relu
163 |         13 - Conv3-256
164 |         14 - Relu
165 |         15 - Conv3-256
166 |         16 - Relu
167 |         17 - Maxpool
168 |         
169 |         18 - conv3-512
170 |         19 - Relu
171 |         20 - Conv3-512
172 |         21 - Relu
173 |         22 - Conv1-512
174 |         23 - Relu
175 |         24 - Maxpool
176 |         
177 |         25 - conv3-512
178 |         26 - Relu
179 |         27 - Conv3-512
180 |         28 - Relu
181 |  
182 |         29 - Conv3-512
183 |         30 - Relu
184 |         31 - Maxpool
185 |         
186 |         32 - Affine FC-4096
187 |         33 - Relu
188 |         34 - Affine FC-4096
189 |         35 - Relu
190 |         36 - Affine FC-1000
191 |         
192 |         37 - Softmax
193 |         
194 |                 
195 |     Inputs:
196 |     - X: The input to the starting layer. If start=0, then this should be an
197 |       array of shape (N, C, 64, 64).
198 |     - start: The index of the layer to start from. start=0 starts from the first
199 |       convolutional layer. Default is 0.
200 |     - end: The index of the layer to end at. start=11 ends at the last
201 |       fully-connected layer, returning class scores. Default is 11.
202 |     - mode: The mode to use, either 'test' or 'train'. We need this because
203 |       batch normalization behaves differently at training time and test time.
204 | 
205 |     Returns:
206 |     - out: Output from the end layer.
207 |     - cache: A cache object that can be passed to the backward method to run the
208 |       network backward over the same range of layers.
209 |     """
210 |     max_pool_layers = [2,4,7,10,13]
211 |     X = X.astype(self.dtype)
212 |     if start is None: start = 0
213 |     if end is None: end = len(self.conv_params) + 2
214 |     layer_caches = []
215 |     pool_params = {'stride':2 , 'pool_height':2, 'pool_width':2 }
216 |     prev_a = X
217 |     for i in xrange(start, end + 1):
218 |       i1 = i + 1
219 |       if 0 <= i < len(self.conv_params):
220 |         # This is a conv layer
221 |         w, b = self.params['W%d' % i1], self.params['b%d' % i1]
222 |         conv_param = self.conv_params[i]
223 |         if((i+1) in max_pool_layers):
224 |             next_a, cache = conv_relu_pool_forward(prev_a, w, b, conv_param, pool_params)
225 |         else:
226 |             next_a, cache = conv_relu_forward(prev_a, w, b, conv_param)
227 |       elif i < len(self.conv_params) + 2:
228 |         # This is the fully-connected hidden layer
229 |         w, b = self.params['W%d' % i1], self.params['b%d' % i1]
230 |         prev_a = prev_a.reshape((-1,w.shape[0]))
231 |         next_a, cache = affine_relu_forward(prev_a, w, b)
232 |       elif i == len(self.conv_params) + 2:
233 |         # This is the last fully-connectsed layer that produces scores
234 |         w, b = self.params['W%d' % i1], self.params['b%d' % i1]
235 |         next_a, cache = affine_forward(prev_a, w, b)
236 |       else:
237 |         raise ValueError('Invalid layer index %d' % i)
238 | 
239 |       layer_caches.append(cache)
240 |       prev_a = next_a
241 | 
242 |     out = prev_a
243 |     cache = (start, end, layer_caches)
244 |     return out, cache
245 | 
246 | 
247 |   def backward(self, dout, cache):
248 |     """
249 |     Run the model backward over a sequence of layers that were previously run
250 |     forward using the self.forward method.
251 | 
252 |     Inputs:
253 |     - dout: Gradient with respect to the ending layer; this should have the same
254 |       shape as the out variable returned from the corresponding call to forward.
255 |     - cache: A cache object returned from self.forward.
256 | 
257 |     Returns:
258 |     - dX: Gradient with respect to the start layer. This will have the same
259 |       shape as the input X passed to self.forward.
260 |     - grads: Gradient of all parameters in the layers. For example if you run
261 |       forward through two convolutional layers, then on the corresponding call
262 |       to backward grads will contain the gradients with respect to the weights,
263 |       biases, and spatial batchnorm parameters of those two convolutional
264 |       layers. The grads dictionary will therefore contain a subset of the keys
265 |       of self.params, and grads[k] and self.params[k] will have the same shape.
266 |     """
267 |     start, end, layer_caches = cache
268 |     dnext_a = dout
269 |     grads = {}
270 |     max_pool_layers = [2,4,7,10,13]
271 | 
272 |     j = len(layer_caches) - 1
273 |     for i in reversed(range(start, end + 1)):
274 |       i1 = i + 1
275 |       if i == len(self.conv_params) + 2:
276 |         # This is the last fully-connected layer
277 |         dprev_a, dw, db = affine_backward(dnext_a, layer_caches[j])
278 |         grads['W%d' % i1] = dw
279 |         grads['b%d' % i1] = db
280 |       elif i >= len(self.conv_params):
281 |         # This is the fully-connected hidden layer
282 |         j = len(layer_caches) - 1
283 |         temp = affine_relu_backward(dnext_a, layer_caches[j])
284 |         dprev_a, dw, db = temp
285 |         if i ==  len(self.conv_params):
286 |           dprev_a = dprev_a.reshape((dprev_a.shape[0],512,7,7))
287 |         grads['W%d' % i1] = dw
288 |         grads['b%d' % i1] = db
289 |       elif 0 <= i < len(self.conv_params):
290 |         # This is a conv layer
291 |         if((i+1) in max_pool_layers):
292 |             temp = conv_relu_pool_backward(dnext_a,layer_caches[j] )
293 |         else:
294 |             temp = conv_relu_backward(dnext_a, layer_caches[j])
295 |         dprev_a, dw, db = temp
296 |         grads['W%d' % i1] = dw
297 |         grads['b%d' % i1] = db
298 |       else:
299 |         raise ValueError('Invalid layer index %d' % i)
300 |       dnext_a = dprev_a
301 |       j = j-1
302 | 
303 |     dX = dnext_a
304 |     return dX, grads
305 | 
306 | 
307 |   def loss(self, X, y=None):
308 |     """
309 |     Classification loss used to train the network.
310 | 
311 |     Inputs:
312 |     - X: Array of data, of shape (N, 3, 64, 64)
313 |     - y: Array of labels, of shape (N,)
314 | 
315 |     If y is None, then run a test-time forward pass and return:
316 |     - scores: Array of shape (N, 100) giving class scores.
317 | 
318 |     If y is not None, then run a training-time forward and backward pass and
319 |     return a tuple of:
320 |     - loss: Scalar giving loss
321 |     - grads: Dictionary of gradients, with the same keys as self.params.
322 |     """
323 |     # Note that we implement this by just caling self.forward and self.backward
324 |     mode = 'test' if y is None else 'train'
325 |     scores, cache = self.forward(X, mode=mode)
326 |     if mode == 'test':
327 |       return scores
328 |     loss, dscores = softmax_loss(scores, y)
329 |     dX, grads = self.backward(dscores, cache)
330 |     return loss, grads
331 | 
332 | 


--------------------------------------------------------------------------------
/library/data_utils.py:
--------------------------------------------------------------------------------
  1 | import cPickle as pickle
  2 | import numpy as np
  3 | import os
  4 | from scipy.misc import imread
  5 | 
  6 | def load_CIFAR_batch(filename):
  7 |   """ load single batch of cifar """
  8 |   with open(filename, 'rb') as f:
  9 |     datadict = pickle.load(f)
 10 |     X = datadict['data']
 11 |     Y = datadict['labels']
 12 |     X = X.reshape(10000, 3, 32, 32).transpose(0,2,3,1).astype("float")
 13 |     Y = np.array(Y)
 14 |     return X, Y
 15 | 
 16 | def load_CIFAR10(ROOT):
 17 |   """ load all of cifar """
 18 |   xs = []
 19 |   ys = []
 20 |   for b in range(1,6):
 21 |     f = os.path.join(ROOT, 'data_batch_%d' % (b, ))
 22 |     X, Y = load_CIFAR_batch(f)
 23 |     xs.append(X)
 24 |     ys.append(Y)    
 25 |   Xtr = np.concatenate(xs)
 26 |   Ytr = np.concatenate(ys)
 27 |   del X, Y
 28 |   Xte, Yte = load_CIFAR_batch(os.path.join(ROOT, 'test_batch'))
 29 |   return Xtr, Ytr, Xte, Yte
 30 | 
 31 | 
 32 | def get_CIFAR10_data(num_training=49000, num_validation=1000, num_test=1000,
 33 |                      subtract_mean=True):
 34 |     """
 35 |     Load the CIFAR-10 dataset from disk and perform preprocessing to prepare
 36 |     it for classifiers. These are the same steps as we used for the SVM, but
 37 |     condensed to a single function.
 38 |     """
 39 |     # Load the raw CIFAR-10 data
 40 |     cifar10_dir = 'cs231n/datasets/cifar-10-batches-py'
 41 |     X_train, y_train, X_test, y_test = load_CIFAR10(cifar10_dir)
 42 |         
 43 |     # Subsample the data
 44 |     mask = range(num_training, num_training + num_validation)
 45 |     X_val = X_train[mask]
 46 |     y_val = y_train[mask]
 47 |     mask = range(num_training)
 48 |     X_train = X_train[mask]
 49 |     y_train = y_train[mask]
 50 |     mask = range(num_test)
 51 |     X_test = X_test[mask]
 52 |     y_test = y_test[mask]
 53 | 
 54 |     # Normalize the data: subtract the mean image
 55 |     if subtract_mean:
 56 |       mean_image = np.mean(X_train, axis=0)
 57 |       X_train -= mean_image
 58 |       X_val -= mean_image
 59 |       X_test -= mean_image
 60 |     
 61 |     # Transpose so that channels come first
 62 |     X_train = X_train.transpose(0, 3, 1, 2).copy()
 63 |     X_val = X_val.transpose(0, 3, 1, 2).copy()
 64 |     X_test = X_test.transpose(0, 3, 1, 2).copy()
 65 | 
 66 |     # Package data into a dictionary
 67 |     return {
 68 |       'X_train': X_train, 'y_train': y_train,
 69 |       'X_val': X_val, 'y_val': y_val,
 70 |       'X_test': X_test, 'y_test': y_test,
 71 |     }
 72 |     
 73 | 
 74 | def load_tiny_imagenet(path, dtype=np.float32, subtract_mean=True):
 75 |   """
 76 |   Load TinyImageNet. Each of TinyImageNet-100-A, TinyImageNet-100-B, and
 77 |   TinyImageNet-200 have the same directory structure, so this can be used
 78 |   to load any of them.
 79 | 
 80 |   Inputs:
 81 |   - path: String giving path to the directory to load.
 82 |   - dtype: numpy datatype used to load the data.
 83 |   - subtract_mean: Whether to subtract the mean training image.
 84 | 
 85 |   Returns: A dictionary with the following entries:
 86 |   - class_names: A list where class_names[i] is a list of strings giving the
 87 |     WordNet names for class i in the loaded dataset.
 88 |   - X_train: (N_tr, 3, 64, 64) array of training images
 89 |   - y_train: (N_tr,) array of training labels
 90 |   - X_val: (N_val, 3, 64, 64) array of validation images
 91 |   - y_val: (N_val,) array of validation labels
 92 |   - X_test: (N_test, 3, 64, 64) array of testing images.
 93 |   - y_test: (N_test,) array of test labels; if test labels are not available
 94 |     (such as in student code) then y_test will be None.
 95 |   - mean_image: (3, 64, 64) array giving mean training image
 96 |   """
 97 |   # First load wnids
 98 |   with open(os.path.join(path, 'wnids.txt'), 'r') as f:
 99 |     wnids = [x.strip() for x in f]
100 | 
101 |   # Map wnids to integer labels
102 |   wnid_to_label = {wnid: i for i, wnid in enumerate(wnids)}
103 | 
104 |   # Use words.txt to get names for each class
105 |   with open(os.path.join(path, 'words.txt'), 'r') as f:
106 |     wnid_to_words = dict(line.split('\t') for line in f)
107 |     for wnid, words in wnid_to_words.iteritems():
108 |       wnid_to_words[wnid] = [w.strip() for w in words.split(',')]
109 |   class_names = [wnid_to_words[wnid] for wnid in wnids]
110 | 
111 |   # Next load training data.
112 |   X_train = []
113 |   y_train = []
114 |   for i, wnid in enumerate(wnids):
115 |     if (i + 1) % 20 == 0:
116 |       print 'loading training data for synset %d / %d' % (i + 1, len(wnids))
117 |     # To figure out the filenames we need to open the boxes file
118 |     boxes_file = os.path.join(path, 'train', wnid, '%s_boxes.txt' % wnid)
119 |     with open(boxes_file, 'r') as f:
120 |       filenames = [x.split('\t')[0] for x in f]
121 |     num_images = len(filenames)
122 |     
123 |     X_train_block = np.zeros((num_images, 3, 64, 64), dtype=dtype)
124 |     y_train_block = wnid_to_label[wnid] * np.ones(num_images, dtype=np.int64)
125 |     for j, img_file in enumerate(filenames):
126 |       img_file = os.path.join(path, 'train', wnid, 'images', img_file)
127 |       img = imread(img_file)
128 |       if img.ndim == 2:
129 |         ## grayscale file
130 |         img.shape = (64, 64, 1)
131 |       X_train_block[j] = img.transpose(2, 0, 1)
132 |     X_train.append(X_train_block)
133 |     y_train.append(y_train_block)
134 |       
135 |   # We need to concatenate all training data
136 |   X_train = np.concatenate(X_train, axis=0)
137 |   y_train = np.concatenate(y_train, axis=0)
138 |   
139 |   # Next load validation data
140 |   with open(os.path.join(path, 'val', 'val_annotations.txt'), 'r') as f:
141 |     img_files = []
142 |     val_wnids = []
143 |     for line in f:
144 |       img_file, wnid = line.split('\t')[:2]
145 |       img_files.append(img_file)
146 |       val_wnids.append(wnid)
147 |     num_val = len(img_files)
148 |     y_val = np.array([wnid_to_label[wnid] for wnid in val_wnids])
149 |     X_val = np.zeros((num_val, 3, 64, 64), dtype=dtype)
150 |     for i, img_file in enumerate(img_files):
151 |       img_file = os.path.join(path, 'val', 'images', img_file)
152 |       img = imread(img_file)
153 |       if img.ndim == 2:
154 |         img.shape = (64, 64, 1)
155 |       X_val[i] = img.transpose(2, 0, 1)
156 | 
157 |   # Next load test images
158 |   # Students won't have test labels, so we need to iterate over files in the
159 |   # images directory.
160 |   img_files = os.listdir(os.path.join(path, 'test', 'images'))
161 |   X_test = np.zeros((len(img_files), 3, 64, 64), dtype=dtype)
162 |   for i, img_file in enumerate(img_files):
163 |     img_file = os.path.join(path, 'test', 'images', img_file)
164 |     img = imread(img_file)
165 |     if img.ndim == 2:
166 |       img.shape = (64, 64, 1)
167 |     X_test[i] = img.transpose(2, 0, 1)
168 | 
169 |   y_test = None
170 |   y_test_file = os.path.join(path, 'test', 'test_annotations.txt')
171 |   if os.path.isfile(y_test_file):
172 |     with open(y_test_file, 'r') as f:
173 |       img_file_to_wnid = {}
174 |       for line in f:
175 |         line = line.split('\t')
176 |         img_file_to_wnid[line[0]] = line[1]
177 |     y_test = [wnid_to_label[img_file_to_wnid[img_file]] for img_file in img_files]
178 |     y_test = np.array(y_test)
179 |   
180 |   mean_image = X_train.mean(axis=0)
181 |   if subtract_mean:
182 |     X_train -= mean_image[None]
183 |     X_val -= mean_image[None]
184 |     X_test -= mean_image[None]
185 | 
186 |   return {
187 |     'class_names': class_names,
188 |     'X_train': X_train,
189 |     'y_train': y_train,
190 |     'X_val': X_val,
191 |     'y_val': y_val,
192 |     'X_test': X_test,
193 |     'y_test': y_test,
194 |     'class_names': class_names,
195 |     'mean_image': mean_image,
196 |   }
197 | 
198 | 
199 | def load_models(models_dir):
200 |   """
201 |   Load saved models from disk. This will attempt to unpickle all files in a
202 |   directory; any files that give errors on unpickling (such as README.txt) will
203 |   be skipped.
204 | 
205 |   Inputs:
206 |   - models_dir: String giving the path to a directory containing model files.
207 |     Each model file is a pickled dictionary with a 'model' field.
208 | 
209 |   Returns:
210 |   A dictionary mapping model file names to models.
211 |   """
212 |   models = {}
213 |   for model_file in os.listdir(models_dir):
214 |     with open(os.path.join(models_dir, model_file), 'rb') as f:
215 |       try:
216 |         models[model_file] = pickle.load(f)['model']
217 |       except pickle.UnpicklingError:
218 |         continue
219 |   return models
220 | 


--------------------------------------------------------------------------------
/library/fast_layers.py:
--------------------------------------------------------------------------------
  1 | import numpy as np
  2 | try:
  3 |   from library.im2col_cython import col2im_cython, im2col_cython
  4 |   from library.im2col_cython import col2im_6d_cython
  5 | except ImportError:
  6 |   print 'run the following from the cs231n directory and try again:'
  7 |   print 'python setup.py build_ext --inplace'
  8 |   print 'You may also need to restart your iPython kernel'
  9 | 
 10 | from library.im2col import *
 11 | 
 12 | 
 13 | def conv_forward_im2col(x, w, b, conv_param):
 14 |   """
 15 |   A fast implementation of the forward pass for a convolutional layer
 16 |   based on im2col and col2im.
 17 |   """
 18 |   N, C, H, W = x.shape
 19 |   num_filters, _, filter_height, filter_width = w.shape
 20 |   stride, pad = conv_param['stride'], conv_param['pad']
 21 | 
 22 |   # Check dimensions
 23 |   assert (W + 2 * pad - filter_width) % stride == 0, 'width does not work'
 24 |   assert (H + 2 * pad - filter_height) % stride == 0, 'height does not work'
 25 | 
 26 |   # Create output
 27 |   out_height = (H + 2 * pad - filter_height) / stride + 1
 28 |   out_width = (W + 2 * pad - filter_width) / stride + 1
 29 |   out = np.zeros((N, num_filters, out_height, out_width), dtype=x.dtype)
 30 | 
 31 |   # x_cols = im2col_indices(x, w.shape[2], w.shape[3], pad, stride)
 32 |   x_cols = im2col_cython(x, w.shape[2], w.shape[3], pad, stride)
 33 |   res = w.reshape((w.shape[0], -1)).dot(x_cols) + b.reshape(-1, 1)
 34 | 
 35 |   out = res.reshape(w.shape[0], out.shape[2], out.shape[3], x.shape[0])
 36 |   out = out.transpose(3, 0, 1, 2)
 37 | 
 38 |   cache = (x, w, b, conv_param, x_cols)
 39 |   return out, cache
 40 | 
 41 | 
 42 | def conv_forward_strides(x, w, b, conv_param):
 43 |   N, C, H, W = x.shape
 44 |   F, _, HH, WW = w.shape
 45 |   stride, pad = conv_param['stride'], conv_param['pad']
 46 | 
 47 |   # Check dimensions
 48 |   #assert (W + 2 * pad - WW) % stride == 0, 'width does not work'
 49 |   #assert (H + 2 * pad - HH) % stride == 0, 'height does not work'
 50 | 
 51 |   # Pad the input
 52 |   p = pad
 53 |   x_padded = np.pad(x, ((0, 0), (0, 0), (p, p), (p, p)), mode='constant')
 54 |   
 55 |   # Figure out output dimensions
 56 |   H += 2 * pad
 57 |   W += 2 * pad
 58 |   out_h = (H - HH) / stride + 1
 59 |   out_w = (W - WW) / stride + 1
 60 | 
 61 |   # Perform an im2col operation by picking clever strides
 62 |   shape = (C, HH, WW, N, out_h, out_w)
 63 |   strides = (H * W, W, 1, C * H * W, stride * W, stride)
 64 |   strides = x.itemsize * np.array(strides)
 65 |   x_stride = np.lib.stride_tricks.as_strided(x_padded,
 66 |                 shape=shape, strides=strides)
 67 |   x_cols = np.ascontiguousarray(x_stride)
 68 |   x_cols.shape = (C * HH * WW, N * out_h * out_w)
 69 | 
 70 |   # Now all our convolutions are a big matrix multiply
 71 |   res = w.reshape(F, -1).dot(x_cols) + b.reshape(-1, 1)
 72 | 
 73 |   # Reshape the output
 74 |   res.shape = (F, N, out_h, out_w)
 75 |   out = res.transpose(1, 0, 2, 3)
 76 | 
 77 |   # Be nice and return a contiguous array
 78 |   # The old version of conv_forward_fast doesn't do this, so for a fair
 79 |   # comparison we won't either
 80 |   out = np.ascontiguousarray(out)
 81 | 
 82 |   cache = (x, w, b, conv_param, x_cols)
 83 |   return out, cache
 84 |   
 85 | 
 86 | def conv_backward_strides(dout, cache):
 87 |   x, w, b, conv_param, x_cols = cache
 88 |   stride, pad = conv_param['stride'], conv_param['pad']
 89 | 
 90 |   N, C, H, W = x.shape
 91 |   F, _, HH, WW = w.shape
 92 |   _, _, out_h, out_w = dout.shape
 93 | 
 94 |   db = np.sum(dout, axis=(0, 2, 3))
 95 | 
 96 |   dout_reshaped = dout.transpose(1, 0, 2, 3).reshape(F, -1)
 97 |   dw = dout_reshaped.dot(x_cols.T).reshape(w.shape)
 98 | 
 99 |   dx_cols = w.reshape(F, -1).T.dot(dout_reshaped)
100 |   dx_cols.shape = (C, HH, WW, N, out_h, out_w)
101 |   dx = col2im_6d_cython(dx_cols, N, C, H, W, HH, WW, pad, stride)
102 | 
103 |   return dx, dw, db
104 | 
105 | 
106 | def conv_backward_im2col(dout, cache):
107 |   """
108 |   A fast implementation of the backward pass for a convolutional layer
109 |   based on im2col and col2im.
110 |   """
111 |   x, w, b, conv_param, x_cols = cache
112 |   stride, pad = conv_param['stride'], conv_param['pad']
113 | 
114 |   db = np.sum(dout, axis=(0, 2, 3))
115 | 
116 |   num_filters, _, filter_height, filter_width = w.shape
117 |   dout_reshaped = dout.transpose(1, 2, 3, 0).reshape(num_filters, -1)
118 |   dw = dout_reshaped.dot(x_cols.T).reshape(w.shape)
119 | 
120 |   dx_cols = w.reshape(num_filters, -1).T.dot(dout_reshaped)
121 |   # dx = col2im_indices(dx_cols, x.shape, filter_height, filter_width, pad, stride)
122 |   dx = col2im_cython(dx_cols, x.shape[0], x.shape[1], x.shape[2], x.shape[3],
123 |                      filter_height, filter_width, pad, stride)
124 | 
125 |   return dx, dw, db
126 | 
127 | 
128 | conv_forward_fast = conv_forward_strides
129 | conv_backward_fast = conv_backward_strides
130 | 
131 | 
132 | def max_pool_forward_fast(x, pool_param):
133 |   """
134 |   A fast implementation of the forward pass for a max pooling layer.
135 | 
136 |   This chooses between the reshape method and the im2col method. If the pooling
137 |   regions are square and tile the input image, then we can use the reshape
138 |   method which is very fast. Otherwise we fall back on the im2col method, which
139 |   is not much faster than the naive method.
140 |   """
141 |   N, C, H, W = x.shape
142 |   pool_height, pool_width = pool_param['pool_height'], pool_param['pool_width']
143 |   stride = pool_param['stride']
144 | 
145 |   same_size = pool_height == pool_width == stride
146 |   tiles = H % pool_height == 0 and W % pool_width == 0
147 |   if same_size and tiles:
148 |     out, reshape_cache = max_pool_forward_reshape(x, pool_param)
149 |     cache = ('reshape', reshape_cache)
150 |   else:
151 |     out, im2col_cache = max_pool_forward_im2col(x, pool_param)
152 |     cache = ('im2col', im2col_cache)
153 |   return out, cache
154 | 
155 | 
156 | def max_pool_backward_fast(dout, cache):
157 |   """
158 |   A fast implementation of the backward pass for a max pooling layer.
159 | 
160 |   This switches between the reshape method an the im2col method depending on
161 |   which method was used to generate the cache.
162 |   """
163 |   method, real_cache = cache
164 |   if method == 'reshape':
165 |     return max_pool_backward_reshape(dout, real_cache)
166 |   elif method == 'im2col':
167 |     return max_pool_backward_im2col(dout, real_cache)
168 |   else:
169 |     raise ValueError('Unrecognized method "%s"' % method)
170 | 
171 | 
172 | def max_pool_forward_reshape(x, pool_param):
173 |   """
174 |   A fast implementation of the forward pass for the max pooling layer that uses
175 |   some clever reshaping.
176 | 
177 |   This can only be used for square pooling regions that tile the input.
178 |   """
179 |   N, C, H, W = x.shape
180 |   pool_height, pool_width = pool_param['pool_height'], pool_param['pool_width']
181 |   stride = pool_param['stride']
182 |   assert pool_height == pool_width == stride, 'Invalid pool params'
183 |   assert H % pool_height == 0
184 |   assert W % pool_height == 0
185 |   x_reshaped = x.reshape(N, C, H / pool_height, pool_height,
186 |                          W / pool_width, pool_width)
187 |   out = x_reshaped.max(axis=3).max(axis=4)
188 | 
189 |   cache = (x, x_reshaped, out)
190 |   return out, cache
191 | 
192 | 
193 | def max_pool_backward_reshape(dout, cache):
194 |   """
195 |   A fast implementation of the backward pass for the max pooling layer that
196 |   uses some clever broadcasting and reshaping.
197 | 
198 |   This can only be used if the forward pass was computed using
199 |   max_pool_forward_reshape.
200 | 
201 |   NOTE: If there are multiple argmaxes, this method will assign gradient to
202 |   ALL argmax elements of the input rather than picking one. In this case the
203 |   gradient will actually be incorrect. However this is unlikely to occur in
204 |   practice, so it shouldn't matter much. One possible solution is to split the
205 |   upstream gradient equally among all argmax elements; this should result in a
206 |   valid subgradient. You can make this happen by uncommenting the line below;
207 |   however this results in a significant performance penalty (about 40% slower)
208 |   and is unlikely to matter in practice so we don't do it.
209 |   """
210 |   x, x_reshaped, out = cache
211 | 
212 |   dx_reshaped = np.zeros_like(x_reshaped)
213 |   out_newaxis = out[:, :, :, np.newaxis, :, np.newaxis]
214 |   mask = (x_reshaped == out_newaxis)
215 |   dout_newaxis = dout[:, :, :, np.newaxis, :, np.newaxis]
216 |   dout_broadcast, _ = np.broadcast_arrays(dout_newaxis, dx_reshaped)
217 |   dx_reshaped[mask] = dout_broadcast[mask]
218 |   dx_reshaped /= np.sum(mask, axis=(3, 5), keepdims=True)
219 |   dx = dx_reshaped.reshape(x.shape)
220 | 
221 |   return dx
222 | 
223 | 
224 | def max_pool_forward_im2col(x, pool_param):
225 |   """
226 |   An implementation of the forward pass for max pooling based on im2col.
227 | 
228 |   This isn't much faster than the naive version, so it should be avoided if
229 |   possible.
230 |   """
231 |   N, C, H, W = x.shape
232 |   pool_height, pool_width = pool_param['pool_height'], pool_param['pool_width']
233 |   stride = pool_param['stride']
234 | 
235 |   assert (H - pool_height) % stride == 0, 'Invalid height'
236 |   assert (W - pool_width) % stride == 0, 'Invalid width'
237 | 
238 |   out_height = (H - pool_height) / stride + 1
239 |   out_width = (W - pool_width) / stride + 1
240 | 
241 |   x_split = x.reshape(N * C, 1, H, W)
242 |   x_cols = im2col(x_split, pool_height, pool_width, padding=0, stride=stride)
243 |   x_cols_argmax = np.argmax(x_cols, axis=0)
244 |   x_cols_max = x_cols[x_cols_argmax, np.arange(x_cols.shape[1])]
245 |   out = x_cols_max.reshape(out_height, out_width, N, C).transpose(2, 3, 0, 1)
246 | 
247 |   cache = (x, x_cols, x_cols_argmax, pool_param)
248 |   return out, cache
249 | 
250 | 
251 | def max_pool_backward_im2col(dout, cache):
252 |   """
253 |   An implementation of the backward pass for max pooling based on im2col.
254 | 
255 |   This isn't much faster than the naive version, so it should be avoided if
256 |   possible.
257 |   """
258 |   x, x_cols, x_cols_argmax, pool_param = cache
259 |   N, C, H, W = x.shape
260 |   pool_height, pool_width = pool_param['pool_height'], pool_param['pool_width']
261 |   stride = pool_param['stride']
262 | 
263 |   dout_reshaped = dout.transpose(2, 3, 0, 1).flatten()
264 |   dx_cols = np.zeros_like(x_cols)
265 |   dx_cols[x_cols_argmax, np.arange(dx_cols.shape[1])] = dout_reshaped
266 |   dx = col2im_indices(dx_cols, (N * C, 1, H, W), pool_height, pool_width,
267 |               padding=0, stride=stride)
268 |   dx = dx.reshape(x.shape)
269 | 
270 |   return dx
271 | 


--------------------------------------------------------------------------------
/library/frameworkpython:
--------------------------------------------------------------------------------
 1 | #!/bin/bash
 2 | 
 3 | # what real Python executable to use
 4 | PYVER=2.7
 5 | PATHTOPYTHON=/usr/local/bin/
 6 | PYTHON=${PATHTOPYTHON}python${PYVER}
 7 | 
 8 | # find the root of the virtualenv, it should be the parent of the dir this script is in
 9 | ENV=`$PYTHON -c "import os; print os.path.abspath(os.path.join(os.path.dirname(\"$0\"), '..'))"`
10 | 
11 | # now run Python with the virtualenv set as Python's HOME
12 | export PYTHONHOME=$ENV
13 | exec $PYTHON "$@"
14 | 


--------------------------------------------------------------------------------
/library/im2col.py:
--------------------------------------------------------------------------------
 1 | import numpy as np
 2 | 
 3 | 
 4 | def get_im2col_indices(x_shape, field_height, field_width, padding=1, stride=1):
 5 |   # First figure out what the size of the output should be
 6 |   N, C, H, W = x_shape
 7 |   assert (H + 2 * padding - field_height) % stride == 0
 8 |   assert (W + 2 * padding - field_height) % stride == 0
 9 |   out_height = (H + 2 * padding - field_height) / stride + 1
10 |   out_width = (W + 2 * padding - field_width) / stride + 1
11 | 
12 |   i0 = np.repeat(np.arange(field_height), field_width)
13 |   i0 = np.tile(i0, C)
14 |   i1 = stride * np.repeat(np.arange(out_height), out_width)
15 |   j0 = np.tile(np.arange(field_width), field_height * C)
16 |   j1 = stride * np.tile(np.arange(out_width), out_height)
17 |   i = i0.reshape(-1, 1) + i1.reshape(1, -1)
18 |   j = j0.reshape(-1, 1) + j1.reshape(1, -1)
19 | 
20 |   k = np.repeat(np.arange(C), field_height * field_width).reshape(-1, 1)
21 | 
22 |   return (k, i, j)
23 | 
24 | 
25 | def im2col_indices(x, field_height, field_width, padding=1, stride=1):
26 |   """ An implementation of im2col based on some fancy indexing """
27 |   # Zero-pad the input
28 |   p = padding
29 |   x_padded = np.pad(x, ((0, 0), (0, 0), (p, p), (p, p)), mode='constant')
30 | 
31 |   k, i, j = get_im2col_indices(x.shape, field_height, field_width, padding,
32 |                                stride)
33 | 
34 |   cols = x_padded[:, k, i, j]
35 |   C = x.shape[1]
36 |   cols = cols.transpose(1, 2, 0).reshape(field_height * field_width * C, -1)
37 |   return cols
38 | 
39 | 
40 | def col2im_indices(cols, x_shape, field_height=3, field_width=3, padding=1,
41 |                    stride=1):
42 |   """ An implementation of col2im based on fancy indexing and np.add.at """
43 |   N, C, H, W = x_shape
44 |   H_padded, W_padded = H + 2 * padding, W + 2 * padding
45 |   x_padded = np.zeros((N, C, H_padded, W_padded), dtype=cols.dtype)
46 |   k, i, j = get_im2col_indices(x_shape, field_height, field_width, padding,
47 |                                stride)
48 |   cols_reshaped = cols.reshape(C * field_height * field_width, -1, N)
49 |   cols_reshaped = cols_reshaped.transpose(2, 0, 1)
50 |   np.add.at(x_padded, (slice(None), k, i, j), cols_reshaped)
51 |   if padding == 0:
52 |     return x_padded
53 |   return x_padded[:, :, padding:-padding, padding:-padding]
54 | 
55 | pass
56 | 


--------------------------------------------------------------------------------
/library/im2col_cython.pyx:
--------------------------------------------------------------------------------
  1 | import numpy as np
  2 | cimport numpy as np
  3 | cimport cython
  4 | 
  5 | # DTYPE = np.float64
  6 | # ctypedef np.float64_t DTYPE_t
  7 | 
  8 | ctypedef fused DTYPE_t:
  9 |     np.float32_t
 10 |     np.float64_t
 11 | 
 12 | def im2col_cython(np.ndarray[DTYPE_t, ndim=4] x, int field_height,
 13 |                   int field_width, int padding, int stride):
 14 |     cdef int N = x.shape[0]
 15 |     cdef int C = x.shape[1]
 16 |     cdef int H = x.shape[2]
 17 |     cdef int W = x.shape[3]
 18 |     
 19 |     cdef int HH = (H + 2 * padding - field_height) / stride + 1
 20 |     cdef int WW = (W + 2 * padding - field_width) / stride + 1
 21 | 
 22 |     cdef int p = padding
 23 |     cdef np.ndarray[DTYPE_t, ndim=4] x_padded = np.pad(x,
 24 |             ((0, 0), (0, 0), (p, p), (p, p)), mode='constant')
 25 | 
 26 |     cdef np.ndarray[DTYPE_t, ndim=2] cols = np.zeros(
 27 |             (C * field_height * field_width, N * HH * WW),
 28 |             dtype=x.dtype)
 29 | 
 30 |     # Moving the inner loop to a C function with no bounds checking works, but does
 31 |     # not seem to help performance in any measurable way.
 32 | 
 33 |     im2col_cython_inner(cols, x_padded, N, C, H, W, HH, WW,
 34 |                         field_height, field_width, padding, stride)
 35 |     return cols
 36 | 
 37 | 
 38 | @cython.boundscheck(False)
 39 | cdef int im2col_cython_inner(np.ndarray[DTYPE_t, ndim=2] cols,
 40 |                              np.ndarray[DTYPE_t, ndim=4] x_padded,
 41 |                              int N, int C, int H, int W, int HH, int WW,
 42 |                              int field_height, int field_width, int padding, int stride) except? -1:
 43 |     cdef int c, ii, jj, row, yy, xx, i, col
 44 | 
 45 |     for c in range(C):
 46 |         for yy in range(HH):
 47 |             for xx in range(WW):
 48 |                 for ii in range(field_height):
 49 |                     for jj in range(field_width):
 50 |                         row = c * field_width * field_height + ii * field_height + jj
 51 |                         for i in range(N):
 52 |                             col = yy * WW * N + xx * N + i
 53 |                             cols[row, col] = x_padded[i, c, stride * yy + ii, stride * xx + jj]
 54 | 
 55 | 
 56 | 
 57 | def col2im_cython(np.ndarray[DTYPE_t, ndim=2] cols, int N, int C, int H, int W,
 58 |                   int field_height, int field_width, int padding, int stride):
 59 |     cdef np.ndarray x = np.empty((N, C, H, W), dtype=cols.dtype)
 60 |     cdef int HH = (H + 2 * padding - field_height) / stride + 1
 61 |     cdef int WW = (W + 2 * padding - field_width) / stride + 1
 62 |     cdef np.ndarray[DTYPE_t, ndim=4] x_padded = np.zeros((N, C, H + 2 * padding, W + 2 * padding),
 63 |                                         dtype=cols.dtype)
 64 | 
 65 |     # Moving the inner loop to a C-function with no bounds checking improves
 66 |     # performance quite a bit for col2im.
 67 |     col2im_cython_inner(cols, x_padded, N, C, H, W, HH, WW, 
 68 |                         field_height, field_width, padding, stride)
 69 |     if padding > 0:
 70 |         return x_padded[:, :, padding:-padding, padding:-padding]
 71 |     return x_padded
 72 | 
 73 | 
 74 | @cython.boundscheck(False)
 75 | cdef int col2im_cython_inner(np.ndarray[DTYPE_t, ndim=2] cols,
 76 |                              np.ndarray[DTYPE_t, ndim=4] x_padded,
 77 |                              int N, int C, int H, int W, int HH, int WW,
 78 |                              int field_height, int field_width, int padding, int stride) except? -1:
 79 |     cdef int c, ii, jj, row, yy, xx, i, col
 80 | 
 81 |     for c in range(C):
 82 |         for ii in range(field_height):
 83 |             for jj in range(field_width):
 84 |                 row = c * field_width * field_height + ii * field_height + jj
 85 |                 for yy in range(HH):
 86 |                     for xx in range(WW):
 87 |                         for i in range(N):
 88 |                             col = yy * WW * N + xx * N + i
 89 |                             x_padded[i, c, stride * yy + ii, stride * xx + jj] += cols[row, col]
 90 | 
 91 | 
 92 | @cython.boundscheck(False)
 93 | @cython.wraparound(False)
 94 | cdef col2im_6d_cython_inner(np.ndarray[DTYPE_t, ndim=6] cols,
 95 |                             np.ndarray[DTYPE_t, ndim=4] x_padded,
 96 |                             int N, int C, int H, int W, int HH, int WW,
 97 |                             int out_h, int out_w, int pad, int stride):
 98 | 
 99 |     cdef int c, hh, ww, n, h, w
100 |     for n in range(N):
101 |         for c in range(C):
102 |             for hh in range(HH):
103 |                 for ww in range(WW):
104 |                     for h in range(out_h):
105 |                         for w in range(out_w):
106 |                             x_padded[n, c, stride * h + hh, stride * w + ww] += cols[c, hh, ww, n, h, w]
107 |     
108 | 
109 | def col2im_6d_cython(np.ndarray[DTYPE_t, ndim=6] cols, int N, int C, int H, int W,
110 |         int HH, int WW, int pad, int stride):
111 |     cdef np.ndarray x = np.empty((N, C, H, W), dtype=cols.dtype)
112 |     cdef int out_h = (H + 2 * pad - HH) / stride + 1
113 |     cdef int out_w = (W + 2 * pad - WW) / stride + 1
114 |     cdef np.ndarray[DTYPE_t, ndim=4] x_padded = np.zeros((N, C, H + 2 * pad, W + 2 * pad),
115 |                                                   dtype=cols.dtype)
116 | 
117 |     col2im_6d_cython_inner(cols, x_padded, N, C, H, W, HH, WW, out_h, out_w, pad, stride)
118 | 
119 |     if pad > 0:
120 |         return x_padded[:, :, pad:-pad, pad:-pad]
121 |     return x_padded 
122 | 


--------------------------------------------------------------------------------
/library/image_utils.py:
--------------------------------------------------------------------------------
 1 | import urllib2, os, tempfile
 2 | import sys
 3 | import cv2
 4 | from socket import error as SocketError
 5 | 
 6 | def image_from_url(url):
 7 |   """
 8 |   Read an image from a URL. Returns a numpy array with the pixel data.
 9 |   We write the image to a temporary file then read it back. Kinda gross.
10 |   """
11 |   try:
12 |     f = urllib2.urlopen(url)
13 |     _, fname = tempfile.mkstemp()
14 |     with open(fname, 'wb') as ff:
15 |       ff.write(f.read())
16 |     img = cv2.imread(fname)
17 |     os.remove(fname)
18 |     return img
19 |   except urllib2.URLError as e:
20 |     print 'URL Error: ', e.reason, url
21 |   except urllib2.HTTPError as e:
22 |     print 'HTTP Error: ', e.code, url
23 |   except ValueError as e:
24 |     print 'Value Error: ', url
25 |   except SocketError as e:
26 |     print 'Socket Error: ', url
27 |   except httplib.BadStatusLine as e:
28 |     print 'Bad StatusLine Error: ', url
29 | 


--------------------------------------------------------------------------------
/library/layer_utils.py:
--------------------------------------------------------------------------------
  1 | from library.layers import *
  2 | from library.fast_layers import *
  3 | 
  4 | 
  5 | def affine_relu_forward(x, w, b):
  6 |   """
  7 |   Convenience layer that perorms an affine transform followed by a ReLU
  8 | 
  9 |   Inputs:
 10 |   - x: Input to the affine layer
 11 |   - w, b: Weights for the affine layer
 12 | 
 13 |   Returns a tuple of:
 14 |   - out: Output from the ReLU
 15 |   - cache: Object to give to the backward pass
 16 |   """
 17 |   a, fc_cache = affine_forward(x, w, b)
 18 |   out, relu_cache = relu_forward(a)
 19 |   cache = (fc_cache, relu_cache)
 20 |   return out, cache
 21 | 
 22 | 
 23 | def affine_relu_backward(dout, cache):
 24 |   """
 25 |   Backward pass for the affine-relu convenience layer
 26 |   """
 27 |   fc_cache, relu_cache = cache
 28 |   da = relu_backward(dout, relu_cache)
 29 |   dx, dw, db = affine_backward(da, fc_cache)
 30 |   return dx, dw, db
 31 | 
 32 | 
 33 | def affine_bn_relu_forward(x, w, b, gamma, beta, bn_param):
 34 |   """
 35 |   Convenience layer that performs an affine transform, batch normalization,
 36 |   and ReLU.
 37 | 
 38 |   Inputs:
 39 |   - x: Array of shape (N, D1); input to the affine layer
 40 |   - w, b: Arrays of shape (D2, D2) and (D2,) giving the weight and bias for
 41 |     the affine transform.
 42 |   - gamma, beta: Arrays of shape (D2,) and (D2,) giving scale and shift
 43 |     parameters for batch normalization.
 44 |   - bn_param: Dictionary of parameters for batch normalization.
 45 | 
 46 |   Returns:
 47 |   - out: Output from ReLU, of shape (N, D2)
 48 |   - cache: Object to give to the backward pass.
 49 |   """
 50 |   a, fc_cache = affine_forward(x, w, b)
 51 |   a_bn, bn_cache = batchnorm_forward(a, gamma, beta, bn_param)
 52 |   out, relu_cache = relu_forward(a_bn)
 53 |   cache = (fc_cache, bn_cache, relu_cache)
 54 |   return out, cache
 55 | 
 56 | 
 57 | def affine_bn_relu_backward(dout, cache):
 58 |   """
 59 |   Backward pass for the affine-batchnorm-relu convenience layer.
 60 |   """
 61 |   fc_cache, bn_cache, relu_cache = cache
 62 |   da_bn = relu_backward(dout, relu_cache)
 63 |   da, dgamma, dbeta = batchnorm_backward(da_bn, bn_cache)
 64 |   dx, dw, db = affine_backward(da, fc_cache)
 65 |   return dx, dw, db, dgamma, dbeta  
 66 | 
 67 | 
 68 | def conv_relu_forward(x, w, b, conv_param):
 69 |   """
 70 |   A convenience layer that performs a convolution followed by a ReLU.
 71 | 
 72 |   Inputs:
 73 |   - x: Input to the convolutional layer
 74 |   - w, b, conv_param: Weights and parameters for the convolutional layer
 75 |   
 76 |   Returns a tuple of:
 77 |   - out: Output from the ReLU
 78 |   - cache: Object to give to the backward pass
 79 |   """
 80 |   a, conv_cache = conv_forward_fast(x, w, b, conv_param)
 81 |   out, relu_cache = relu_forward(a)
 82 |   cache = (conv_cache, relu_cache)
 83 |   return out, cache
 84 | 
 85 | 
 86 | def conv_relu_backward(dout, cache):
 87 |   """
 88 |   Backward pass for the conv-relu convenience layer.
 89 |   """
 90 |   conv_cache, relu_cache = cache
 91 |   da = relu_backward(dout, relu_cache)
 92 |   dx, dw, db = conv_backward_fast(da, conv_cache)
 93 |   return dx, dw, db
 94 | 
 95 | 
 96 | def conv_bn_relu_forward(x, w, b, gamma, beta, conv_param, bn_param):
 97 |   a, conv_cache = conv_forward_fast(x, w, b, conv_param)
 98 |   an, bn_cache = spatial_batchnorm_forward(a, gamma, beta, bn_param)
 99 |   out, relu_cache = relu_forward(an)
100 |   cache = (conv_cache, bn_cache, relu_cache)
101 |   return out, cache
102 | 
103 | 
104 | def conv_bn_relu_backward(dout, cache):
105 |   conv_cache, bn_cache, relu_cache = cache
106 |   dan = relu_backward(dout, relu_cache)
107 |   da, dgamma, dbeta = spatial_batchnorm_backward(dan, bn_cache)
108 |   dx, dw, db = conv_backward_fast(da, conv_cache)
109 |   return dx, dw, db, dgamma, dbeta
110 | 
111 | 
112 | def conv_relu_pool_forward(x, w, b, conv_param, pool_param):
113 |   """
114 |   Convenience layer that performs a convolution, a ReLU, and a pool.
115 | 
116 |   Inputs:
117 |   - x: Input to the convolutional layer
118 |   - w, b, conv_param: Weights and parameters for the convolutional layer
119 |   - pool_param: Parameters for the pooling layer
120 | 
121 |   Returns a tuple of:
122 |   - out: Output from the pooling layer
123 |   - cache: Object to give to the backward pass
124 |   """
125 |   a, conv_cache = conv_forward_fast(x, w, b, conv_param)
126 |   s, relu_cache = relu_forward(a)
127 |   out, pool_cache = max_pool_forward_fast(s, pool_param)
128 |   cache = (conv_cache, relu_cache, pool_cache)
129 |   return out, cache
130 | 
131 | 
132 | def conv_relu_pool_backward(dout, cache):
133 |   """
134 |   Backward pass for the conv-relu-pool convenience layer
135 |   """
136 |   conv_cache, relu_cache, pool_cache = cache
137 |   #print pool_cache
138 |   ds = max_pool_backward_fast(dout, pool_cache)
139 |   da = relu_backward(ds, relu_cache)
140 |   dx, dw, db = conv_backward_fast(da, conv_cache)
141 |   return dx, dw, db
142 | 
143 | 


--------------------------------------------------------------------------------
/library/layers.py:
--------------------------------------------------------------------------------
  1 | import numpy as np
  2 | 
  3 | 
  4 | def affine_forward(x, w, b):
  5 |   """
  6 |   Computes the forward pass for an affine (fully-connected) layer.
  7 | 
  8 |   The input x has shape (N, d_1, ..., d_k) where x[i] is the ith input.
  9 |   We multiply this against a weight matrix of shape (D, M) where
 10 |   D = \prod_i d_i
 11 | 
 12 |   Inputs:
 13 |   x - Input data, of shape (N, d_1, ..., d_k)
 14 |   w - Weights, of shape (D, M)
 15 |   b - Biases, of shape (M,)
 16 |   
 17 |   Returns a tuple of:
 18 |   - out: output, of shape (N, M)
 19 |   - cache: (x, w, b)
 20 |   """
 21 |   out = x.reshape(x.shape[0], -1).dot(w) + b
 22 |   cache = (x, w, b)
 23 |   return out, cache
 24 | 
 25 | 
 26 | def affine_backward(dout, cache):
 27 |   """
 28 |   Computes the backward pass for an affine layer.
 29 | 
 30 |   Inputs:
 31 |   - dout: Upstream derivative, of shape (N, M)
 32 |   - cache: Tuple of:
 33 |     - x: Input data, of shape (N, d_1, ... d_k)
 34 |     - w: Weights, of shape (D, M)
 35 | 
 36 |   Returns a tuple of:
 37 |   - dx: Gradient with respect to x, of shape (N, d1, ..., d_k)
 38 |   - dw: Gradient with respect to w, of shape (D, M)
 39 |   - db: Gradient with respect to b, of shape (M,)
 40 |   """
 41 |   x, w, b = cache
 42 |   dx = dout.dot(w.T).reshape(x.shape)
 43 |   dw = x.reshape(x.shape[0], -1).T.dot(dout)
 44 |   db = np.sum(dout, axis=0)
 45 |   return dx, dw, db
 46 | 
 47 | 
 48 | def relu_forward(x):
 49 |   """
 50 |   Computes the forward pass for a layer of rectified linear units (ReLUs).
 51 | 
 52 |   Input:
 53 |   - x: Inputs, of any shape
 54 | 
 55 |   Returns a tuple of:
 56 |   - out: Output, of the same shape as x
 57 |   - cache: x
 58 |   """
 59 |   out = np.maximum(0, x)
 60 |   cache = x
 61 |   return out, cache
 62 | 
 63 | 
 64 | def relu_backward(dout, cache):
 65 |   """
 66 |   Computes the backward pass for a layer of rectified linear units (ReLUs).
 67 | 
 68 |   Input:
 69 |   - dout: Upstream derivatives, of any shape
 70 |   - cache: Input x, of same shape as dout
 71 | 
 72 |   Returns:
 73 |   - dx: Gradient with respect to x
 74 |   """
 75 |   x = cache
 76 |   dx = np.where(x > 0, dout, 0)
 77 |   return dx
 78 | 
 79 | 
 80 | def batchnorm_forward(x, gamma, beta, bn_param):
 81 |   """
 82 |   Forward pass for batch normalization.
 83 |   
 84 |   During training the sample mean and (uncorrected) sample variance are
 85 |   computed from minibatch statistics and used to normalize the incoming data.
 86 |   During training we also keep an exponentially decaying running mean of the mean
 87 |   and variance of each feature, and these averages are used to normalize data
 88 |   at test-time.
 89 | 
 90 |   At each timestep we update the running averages for mean and variance using
 91 |   an exponential decay based on the momentum parameter:
 92 | 
 93 |   running_mean = momentum * running_mean + (1 - momentum) * sample_mean
 94 |   running_var = momentum * running_var + (1 - momentum) * sample_var
 95 | 
 96 |   Note that the batch normalization paper suggests a different test-time
 97 |   behavior: they compute sample mean and variance for each feature using a
 98 |   large number of training images rather than using a running average. For
 99 |   this implementation we have chosen to use running averages instead since
100 |   they do not require an additional estimation step; the torch7 implementation
101 |   of batch normalization also uses running averages.
102 | 
103 |   Input:
104 |   - x: Data of shape (N, D)
105 |   - gamma: Scale parameter of shape (D,)
106 |   - beta: Shift paremeter of shape (D,)
107 |   - bn_param: Dictionary with the following keys:
108 |     - mode: 'train' or 'test'; required
109 |     - eps: Constant for numeric stability
110 |     - momentum: Constant for running mean / variance.
111 |     - running_mean: Array of shape (D,) giving running mean of features
112 |     - running_var Array of shape (D,) giving running variance of features
113 | 
114 |   Returns a tuple of:
115 |   - out: of shape (N, D)
116 |   - cache: A tuple of values needed in the backward pass
117 |   """
118 |   mode = bn_param['mode']
119 |   eps = bn_param.get('eps', 1e-5)
120 |   momentum = bn_param.get('momentum', 0.9)
121 | 
122 |   N, D = x.shape
123 |   running_mean = bn_param.get('running_mean', np.zeros(D, dtype=x.dtype))
124 |   running_var = bn_param.get('running_var', np.zeros(D, dtype=x.dtype))
125 | 
126 |   out, cache = None, None
127 |   if mode == 'train':
128 |     # Compute output
129 |     mu = x.mean(axis=0)
130 |     xc = x - mu
131 |     var = np.mean(xc ** 2, axis=0)
132 |     std = np.sqrt(var + eps)
133 |     xn = xc / std
134 |     out = gamma * xn + beta
135 | 
136 |     cache = (mode, x, gamma, xc, std, xn, out)
137 | 
138 |     # Update running average of mean
139 |     running_mean *= momentum
140 |     running_mean += (1 - momentum) * mu
141 | 
142 |     # Update running average of variance
143 |     running_var *= momentum
144 |     running_var += (1 - momentum) * var
145 |   elif mode == 'test':
146 |     # Using running mean and variance to normalize
147 |     std = np.sqrt(running_var + eps)
148 |     xn = (x - running_mean) / std
149 |     out = gamma * xn + beta
150 |     cache = (mode, x, xn, gamma, beta, std)
151 |   else:
152 |     raise ValueError('Invalid forward batchnorm mode "%s"' % mode)
153 | 
154 |   # Store the updated running means back into bn_param
155 |   bn_param['running_mean'] = running_mean
156 |   bn_param['running_var'] = running_var
157 | 
158 |   return out, cache
159 | 
160 | 
161 | def batchnorm_backward(dout, cache):
162 |   """
163 |   Backward pass for batch normalization.
164 |   
165 |   For this implementation, you should write out a computation graph for
166 |   batch normalization on paper and propagate gradients backward through
167 |   intermediate nodes.
168 |   
169 |   Inputs:
170 |   - dout: Upstream derivatives, of shape (N, D)
171 |   - cache: Variable of intermediates from batchnorm_forward.
172 |   
173 |   Returns a tuple of:
174 |   - dx: Gradient with respect to inputs x, of shape (N, D)
175 |   - dgamma: Gradient with respect to scale parameter gamma, of shape (D,)
176 |   - dbeta: Gradient with respect to shift parameter beta, of shape (D,)
177 |   """
178 |   mode = cache[0]
179 |   if mode == 'train':
180 |     mode, x, gamma, xc, std, xn, out = cache
181 | 
182 |     N = x.shape[0]
183 |     dbeta = dout.sum(axis=0)
184 |     dgamma = np.sum(xn * dout, axis=0)
185 |     dxn = gamma * dout
186 |     dxc = dxn / std
187 |     dstd = -np.sum((dxn * xc) / (std * std), axis=0)
188 |     dvar = 0.5 * dstd / std
189 |     dxc += (2.0 / N) * xc * dvar
190 |     dmu = np.sum(dxc, axis=0)
191 |     dx = dxc - dmu / N
192 |   elif mode == 'test':
193 |     mode, x, xn, gamma, beta, std = cache
194 |     dbeta = dout.sum(axis=0)
195 |     dgamma = np.sum(xn * dout, axis=0)
196 |     dxn = gamma * dout
197 |     dx = dxn / std
198 |   else:
199 |     raise ValueError(mode)
200 | 
201 |   return dx, dgamma, dbeta
202 | 
203 | 
204 | def spatial_batchnorm_forward(x, gamma, beta, bn_param):
205 |   """
206 |   Computes the forward pass for spatial batch normalization.
207 |   
208 |   Inputs:
209 |   - x: Input data of shape (N, C, H, W)
210 |   - gamma: Scale parameter, of shape (C,)
211 |   - beta: Shift parameter, of shape (C,)
212 |   - bn_param: Dictionary with the following keys:
213 |     - mode: 'train' or 'test'; required
214 |     - eps: Constant for numeric stability
215 |     - momentum: Constant for running mean / variance. momentum=0 means that
216 |       old information is discarded completely at every time step, while
217 |       momentum=1 means that new information is never incorporated. The
218 |       default of momentum=0.9 should work well in most situations.
219 |     - running_mean: Array of shape (D,) giving running mean of features
220 |     - running_var Array of shape (D,) giving running variance of features
221 |     
222 |   Returns a tuple of:
223 |   - out: Output data, of shape (N, C, H, W)
224 |   - cache: Values needed for the backward pass
225 |   """
226 |   N, C, H, W = x.shape
227 |   x_flat = x.transpose(0, 2, 3, 1).reshape(-1, C)
228 |   out_flat, cache = batchnorm_forward(x_flat, gamma, beta, bn_param)
229 |   out = out_flat.reshape(N, H, W, C).transpose(0, 3, 1, 2)
230 |   return out, cache
231 | 
232 | 
233 | def spatial_batchnorm_backward(dout, cache):
234 |   """
235 |   Computes the backward pass for spatial batch normalization.
236 |   
237 |   Inputs:
238 |   - dout: Upstream derivatives, of shape (N, C, H, W)
239 |   - cache: Values from the forward pass
240 |   
241 |   Returns a tuple of:
242 |   - dx: Gradient with respect to inputs, of shape (N, C, H, W)
243 |   - dgamma: Gradient with respect to scale parameter, of shape (C,)
244 |   - dbeta: Gradient with respect to shift parameter, of shape (C,)
245 |   """
246 |   N, C, H, W = dout.shape
247 |   dout_flat = dout.transpose(0, 2, 3, 1).reshape(-1, C)
248 |   dx_flat, dgamma, dbeta = batchnorm_backward(dout_flat, cache)
249 |   dx = dx_flat.reshape(N, H, W, C).transpose(0, 3, 1, 2)
250 |   return dx, dgamma, dbeta
251 | 
252 | 
253 | def svm_loss(x, y):
254 |   """
255 |   Computes the loss and gradient using for multiclass SVM classification.
256 | 
257 |   Inputs:
258 |   - x: Input data, of shape (N, C) where x[i, j] is the score for the jth class
259 |     for the ith input.
260 |   - y: Vector of labels, of shape (N,) where y[i] is the label for x[i] and
261 |     0 <= y[i] < C
262 | 
263 |   Returns a tuple of:
264 |   - loss: Scalar giving the loss
265 |   - dx: Gradient of the loss with respect to x
266 |   """
267 |   N = x.shape[0]
268 |   correct_class_scores = x[np.arange(N), y]
269 |   margins = np.maximum(0, x - correct_class_scores[:, np.newaxis] + 1.0)
270 |   margins[np.arange(N), y] = 0
271 |   loss = np.sum(margins) / N
272 |   num_pos = np.sum(margins > 0, axis=1)
273 |   dx = np.zeros_like(x)
274 |   dx[margins > 0] = 1
275 |   dx[np.arange(N), y] -= num_pos
276 |   dx /= N
277 |   return loss, dx
278 | 
279 | 
280 | def softmax_loss(x, y):
281 |   """
282 |   Computes the loss and gradient for softmax classification.
283 | 
284 |   Inputs:
285 |   - x: Input data, of shape (N, C) where x[i, j] is the score for the jth class
286 |     for the ith input.
287 |   - y: Vector of labels, of shape (N,) where y[i] is the label for x[i] and
288 |     0 <= y[i] < C
289 | 
290 |   Returns a tuple of:
291 |   - loss: Scalar giving the loss
292 |   - dx: Gradient of the loss with respect to x
293 |   """
294 |   probs = np.exp(x - np.max(x, axis=1, keepdims=True))
295 |   probs /= np.sum(probs, axis=1, keepdims=True)
296 |   N = x.shape[0]
297 |   loss = -np.sum(np.log(probs[np.arange(N), y])) / N
298 |   dx = probs.copy()
299 |   dx[np.arange(N), y] -= 1
300 |   dx /= N
301 |   return loss, dx
302 | 
303 | 


--------------------------------------------------------------------------------
/library/localization.py:
--------------------------------------------------------------------------------
  1 | #Load the model and dependencies
  2 | import time, os, json
  3 | import numpy as np
  4 | from scipy.misc import imread, imresize
  5 | import matplotlib.pyplot as plt
  6 | import pickle
  7 | import cv2, numpy as np
  8 | from deconv_utils import *
  9 | import copy
 10 | from library.classifiers.pretrained_vgg16 import PretrainedVGG
 11 | 
 12 | #Global Variables
 13 | percentile_threshold = 40
 14 | 
 15 | #Function to get the bounding box given an image
 16 | def bbox(im, model, layer=11, n_neurons=5, kmax=10,class_no=0):
 17 |     #Inputs
 18 |     # im = Output of Image file read by cv2 - Not Resized
 19 |     # model = CNN model
 20 |     # layer = Layer from which the image has to be extracted
 21 |     # n_neurons = Number of Neurons to use to build the image
 22 |     # k_max = Maximum Number of neurons to evaluate
 23 |     
 24 |     mask, cache = get_localization_mask(im, model, layer, n_neurons, kmax,class_no)
 25 |     class_no,original_size = cache
 26 |     return get_box_from_mask(mask, original_size)
 27 | 
 28 | def get_localization_mask(im, model, layer, n_neurons, kmax, class_no=0 ):
 29 |     #Get localization Mask for Image
 30 |     
 31 |     #Inputs
 32 |     # im = Output of Image file read by cv2 - Not Resized
 33 |     # model = CNN model
 34 |     
 35 |     original_size = im.shape
 36 |     im = resize_image(im)
 37 |     im = process_image(im)
 38 |     activs, caches = get_activs(model, im)
 39 |     
 40 |     if (class_no==0):
 41 |         class_no = np.argmax(activs[15][0])
 42 |     
 43 |     back_grad = get_backgrad(activs, model, class_no, layer, caches)
 44 |     
 45 |     #Get the Filters of Interest in Sorted Order
 46 |     amax = filter_of_intr(activs,back_grad,kmax,layer)
 47 |     filter_scores = get_filter_scores(amax, model, im, activs,caches, layer, class_no, percentile_thresh=40,use_blob=True)
 48 |     
 49 |     #Union Blobs
 50 |     sorted_scores = sorted(filter_scores,key=lambda x:-x[1])
 51 |     mask = np.zeros(im.shape)
 52 |     for k in range(n_neurons):
 53 |         i,n_score,blob,xmin,xmax,ymin,ymax=sorted_scores[k]
 54 |         mask = (mask+blob)>0
 55 |         
 56 |     localization_cache = class_no,original_size
 57 |     return mask, localization_cache
 58 | 
 59 | def get_box_from_mask(mask, original_size=0):
 60 |     #To get the co-ordinates of Rectangle around the box
 61 |     #Inputs
 62 |     # Mask - N X D X W X H
 63 |     # Original Size is tuple - N X D_orig X W_orig X H_orig
 64 |     
 65 |     mask = mask[0]
 66 |     flat_mask = np.sum(mask, axis = 0)
 67 |     
 68 |     col_max = np.max(flat_mask, axis = 0)
 69 |     row_max = np.max(flat_mask, axis = 1)
 70 |     
 71 |     col_idxs = np.where(col_max>0)
 72 |     xmin = col_idxs[0][0]
 73 |     xmax = col_idxs[0][-1]
 74 |     
 75 |     row_idxs = np.where(row_max>0)
 76 |     ymin = row_idxs[0][0]
 77 |     ymax = row_idxs[0][-1]
 78 |     
 79 |     if (original_size!=0):
 80 |         W_orig, H_orig,_ = original_size
 81 |         xmin = int(xmin*H_orig/224)
 82 |         xmax = int(xmax*H_orig/224)
 83 |         ymin = int(ymin*W_orig/224)
 84 |         ymax = int(ymax*W_orig/224)
 85 | 
 86 |     bbox = (xmin, xmax, ymin, ymax)
 87 |     return bbox
 88 |     
 89 | def visualize(im, bbox_cords):
 90 |     xmin, xmax, ymin, ymax = bbox_cords
 91 |     im = im.astype(np.uint8)
 92 |     new_im = im.copy()
 93 |     mask = np.zeros(im.shape)
 94 |     mask[ymin:ymax,xmin:xmax] = 1
 95 |     new_im = new_im*mask
 96 |     new_im = new_im.astype(np.uint8)
 97 |     new_im = cv2.cvtColor(new_im, cv2.cv.CV_BGR2RGB)
 98 |     plt.imshow(new_im)
 99 |     
100 | def calculate_area(c):
101 |     xmin,xmax,ymin,ymax = c
102 |     return ((xmax-xmin)*(ymax-ymin))
103 | 
104 | def calculate_overlap(xL1,xH1, xL2, xH2):
105 |     if(xH1>=xL2):
106 |         if(xH2>=xH1):
107 |             return (xH1-xL2)
108 |         else:
109 |             #Box 2 axis lies inside 1
110 |             return (xH2-xL2)
111 |     else:
112 |         return 0
113 |     
114 | def eval_precision(c1,c2):
115 |     xmin1, xmax1, ymin1, ymax1 = c1
116 |     xmin2, xmax2, ymin2, ymax2 = c2
117 |    
118 |     #Order by X
119 |     if (xmin2>=xmin1):
120 |         x_overlap = calculate_overlap(xmin1,xmax1,xmin2,xmax2)
121 |     else:
122 |         x_overlap = calculate_overlap(xmin2,xmax2,xmin1,xmax1)
123 | 
124 |     if (ymin2>=ymin1):
125 |         y_overlap = calculate_overlap(ymin1,ymax1,ymin2,ymax2)
126 |     else:
127 |         y_overlap = calculate_overlap(ymin2,ymax2,ymin1,ymax1)
128 |    
129 |     intersection = x_overlap*y_overlap
130 |     union = calculate_area(c1) + calculate_area(c2) - intersection
131 |     
132 |     return float(intersection)/float(union)
133 |     
134 |     
135 | def main():
136 |     #Sample Code for an image
137 |     #Load the Image
138 |     imgf = 'Images/dog.jpg'
139 |     im = cv2.imread(imgf)
140 |     
141 |     #Load the VGG Model and ImageNet Class Mappings
142 |     model = PretrainedVGG(h5_file = 'Data/vgg16_weights.h5')
143 |     CLASSES = pickle.load(open('Data/CLASSES.pkl'))
144 |     
145 |     #Layer from which the image will be segment
146 |     layer = 11
147 |     #No of neurons to evaluate from ranked list
148 |     kmax = 10
149 |     #No of neurons to use
150 |     n_neurons = 5
151 |     
152 |     #Get Bounding Box
153 |     bbox_coords = bbox(im,model,layer,n_neurons,kmax)
154 |     #Visualize the Box on the original image
155 |     visualize(im, bbox_cords)
156 |     return bbox_cords


--------------------------------------------------------------------------------
/library/setup.py:
--------------------------------------------------------------------------------
 1 | from distutils.core import setup
 2 | from distutils.extension import Extension
 3 | from Cython.Build import cythonize
 4 | import numpy
 5 | 
 6 | extensions = [
 7 |   Extension('im2col_cython', ['im2col_cython.pyx'],
 8 |             include_dirs = [numpy.get_include()]
 9 |   ),
10 | ]
11 | 
12 | setup(
13 |     ext_modules = cythonize(extensions),
14 | )
15 | 


--------------------------------------------------------------------------------
/validation_script.py:
--------------------------------------------------------------------------------
  1 | # Script to Run the validation for Localization Algorithm
  2 | # Takes Input a Data file(.pkl) which containts the dataset
  3 | # Data set input needed is a pkl file with row format:
  4 | # class_wnid,imgid,class_idx,xmlidx,url,xmin,xmax,ymin,ymax = ret
  5 | # It uses a VGG16 model and writes the results in the same folder as input with _results.csv as suffix to input file name
  6 | # hyper Parameters used by algorithm are currently hardcoded in the script
  7 | 
  8 | import types
  9 | from library.localization import *
 10 | from deconv_utils import *
 11 | from library.image_utils import *
 12 | import csv
 13 | import sys
 14 | import datetime
 15 | import time
 16 | import os
 17 | 
 18 | #############################################################
 19 | #Image Processing Inputs
 20 | Param.num_dilation = 15
 21 | #############################################################
 22 | 
 23 | def get_time_stamp():
 24 |     return datetime.datetime.fromtimestamp(time.time()).strftime('%Y-%m-%d %H:%M:%S')
 25 | 
 26 | def print_and_log(text, filename):
 27 |     print (get_time_stamp()+text)
 28 |     filename.write(get_time_stamp()+ " "+text+"\n")
 29 | 
 30 | #HyperParams
 31 | kmax_input = 30
 32 | n_neurons_input = 5
 33 | 
 34 | #Validation & Model File Inputs
 35 | data_file = sys.argv[1]
 36 | path = os.path.dirname(data_file)
 37 | name = os.path.basename(data_file)
 38 | data_extn = ".pkl"
 39 | results_path = path + "/results_k"+str(kmax_input)+"_n"+str(n_neurons_input)+"/"
 40 | results_file = results_path+ name + "_results.csv"
 41 | log_file = results_path+ name + "_log.txt"
 42 | model_file = 'Data/vgg16_weights.h5'
 43 | class_file = 'Data/CLASSES.pkl'
 44 | 
 45 | start_index = 0
 46 | result_csv = None
 47 | try:
 48 |     #Check if csv file exists and read the last index outputted
 49 |     results_csv = open(results_file,'r')
 50 |     lastline = results_csv.read()
 51 |     if lastline!="":
 52 |         lastline = lastline.split('\n')[-2]
 53 |         start_index = int(lastline.split(',')[0]) + 1
 54 | except IOError as e:
 55 |     pass
 56 | finally:
 57 |     if type(result_csv)!=types.NoneType:
 58 |         close(results_csv)
 59 | 
 60 | try:
 61 |     #Write to CSV file
 62 |     results_csv = open(results_file,'a')
 63 |     log_txt = open(log_file,'a')
 64 |     msg = ("#################################################################################################\n")
 65 |     print_and_log(msg, log_txt)
 66 | 
 67 |     #Load the VGG Model and ImageNet Class Mappings
 68 |     from library.classifiers.pretrained_vgg16 import PretrainedVGG
 69 |     model = PretrainedVGG(h5_file = model_file)
 70 |     print_and_log("Loaded Model", log_txt)
 71 | 
 72 |     CLASSES = pickle.load(open(class_file))
 73 |     print_and_log("Loaded Class File",log_txt)
 74 | 
 75 |     #Read Input File for the candidates
 76 |     candidates = pickle.load(open(data_file+data_extn))
 77 |     print_and_log("Loaded Input Data Set",log_txt)
 78 | 
 79 |     #Read Input File and start at the index higher than that of last line
 80 | 
 81 |     for index in range(start_index,len(candidates)):
 82 |         log_txt.flush()
 83 |         ret = candidates[index]
 84 |         class_wnid,imgid,class_idx,xmlidx,url,xmin,xmax,ymin,ymax = ret
 85 |         im = image_from_url(url)
 86 |         if type(im) == types.NoneType:
 87 |             #Debug Print
 88 |             msg = "[%d]: Skip (%d,%d): %s : URL is bad."%(index, class_idx,xmlidx, url)
 89 |             print_and_log(msg,log_txt)
 90 |         elif np.mean(im)>=253:
 91 |             #Debug Print
 92 |             msg = "[%d]: Skip (%d,%d): %s : URL is Empty(White) Image"%(index, class_idx,xmlidx,url)
 93 |             print_and_log(msg,log_txt)
 94 |         else:  
 95 |             bbox_coords = bbox(im,model,class_no=class_idx,n_neurons = n_neurons_input,kmax=kmax_input)
 96 |             xmin_out, xmax_out, ymin_out, ymax_out = bbox_coords
 97 |             precision = eval_precision(bbox_coords, (xmin,xmax,ymin,ymax))
 98 |             #Write Results to file
 99 |             result_row = "%d, %s, %d, %d, %f,"%(index, imgid, class_idx, xmlidx, precision)
100 |             result_row += "%d, %d, %d, %d,"%(xmin, xmax, ymin, ymax,)
101 |             result_row += "%d, %d, %d, %d"%(xmin_out, xmax_out, ymin_out, ymax_out)
102 |             result_row += "\n"
103 |             results_csv.write(result_row)
104 |             results_csv.flush()
105 |             #Debug Print
106 |             msg = "[%d]: %d,%d,%f"%(index, class_idx,xmlidx,precision)
107 |             print_and_log(msg,log_txt)
108 | except IOError as e:
109 |     print 'IO Error: ', e.code
110 | finally:
111 |     close(results_csv)
112 |     close(log_txt)


--------------------------------------------------------------------------------