%PDF-1.3 1 0 obj << /Kids [ 3 0 R 4 0 R 5 0 R 6 0 R 7 0 R 8 0 R 9 0 R 10 0 R 11 0 R 12 0 R ] /Type /Pages /Count 10 >> endobj 2 0 obj << /Title (What\047s in a Question\072 Using Visual Questions as a Form of Supervision) /Producer (PyPDF2) /Author (Siddha Ganju\054 Olga Russakovsky\054 Abhinav Gupta) /Subject (2017 IEEE Conference on Computer Vision and Pattern Recognition) >> endobj 3 0 obj << /Parent 1 0 R /Rotate 0 /Contents 14 0 R /Resources << /XObject << /x8 15 0 R /x10 18 0 R /x6 21 0 R /x12 24 0 R /R41 27 0 R /R40 28 0 R >> /ExtGState << /s9 29 0 R /R28 32 0 R /s11 33 0 R /a0 << /CA 1 /ca 1 >> /s5 36 0 R /s7 39 0 R >> /Font << /F2 42 0 R /R29 43 0 R /F1 47 0 R /R37 48 0 R /R35 52 0 R /R33 55 0 R /R31 59 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /Group 63 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 64 0 R 65 0 R 66 0 R 67 0 R 68 0 R 69 0 R 70 0 R 71 0 R 72 0 R 73 0 R 74 0 R 75 0 R 76 0 R 77 0 R 78 0 R 79 0 R 80 0 R 81 0 R ] /Type /Page >> endobj 4 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 82 0 R /Resources << /ExtGState << /R28 32 0 R >> /Font << /R29 43 0 R /F2 83 0 R /R102 84 0 R /R33 55 0 R /R37 48 0 R /F1 87 0 R /R35 52 0 R /R104 88 0 R /R31 59 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 91 0 R 92 0 R 93 0 R 94 0 R 95 0 R 96 0 R 97 0 R 98 0 R 99 0 R 100 0 R 101 0 R 102 0 R 103 0 R 104 0 R 105 0 R 106 0 R 107 0 R 108 0 R 109 0 R 110 0 R 111 0 R 112 0 R 113 0 R 114 0 R 115 0 R 116 0 R 117 0 R 118 0 R 119 0 R 120 0 R 121 0 R 122 0 R 123 0 R 124 0 R 125 0 R 126 0 R 127 0 R 128 0 R 129 0 R 130 0 R 131 0 R 132 0 R 133 0 R 134 0 R 135 0 R 136 0 R 137 0 R 138 0 R 139 0 R 140 0 R 141 0 R 142 0 R ] >> endobj 5 0 obj << /Parent 1 0 R /Rotate 0 /Contents 143 0 R /Resources << /XObject << /R143 144 0 R /R127 145 0 R /R126 146 0 R /R125 147 0 R /R133 148 0 R /R134 149 0 R /R135 150 0 R /R136 151 0 R /R129 152 0 R /R128 153 0 R /R145 154 0 R /R144 157 0 R /R147 158 0 R /R146 159 0 R /R152 160 0 R /R153 161 0 R /R150 162 0 R /R131 163 0 R /R151 164 0 R /R149 165 0 R /R148 166 0 R /R132 167 0 R /R130 168 0 R >> /ExtGState << /R28 32 0 R /R140 169 0 R /R137 170 0 R >> /Font << /R29 43 0 R /F2 171 0 R /F1 172 0 R /R33 55 0 R /R37 48 0 R /R35 52 0 R /R104 88 0 R /R31 59 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R139 173 0 R /R142 155 0 R >> >> /Group 63 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 175 0 R 176 0 R 177 0 R 178 0 R 179 0 R 180 0 R 181 0 R 182 0 R 183 0 R 184 0 R 185 0 R 186 0 R 187 0 R 188 0 R 189 0 R ] /Type /Page >> endobj 6 0 obj << /Parent 1 0 R /Rotate 0 /Contents 190 0 R /Resources << /XObject << /R192 191 0 R /R193 192 0 R /R190 193 0 R /R191 194 0 R /R196 195 0 R /R197 196 0 R /R194 197 0 R /R195 198 0 R /R189 199 0 R /R188 200 0 R /R186 201 0 R /R185 202 0 R /R187 204 0 R /R184 203 0 R >> /ExtGState << /R28 32 0 R >> /Font << /R29 43 0 R /F2 205 0 R /R102 84 0 R /R37 48 0 R /F1 206 0 R /R35 52 0 R /R33 55 0 R /R31 59 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /Group 63 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 207 0 R 208 0 R 209 0 R 210 0 R 211 0 R 212 0 R 213 0 R 214 0 R 215 0 R 216 0 R 217 0 R 218 0 R 219 0 R 220 0 R 221 0 R 222 0 R 223 0 R 224 0 R 225 0 R 226 0 R 227 0 R 228 0 R 229 0 R ] /Type /Page >> endobj 7 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 230 0 R /Resources << /XObject << /R212 231 0 R >> /ExtGState << /R211 232 0 R /R28 32 0 R >> /Font << /F2 233 0 R /R213 234 0 R /R215 238 0 R /R217 243 0 R /R102 84 0 R /R29 43 0 R /F1 246 0 R /R37 48 0 R /R35 52 0 R /R31 59 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 247 0 R 248 0 R 249 0 R 250 0 R 251 0 R 252 0 R 253 0 R ] >> endobj 8 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 254 0 R /Resources << /ExtGState << /R28 32 0 R >> /Font << /F2 255 0 R /R35 52 0 R /R215 238 0 R /R217 243 0 R /R102 84 0 R /R29 43 0 R /F1 256 0 R /R249 257 0 R /R37 48 0 R /R245 260 0 R /R251 263 0 R /R33 55 0 R /R247 267 0 R /R31 59 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 271 0 R 272 0 R 273 0 R 274 0 R 275 0 R 276 0 R 277 0 R 278 0 R 279 0 R 280 0 R 281 0 R 282 0 R 283 0 R 284 0 R 285 0 R 286 0 R 287 0 R 288 0 R 289 0 R ] >> endobj 9 0 obj << /Parent 1 0 R /Rotate 0 /Contents 290 0 R /Resources << /XObject << /R266 291 0 R /R267 292 0 R /R268 293 0 R /R269 294 0 R /R279 295 0 R /R278 296 0 R /R273 297 0 R /R272 298 0 R /R271 299 0 R /R270 300 0 R /R277 301 0 R /R276 302 0 R /R275 303 0 R /R274 304 0 R /R288 305 0 R /R289 306 0 R /R282 307 0 R /R283 308 0 R /R280 309 0 R /R281 310 0 R /R286 311 0 R /R287 312 0 R /R284 313 0 R /R285 314 0 R >> /ExtGState << /R28 32 0 R >> /Font << /F2 315 0 R /R213 234 0 R /R215 238 0 R /R217 243 0 R /R29 43 0 R /F1 316 0 R /R37 48 0 R /R35 52 0 R /R33 55 0 R /R31 59 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /Group 63 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 317 0 R 318 0 R 319 0 R 320 0 R 321 0 R 322 0 R 323 0 R 324 0 R 325 0 R ] /Type /Page >> endobj 10 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 326 0 R /Resources << /XObject << /R306 327 0 R /R305 328 0 R /R304 329 0 R >> /ExtGState << /R28 32 0 R >> /Font << /R29 43 0 R /F2 330 0 R /R35 52 0 R /R247 267 0 R /R37 48 0 R /F1 331 0 R /R245 260 0 R /R33 55 0 R /R31 59 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 332 0 R 333 0 R 334 0 R 335 0 R 336 0 R 337 0 R 338 0 R 339 0 R 340 0 R ] >> endobj 11 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 341 0 R /Resources << /ExtGState << /R28 32 0 R >> /Font << /R29 43 0 R /F2 342 0 R /F1 343 0 R /R33 55 0 R /R31 59 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 344 0 R 345 0 R 346 0 R 347 0 R 348 0 R 349 0 R 350 0 R 351 0 R 352 0 R 353 0 R 354 0 R 355 0 R 356 0 R 357 0 R 358 0 R 359 0 R 360 0 R 361 0 R 362 0 R 363 0 R 364 0 R 365 0 R 366 0 R 367 0 R 368 0 R 369 0 R 370 0 R 371 0 R 372 0 R 373 0 R 374 0 R 375 0 R 376 0 R 377 0 R 378 0 R 379 0 R 380 0 R 381 0 R 382 0 R 383 0 R 384 0 R 385 0 R 386 0 R 387 0 R 388 0 R 389 0 R 390 0 R 391 0 R 392 0 R 393 0 R 394 0 R ] >> endobj 12 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 395 0 R /Resources << /ExtGState << /R28 32 0 R >> /Font << /F2 396 0 R /F1 397 0 R /R33 55 0 R /R31 59 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 398 0 R 399 0 R 400 0 R 401 0 R 402 0 R 403 0 R 404 0 R 405 0 R 406 0 R 407 0 R 408 0 R 409 0 R 410 0 R 411 0 R 412 0 R 413 0 R 414 0 R 415 0 R 416 0 R 417 0 R 418 0 R 419 0 R 420 0 R 421 0 R 422 0 R 423 0 R 424 0 R 425 0 R 426 0 R 427 0 R 428 0 R 429 0 R 430 0 R 431 0 R 432 0 R 433 0 R 434 0 R 435 0 R 436 0 R 437 0 R 438 0 R 439 0 R 440 0 R 441 0 R 442 0 R 443 0 R ] >> endobj 13 0 obj << /Type /Catalog /Pages 1 0 R >> endobj 14 0 obj << /Length 19926 >> stream q q q 0.10000 0 0 0.10000 0 0 cm /R28 gs 0 g q 10 0 0 10 0 0 cm BT /R29 14.34620 Tf 1 0 0 1 79.76020 675.06700 Tm [ (What\047) 37 (s) -249.99700 (in) -249.99200 (a) -250.00800 (Question\072) -310.01300 (Using) -249.99100 (V) 37.01370 (isual) -250.00500 (Questions) -250.01000 (as) -249.99300 (a) -250.00800 (F) 24.98890 (orm) -249.99500 (of) -250.01200 (Super) 10.00310 (vision) ] TJ /R31 11.95520 Tf 46.06170 -37.85820 Td [ (Siddha) -250.00400 (Ganju) -4861.01000 (Olg) 5 (a) -249.99700 (Russak) 9.99010 (o) 14.98510 (vsk) 14.98920 (y) ] TJ 101.18300 -13.94690 Td [ (Carne) 15.01370 (gie) -250 (Mellon) -249.99700 (Uni) 24.99570 (v) 14.98510 (ersity) ] TJ 169.25500 13.94690 Td [ (Abhina) 19.99650 (v) -250.00200 (Gupta) ] TJ /R29 11.95520 Tf -250.26500 -54.99410 Td (Abstract) Tj /R33 9.96260 Tf -83.92770 -24.58980 Td [ (Collecting) -406.00200 (fully) -407.00800 (annotated) -406.00900 (ima) 10.01300 (g) 10.00320 (e) -406.01600 (datasets) -406.01300 (is) -407.01800 (c) 15.01220 (halleng\055) ] TJ -11.95510 -11.95510 Td [ (ing) -360.98100 (and) -360.98300 (e) 19.99240 (xpensive) 14.98400 (\056) -642.98800 (Many) -360.99100 (types) -360.98900 (of) -360.98600 (weak) -360.98900 (supervision) -360.99400 (have) ] TJ 11.95510 TL T* [ (been) -354.01800 (e) 19.99240 (xplor) 36.99320 (ed\072) -517.99400 (weak) -354.01000 (manual) -353.98900 (annotations\054) -381.00800 (web) -354.00500 (sear) 36.98650 (c) 15.01220 (h) -354.01400 (r) 37.01830 (e\055) ] TJ T* [ (sults\054) -381 (tempor) 15 (al) -354.00700 (continuity) 55.01170 (\054) -380.99300 (ambient) -353.99100 (sound) -355.01200 (and) -354.00500 (other) 10.01790 (s\056) -624.00400 (W) 91.98710 (e) ] TJ 11.95590 TL T* [ (focus) -272.99400 (on) -273.00300 (one) -273.00800 (particular) -273.00300 (une) 20.00280 (xplor) 36.99260 (ed) -273.01200 (mode\072) -356.99400 (visual) -272.98600 (questions) ] TJ 11.95510 TL T* [ (that) -318.98100 (ar) 36.98520 (e) -319.99000 (ask) 10.00510 (ed) -319.00400 (about) -319.01800 (ima) 10.01300 (g) 10.00320 (es\056) -517.98800 (The) -320.01000 (k) 9.99343 (e) 29.99010 (y) -319.01000 (observation) -319 (that) -320 (in\055) ] TJ T* [ (spir) 36.99870 (es) -229.99900 (our) -231.00400 (work) -229.98100 (is) -231.00700 (that) -230.01600 (the) -231.01300 (question) -230.00500 (itself) -230.99100 (pr) 44.98510 (o) 10.00320 (vides) -229.99100 (useful) -230.99300 (in\055) ] TJ T* [ (formation) -294.99900 (about) -294.98300 (the) -295.00200 (ima) 10.01300 (g) 10.00320 (e) -295.01400 (\050e) 15.00980 (ven) -295.01300 (without) -296.01900 (the) -295.00200 (answer) -295.01700 (being) ] TJ T* [ (available\051\056) -644.01800 (F) 105.00600 (or) -360.98800 (instance) 9.98608 (\054) -388.99100 (the) -360.99100 (question) -361.00100 (\223what) -360.99400 (is) -362.00600 (t) 0.98758 (he) -362.01800 (br) 36.98650 (eed) ] TJ T* [ (of) -354.98700 (the) -354.01100 (do) 10.00810 (g\077\224) -623.98900 (informs) -354.01900 (the) -354.99200 (AI) -354.01800 (that) -355.01400 (the) -354.01200 (animal) -355.00700 (in) -354.00700 (the) -354.99200 (scene) ] TJ 11.95590 TL T* [ (is) -332.99000 (a) -332.01800 (do) 10.00810 (g) -332.99800 (and) -332.98800 (that) -331.99900 (ther) 36.98770 (e) -333.00800 (is) -332.99000 (only) -332.99100 (one) -332.01700 (do) 10.00810 (g) -332.99800 (pr) 36.98650 (esent\056) -558.01200 (W) 91.98710 (e) -333.00800 (mak) 10 (e) ] TJ 11.95510 TL T* [ (thr) 36.99260 (ee) -331.01200 (contrib) 19.98870 (utions\072) -470.98800 (\0501\051) -331.00500 (pr) 44.98390 (o) 10.00320 (viding) -331.01800 (an) -330.99400 (e) 19.99180 (xtensive) -330.99600 (qualitative) ] TJ T* [ (and) -387.99800 (quantitative) -386.98100 (analysis) -388.01600 (of) -386.98200 (the) -388.00600 (information) -387.00900 (contained) -387.98200 (in) ] TJ T* [ (human) -310.00400 (visual) -308.98200 (questions\054) -324.99800 (\0502\051) -309.98900 (pr) 44.98390 (oposing) -309.98700 (two) -308.98300 (simple) -310.00200 (b) 20.00160 (ut) -310.01400 (sur) 20.01380 (\055) ] TJ T* [ (prisingly) -196.98200 (ef) 18 (fective) -196.98300 <6d6f646902636174696f6e73> -197.98900 (to) -197.01400 (the) -197.01900 (standar) 37.01340 (d) -196.98200 (visual) -198.01800 (ques\055) ] TJ 11.95630 TL T* [ (tion) -238.99400 (answering) -239.01000 (models) -238.99700 (that) -238.99400 (allow) -238.98100 (them) -239.00800 (to) -239.00700 (mak) 10 (e) -240.00400 (use) -239.01200 (of) -239.00700 (weak) ] TJ 11.95470 TL T* [ (supervision) -233.99500 (in) -233.98800 (the) -235.01200 (form) -234.01900 (of) -233.98800 (unanswer) 37.01590 (ed) -233.99900 (questions) -235.01700 (associated) ] TJ T* [ (with) -275.02000 (i) 0.98758 (ma) 10.00630 (g) 10.00320 (es) -275.01100 (and) -274.99800 (\0503\051) -273.99500 (demonstr) 15.01100 (ating) -274.98400 (that) -274.00800 (a) -275.00800 (simple) -274.00800 (data) -274.99100 (aug\055) ] TJ T* [ (mentation) -204.98500 (str) 14.99750 (ate) 40 (gy) -204.00400 (inspir) 37.01100 (ed) -204.98400 (by) -204.00400 (our) -205.00900 (insights) -205.01500 (r) 37.01830 (esults) -204.01700 (in) -205.01200 (a) ] TJ /R35 9.96260 Tf 215.21800 0 Td (7) Tj /R37 9.96260 Tf 4.98086 0 Td [ (\072) -0.79889 ] TJ /R35 9.96260 Tf 2.76719 0 Td [ (1\045) -0.29897 ] TJ /R33 9.96260 Tf -222.96600 -11.95510 Td [ (impr) 44.99370 (o) 10.00320 (vement) -249.99600 (on) -249.98800 (the) -249.99000 (standar) 37.01340 (d) -249.99300 (VQA) -249.98600 (benc) 15.01960 (hmark\056) ] TJ /R29 11.95520 Tf 37.90590 TL T* [ (1\056) -249.99000 (Intr) 18.01460 (oduction) ] TJ /R31 9.96260 Tf 11.95510 -19.60900 Td [ (Supervised) -202.00900 (learning) -201.98800 (has) -203.01800 (sho) 24.99220 (wn) -201.99800 (great) -202.00600 (promise) -202.01800 (in) -203.01300 (de) 25.01540 (v) 14.98280 (elop\055) ] TJ -11.95510 -11.95510 Td [ (ing) -320.98800 (visual) -321.99800 (AI\056) -321.00300 (Ho) 24.98360 (we) 25.01420 (v) 14.98280 (er) 39.98600 (\054) -339 (collecting) -321.01400 (manually) -321.01300 (annotated) -322.00300 (vi\055) ] TJ 11.95510 TL T* [ (sual) -225.00700 (datasets) -223.98300 (is) -225.00800 (both) -224.01900 (challenging) -224.98800 (and) -225.01500 (e) 15.01220 (xpensi) 25 (v) 14.98280 (e) -224.00700 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 241.20500 249.05200 Tm (36) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 251.16800 249.05200 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 255.90000 249.05200 Tm (14) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 265.86300 249.05200 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 270.59500 249.05200 Tm (28) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 280.55700 249.05200 Tm (\135\056) Tj -230.44500 -11.95470 Td [ (Using) -191.98100 (cheaper) -190.99900 (and) -192.00100 (weak) 9.99711 (er) -192.01500 (supervision) -192.00200 (is) -191.01500 (a) -192.01200 (gro) 24.98360 (wing) -191.98800 (research) ] TJ 11.95510 TL T* [ (direction) -387.98900 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 92.71210 225.14200 Tm (46) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 102.67500 225.14200 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 109.03100 225.14200 Tm (42) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 118.99300 225.14200 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 125.35900 225.14200 Tm (8) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 130.34100 225.14200 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 136.69700 225.14200 Tm (12) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 146.65900 225.14200 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 153.01600 225.14200 Tm (35) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 162.97800 225.14200 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 169.33400 225.14200 Tm (7) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 174.31600 225.14200 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 180.67200 225.14200 Tm (34) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 190.63400 225.14200 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 197 225.14200 Tm (4) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 201.98200 225.14200 Tm [ (\135\056) -723.99700 (As) -388.00400 (AI) -388.99400 (is) -388.00100 (increas\055) ] TJ -151.87000 -11.95510 Td [ (ingly) -353.01000 (inte) 14.99320 (grated) -354.01400 (into) -353.01500 (our) -352.99300 (daily) -354 (li) 24.99090 (v) 14.98280 (es\054) -378.99200 (computer) -353.01700 (vision) -353.98500 (sys\055) ] TJ 11.95590 TL T* [ (tems) -266.99300 (will) -266.99000 (ha) 19.99730 (v) 14.98280 (e) -268 (access) -267.01900 (to) -267.00200 (increasingly) -266.98200 (di) 24.98850 (v) 14.98280 (erse) -266.98000 (sources) -267.00400 (of) -267.99200 (in\055) ] TJ 11.95510 TL T* [ (formation) -323.01300 (by) -322.99500 (constantly) -323.00500 (observing) -323.01300 (human\055human\054) -340.99700 (human\055) ] TJ T* [ (object\054) -229.00200 (human\055en) 40.01420 (vironment) -224.00200 (and) -223.99600 (human\055AI) -223.98200 (interactions\056) -301.99100 (Ef\055) ] TJ T* [ <026369656e746c79> -301.01000 (util) 1.00657 (izing) -300.98900 (all) -300.99900 (this) -300.00300 (information) -301.00500 (will) -300.00400 (become) -300.99900 (increas\055) ] TJ T* [ (ingly) -236.99000 (critical) -237.98700 (as) -237.01800 (we) -238 (aim) -237.00500 (to) -237.98700 (de) 25.01670 (v) 14.98280 (elop) -237.00700 (lar) 17.99700 (ge\055scale\054) -240 (accurate) -238.01200 (and) ] TJ T* [ (adaptable) -249.98500 (visual) -250.01100 (systems\056) ] TJ 11.95510 -12.63630 Td [ (V) 59.99310 (isual) -371.99600 (Question) -371.98200 (Answering) -371.98900 (\050VQA\051) -372.01400 (has) -372.00900 (become) -372.00700 (a) -371.98200 (ne) 25.01540 (w) ] TJ -11.95510 -11.95470 Td [ (mode) -256.01800 (of) -255.01400 (interaction) -256.00900 (between) -255.98900 (humans) -254.98700 (and) -255.99200 (AI) -255.99600 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 233.39400 116.86500 Tm (3) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 238.37500 116.86500 Tm [ (\135\056) -327.00400 (Presently) 65.00140 (\054) ] TJ -188.26300 -11.95510 Td [ (VQA) -412.00100 (is) -413.01600 (mostly) -412.00600 (used) -413.01700 (as) -412.00800 (means) -411.99500 (for) -412.99100 (e) 25.01050 (v) 24.98110 (aluating) -411.98300 (visual) -413.00200 (rea\055) ] TJ 11.95510 TL T* [ (soning) -303.01400 (capabilities) -303.99300 (of) -303.00700 (computers\056) -470.00900 (Ho) 24.98360 (we) 25.01300 (v) 14.98280 (er) 39.98600 (\054) -315.98400 (going) -304 (forw) 10 (ard) ] TJ T* [ (this) -480.01200 (is) -479.00500 (lik) 10.01790 (ely) -479.99100 (to) -479.98600 (become) -479.99000 (a) -478.98400 (natural) -479.98600 (human\055AI) -480.01800 (interaction) ] TJ ET Q q 2362.20000 0 0 718.03000 3088.62000 5203.73000 cm /R41 Do Q q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 308.86200 502.04200 Tm [ (Figure) -292.98500 (1\072) -396 (W) 79.98660 (e) -293.01500 (e) 15.01220 (xamine) -292.99000 (ho) 24.98600 (w) -293.00500 (much) -294.01000 (information) -293.00500 (is) -292.99500 (contained) ] TJ 11.95630 TL T* [ (in) -306.01500 (a) -307.01200 (visual) -306 (question\054) -319.98600 (and) -307.00300 (demonstrate) -305.99800 (that) -306.01300 (this) -306.98300 (information) ] TJ 11.95510 TL T* [ (can) -250 (be) -249.99700 (ef) 25.00810 (fecti) 25.01790 (v) 14.98280 (ely) -249.99000 (used) -249.98500 (in) -249.98800 (training) -250.00700 (computer) -250.01200 (vision) -250.00200 (models\056) ] TJ 35.47500 TL T* [ (paradigm\056) -415.98400 (It) -285.00600 (will) -284.98700 (become) -285.00100 (commonplace) -285.01600 (for) -285.01100 (humans) -286.00100 (to) -285.00100 (ask) ] TJ 11.95470 TL T* [ (computers) -236.99000 (visual) -238.01200 (questions\054) -239.99400 (such) -237.00900 (as\054) -239.99400 (\223Where) -236.99000 (did) -237.98000 (I) -236.98500 (lea) 19.99670 (v) 14.98280 (e) -238.00400 (my) ] TJ T* [ (k) 10.00320 (e) 15.01220 (ys\077\224\054) -273.98900 (\223What) -267.99000 (breed) -269.01600 (of) -269.01100 (dog) -269 (is) -269 (this\077\224\054) -272.99400 (\223Ha) 19.99180 (v) 14.98280 (e) -269.01900 (I) -269.01900 (met) -269 (this) -268.98900 (per) 19.99180 (\055) ] TJ T* [ (son) -260.99800 (before\077\224) -343.99700 (or) -261.01300 (\223Wh) 4.98201 (y) -261.99100 (is) -261.00300 (she) -261.01000 (doing) -260.98600 (that\077\224\056) -344.01100 (Instead) -261.00500 (of) -261.99500 (vie) 24.98600 (w\055) ] TJ 11.95590 TL T* [ (ing) -348.00300 (this) -347.99600 (as) -348.01800 (single\055sided) -348.01800 (interaction) -347.99100 (with) -347.99600 (humans) -347.99100 (soliciting) ] TJ 11.95510 TL T* [ (information) -219.99800 (from) -221.01200 (AI) -220.00300 (systems\054) -227.02100 (we) -220.00300 (consider) -221.00700 (ho) 24.98600 (w) -219.99800 (visual) -220.99700 (ques\055) ] TJ T* [ (tions) -301.99900 (themselv) 15.01220 (es) -303.00600 (can) -301.99100 (serv) 14.98280 (e) -303.01300 (as) -301.98900 (a) -303.01300 (form) -302.01800 (of) -303.00800 (supervision) -301.98400 (to) -302.99800 (im\055) ] TJ T* [ (pro) 14.98520 (v) 14.98280 (e) -250.00200 (computer) -250.01500 (vision) -250.00200 (systems) -249.99700 (\050Figure) ] TJ ET Q 1 0 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 466.57100 358.97000 Tm (1) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 471.55200 358.97000 Tm (\051\056) Tj -150.73500 -12.95630 Td [ (In) -416.98700 (contrast) -415.98900 (to) -417.01600 (e) 15.01220 (xisting) -416.00400 (w) 10.00320 (orks) -417.01400 (that) -417.01400 (focus) -416 (on) -417.01900 (impro) 15.00240 (ving) ] TJ -11.95510 -11.95510 Td [ (AI\047) 55.00800 (s) -259.99100 (VQA) -259.98800 (capabilities) -258.98100 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 404.05400 334.05900 Tm (16) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 414.01700 334.05900 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 419.09800 334.05900 Tm (26) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 429.06000 334.05900 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 434.14100 334.05900 Tm (33) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 444.10400 334.05900 Tm [ (\135\054) -261.99500 (we) -259.99600 (stri) 24.99580 (v) 14.98280 (e) -260 (to) -259.98600 (understand) ] TJ -135.24100 -11.95470 Td [ (ho) 24.98600 (w) -307.98500 (much) -306.99000 (information) -307.98300 (is) -308.01200 (contained) -308.00700 (within) -307.01200 (the) ] TJ /R33 9.96260 Tf 190.56500 0 Td (question) Tj /R31 9.96260 Tf 36.82810 0 Td (it\055) Tj -227.39300 -11.95510 Td [ (self\054) -241.99100 (e) 25.01050 (v) 14.98280 (en) -240 (when) -241.01100 (the) -239.99200 (answer) -240 (is) -239.98400 (not) -240.99900 (pro) 14.98280 (vided) -239.98000 (\050as) -240 (w) 10.00320 (ould) -240.01400 (be) -241.01900 (the) ] TJ 11.95510 TL T* [ (case) -264.00300 (of) -263.99300 (human\055AI) -264.01500 (con) 39.99820 (v) 14.98280 (ersations\051\056) -353.01500 (E\056g\056\054) -267 (the) -265.00500 (question) -264 (\223What) ] TJ 11.95590 TL T* [ (breed) -273.01600 (of) -273.01100 (dog) -272.99900 (is) -274.02000 (this\077\224) -379.01800 (pro) 14.98280 (vides) -272.99400 (information) -273.00800 (that) -274.01800 (the) -273.00300 (animal) ] TJ 11.95510 TL T* [ (in) -273.00100 (the) -273.00600 (scene) -273.01600 (is) -273.00100 (a) -273.01800 (dog) -272.99900 (and) -273.00800 (suggests) -273.01800 (that) -272.99900 (there) -273.99300 (is) -272.99900 (a) -273.01800 (single) -272.98900 (dog) ] TJ T* [ (present\056) -675.99000 (The) -372.01100 (question) -371 (\223Wh) 4.97956 (y) -372.01100 (is) -372.00200 (he) -372.01600 (doing) -371.98700 (that\077\224) -676 (suggests) ] TJ T* [ (that) -338.98800 (the) -338.01500 (depicted) -338.98500 (beha) 20 (vior) -338.98800 (is) -338.00700 (unusual) -339.01200 (or) -339.00200 (une) 14.98280 (xpected\056) -576.01200 (This) ] TJ T* [ (type) -324.99200 (of) -325.00200 (fr) 1.01209 (ee\054) -343.98900 (natural) -324.99200 (and) -324.01900 (open\055ended) -324.99000 (supervision) -325 (can) -325.00900 (pa) 19.99670 (v) 14.98280 (e) ] TJ T* [ (the) -249.99000 (w) 10 (ay) -249.99700 (to) -249.98500 (de) 25.01540 (v) 14.98280 (eloping) -250.00700 (richer) -250.00200 (cogniti) 25.01050 (v) 14.98280 (e) -250.00200 (AI\056) ] TJ 11.95510 -12.95630 Td [ (W) 79.98660 (e) -478.00400 (set) -477.98900 (out) -479 (to) -477.98600 (in) 40.00560 (v) 14.98280 (estig) 5 (ate) -477.99900 (this) -478.01300 (h) 4.98446 (ypothesis) -477.99400 (that) -479.00300 (human) ] TJ -11.95510 -11.95470 Td [ (questions) -323.98300 (can) -323.98800 (be) -323.98500 (ef) 25.00810 (fec) 1.00718 (ti) 24.99090 (v) 14.98280 (ely) -324.01500 (used) -324.01500 (to) -324.01500 (impro) 15.00240 (v) 14.98280 (e) -323.99000 (computer) -324 (vi\055) ] TJ T* [ (sion) -280.01000 (capabi) 0.98513 (lities\056) -399 (W) 79.98660 (e) -279.99700 (be) 15.01710 (gin) -278.99300 (by) -279.98300 (pro) 14.98280 (viding) -279.01700 (e) 15.01220 (xtensi) 25 (v) 14.98280 (e) -279.99700 (qualita\055) ] TJ T* [ (ti) 24.99090 (v) 14.98280 (e) -319.99100 (and) -320.02000 (quantitati) 25.01790 (v) 14.98280 (e) -319.99100 (analysis) -319.99600 (of) -321.00500 (the) -320.01500 (information) -319.98100 (contained) ] TJ 11.95590 TL T* [ (in) -294.99700 (a) -295.01400 (visual) -294.00200 (question) -295.01200 (using) -294.98000 (the) -295 (lar) 17.99700 (ge\055s) 0.98513 (cale) -295.01400 (VQA) -295 (dataset) -294.99500 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 534.32600 165.68600 Tm (3) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 539.30700 165.68600 Tm (\135\056) Tj -230.44500 -11.95510 Td [ (W) 79.98660 (e) -254.00200 (propose) -254.99700 (tw) 10.00810 (o) -253.99200 (simple) -254.98200 (b) 20.00160 (ut) -253.98700 (surprisingly) -253.98200 (ef) 25.00560 (fecti) 25.01540 (v) 14.98280 (e) -254.98200 <6d6f64690263612d> ] TJ 11.95510 TL T* [ (tions) -349.99100 (to) -348.98800 (the) -350.01300 (iBO) 34.99660 (WIMG) -350.00800 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 410.45100 141.77600 Tm (48) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R31 9.96260 Tf 1 0 0 1 420.41300 141.77600 Tm [ (\135) -349.00500 (VQA) -350.01000 (model) -349.99600 (that) -348.98600 (allo) 24.99090 (ws) -350.01000 (it) -350.00500 (to) ] TJ -111.55100 -11.95510 Td [ (mak) 10.01060 (e) -301.01300 (use) -299.98200 (of) -301.00600 (weak) -299.98900 (supervision) -301.00400 (in) -300.01900 (the) -300.99900 (form) -300.99900 (of) -299.98900 (images) -300.98900 (asso\055) ] TJ 11.95510 TL T* [ (ciated) -253.98900 (with) ] TJ /R33 9.96260 Tf 46.55980 0 Td [ (unanswer) 37.01590 (ed) ] TJ /R31 9.96260 Tf 50.57700 0 Td [ (questions\056) -321 (This) -254.01200 (pro) 14.98280 (v) 14.98280 (es) -253.99700 (our) -253.99200 (h) 4.98446 (ypoth\055) ] TJ -97.13670 -11.95470 Td [ (esis) -278.99800 (t) 0.98758 (hat) -279.00500 (unanswered) -279 (questions) -277.98800 (can) -279.01700 (be) -277.99300 (ef) 25.00560 (fecti) 25.01540 (v) 14.98280 (ely) -279.00200 (used) -277.98300 (as) -279.01200 (a) ] TJ 11.95630 TL T* [ (form) -249.98800 (of) -249.99500 (visual) -250.01000 (supervision\056) ] TJ 11.95510 -12.95510 Td [ (Inspired) -453.00500 (by) -451.99300 (the) -453.01500 (insights) -453.01300 (from) -451.99300 (our) -453.01800 (initial) -452.98800 (e) 15.01220 (xperiments\054) ] TJ -25.69610 -29.88790 Td (1) Tj ET Q Q Q q q 1 1 1 rg /a0 gs 48.40600 786.42200 515.18800 -52.69900 re f q /s5 gs /x6 Do Q q /s7 gs /x8 Do Q q /s9 gs /x10 Do Q q /s11 gs /x12 Do Q Q Q Q q 1 0 0 1 0 0 cm BT /F1 12 Tf 14.40000 TL ET 1 1 1 rg n 270 47 72 14 re f* 0.50000 0.50000 0.50000 rg BT /F2 9 Tf 10.80000 TL ET BT 1 0 0 1 299.25000 50 Tm (241) Tj T* ET Q endstream endobj 15 0 obj << /Filter /FlateDecode /Resources << /ExtGState << /a0 << /CA 1 /ca 1 >> >> /XObject << /x18 16 0 R >> >> /Length 28 /Group << /Type /Group /S /Transparency /CS /DeviceRGB /I true >> /BBox [ 78 746 96 765 ] /Type /XObject /Subtype /Form >> stream x+O4PH/VЯ0Pp 0 endstream endobj 16 0 obj << /Filter /FlateDecode /Resources 17 0 R /Length 107 /Type /XObject /BBox [ 78 746 96 765 ] /Subtype /Form >> stream xe AC̬wʠ =p,?]%+H-
Jc "82w8VSnGW;"
endstream
endobj
17 0 obj
<<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
>>
endobj
18 0 obj
<<
/Filter /FlateDecode
/Resources <<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
/XObject <<
/x21 19 0 R
>>
>>
/Length 28
/Group <<
/Type /Group
/S /Transparency
/CS /DeviceRGB
/I true
>>
/BBox [ 96 752 110 775 ]
/Type /XObject
/Subtype /Form
>>
stream
x+O4PH/VЯ02Tp
0w
endstream
endobj
19 0 obj
<<
/Filter /FlateDecode
/Resources 20 0 R
/Length 118
/Type /XObject
/BBox [ 96 752 110 775 ]
/Subtype /Form
>>
stream
xe=QSp#lZq|ddLB:D[c¸>*FI&(\T=s&GuJѩIY\I}#
.!.
endstream
endobj
20 0 obj
<<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
>>
endobj
21 0 obj
<<
/Filter /FlateDecode
/Resources <<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
/XObject <<
/x15 22 0 R
>>
>>
/Length 28
/Group <<
/Type /Group
/S /Transparency
/CS /DeviceRGB
/I true
>>
/BBox [ 67 752 84 775 ]
/Type /XObject
/Subtype /Form
>>
stream
x+O4PH/VЯ04Up
0
endstream
endobj
22 0 obj
<<
/Filter /FlateDecode
/Resources 23 0 R
/Length 228
/Type /XObject
/BBox [ 67 752 84 775 ]
/Subtype /Form
>>
stream
xeQKn!s ?FPav6R٪TS.
b];15YyR
{7QL.\:Rv/x9l+L7h%1!}i/AI(kz"U&,YO![R hg{3}4/GyYF:!w}Gn+'xJcO9i뽼_-:`
endstream
endobj
23 0 obj
<<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
>>
endobj
24 0 obj
<<
/Filter /FlateDecode
/Resources <<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
/XObject <<
/x24 25 0 R
>>
>>
/Length 28
/Group <<
/Type /Group
/S /Transparency
/CS /DeviceRGB
/I true
>>
/BBox [ 132 751 480 772 ]
/Type /XObject
/Subtype /Form
>>
stream
x+O4PH/VЯ02Qp
0
endstream
endobj
25 0 obj
<<
/Filter /FlateDecode
/Resources 26 0 R
/Length 53223
/Type /XObject
/BBox [ 132 751 480 772 ]
/Subtype /Form
>>
stream
xtI:6%Q㨈?7rA= u%6 ?Y(WbWo{B>9
x`Znϳ|8{3?0x*z ǃ|,@:w>`c|*ϻⳅKO3`g
:_|}}><.6`Z{{3]#<_o"~:ͺgk7/Ұ@|K yp ]03ʷCmş8˽Y?>(3!Bwqs.Z8,~~=rMT̩y+/*w: uBZ_`ߵp`%M?ɝ1ɳw=vDۉy&xb4Q>d@ sg~lA