%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 (Learning to Disambiguate by Asking Discriminative Questions) /Producer (PyPDF2) /Author (Yining Li\054 Chen Huang\054 Xiaoou Tang\054 Chen Change Loy) /Subject (2017 IEEE International Conference on Computer Vision) >> endobj 3 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 14 0 R /Resources << /XObject << /x8 15 0 R /R25 18 0 R /x6 19 0 R /x12 22 0 R /R24 25 0 R /x10 26 0 R >> /ExtGState << /s9 29 0 R /R13 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 50 0 R /R14 51 0 R /R16 54 0 R /R18 58 0 R /R20 62 0 R /R22 65 0 R /R32 69 0 R /R26 72 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 75 0 R 76 0 R 77 0 R 78 0 R 79 0 R ] >> endobj 4 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 80 0 R /Resources << /ExtGState << /R13 32 0 R >> /Font << /F2 81 0 R /F1 82 0 R /R14 51 0 R /R16 54 0 R /R22 65 0 R /R87 83 0 R /R85 88 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 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 ] >> endobj 5 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 136 0 R /Resources << /ExtGState << /R13 32 0 R >> /Font << /F2 137 0 R /R126 138 0 R /F1 142 0 R /R14 51 0 R /R16 54 0 R /R22 65 0 R /R87 83 0 R /R85 88 0 R /R128 143 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 147 0 R 148 0 R 149 0 R 150 0 R 151 0 R 152 0 R 153 0 R 154 0 R 155 0 R 156 0 R 157 0 R 158 0 R 159 0 R 160 0 R 161 0 R 162 0 R 163 0 R 164 0 R 165 0 R 166 0 R 167 0 R 168 0 R 169 0 R 170 0 R 171 0 R 172 0 R 173 0 R 174 0 R 175 0 R 176 0 R 177 0 R ] >> endobj 6 0 obj << /Parent 1 0 R /Rotate 0 /Contents 178 0 R /Resources << /XObject << /R163 179 0 R /R157 188 0 R /R154 189 0 R /R155 190 0 R /R167 191 0 R /R156 196 0 R /R164 197 0 R /R147 198 0 R /R158 199 0 R /R146 200 0 R /R152 201 0 R /R153 202 0 R >> /ExtGState << /R145 180 0 R /R13 32 0 R /R160 181 0 R /R159 182 0 R >> /Font << /R161 183 0 R /R165 192 0 R /R174 210 0 R /R170 215 0 R /R168 218 0 R /R149 203 0 R /R172 223 0 R /F2 228 0 R /F1 229 0 R /R14 51 0 R /R16 54 0 R /R87 83 0 R /R85 88 0 R /R26 72 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /Group 187 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 230 0 R 231 0 R 232 0 R 233 0 R 234 0 R 235 0 R 236 0 R 237 0 R 238 0 R ] /Type /Page >> endobj 7 0 obj << /Parent 1 0 R /Rotate 0 /Contents 239 0 R /Resources << /XObject << /R192 240 0 R /R193 242 0 R /R191 241 0 R /R185 243 0 R /R214 244 0 R /R204 250 0 R /R186 251 0 R /R25 18 0 R /R24 25 0 R >> /ExtGState << /R145 180 0 R /R13 32 0 R /R213 245 0 R /R160 181 0 R >> /Font << /R211 273 0 R /R201 259 0 R /R197 266 0 R /R194 246 0 R /R205 277 0 R /R188 252 0 R /R209 280 0 R /F2 283 0 R /F1 284 0 R /R87 83 0 R /R85 88 0 R /R207 285 0 R /R161 183 0 R /R174 210 0 R /R168 218 0 R /R170 215 0 R /R172 223 0 R /R14 51 0 R /R16 54 0 R /R26 72 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /Group 187 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 289 0 R 290 0 R 291 0 R ] /Type /Page >> endobj 8 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 292 0 R /Resources << /ExtGState << /R13 32 0 R >> /Font << /R211 273 0 R /R168 218 0 R /R207 285 0 R /R174 210 0 R /R209 280 0 R /R170 215 0 R /F2 293 0 R /F1 294 0 R /R14 51 0 R /R16 54 0 R /R22 65 0 R /R87 83 0 R /R85 88 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 295 0 R 296 0 R 297 0 R 298 0 R 299 0 R 300 0 R 301 0 R 302 0 R 303 0 R 304 0 R 305 0 R 306 0 R 307 0 R ] >> endobj 9 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 308 0 R /Resources << /ExtGState << /R13 32 0 R >> /Font << /R211 273 0 R /R168 218 0 R /R207 285 0 R /R174 210 0 R /R209 280 0 R /R172 223 0 R /F2 309 0 R /R126 138 0 R /F1 310 0 R /R14 51 0 R /R16 54 0 R /R254 311 0 R /R256 314 0 R /R87 83 0 R /R85 88 0 R /R22 65 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /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 326 0 R 327 0 R 328 0 R 329 0 R 330 0 R 331 0 R ] >> endobj 10 0 obj << /Parent 1 0 R /Rotate 0 /Contents 332 0 R /Resources << /XObject << /R282 333 0 R /R283 334 0 R /R280 337 0 R /R281 338 0 R /R269 340 0 R /R279 335 0 R /R278 336 0 R /R273 341 0 R /R272 342 0 R /R271 343 0 R /R270 344 0 R /R277 339 0 R /R276 345 0 R >> /ExtGState << /R145 180 0 R /R13 32 0 R >> /Font << /R161 183 0 R /R194 246 0 R /R168 218 0 R /R149 203 0 R /F2 350 0 R /F1 351 0 R /R14 51 0 R /R16 54 0 R /R274 346 0 R /R87 83 0 R /R85 88 0 R /R26 72 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /Group 187 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 352 0 R 353 0 R 354 0 R 355 0 R 356 0 R 357 0 R ] /Type /Page >> endobj 11 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 358 0 R /Resources << /ExtGState << /R13 32 0 R >> /Font << /F2 359 0 R /F1 360 0 R /R22 65 0 R /R14 51 0 R /R16 54 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 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 395 0 R 396 0 R 397 0 R 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 ] >> endobj 12 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 412 0 R /Resources << /ExtGState << /R13 32 0 R >> /Font << /F2 413 0 R /F1 414 0 R /R22 65 0 R /R16 54 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 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 ] >> endobj 13 0 obj << /Type /Catalog /Pages 1 0 R >> endobj 14 0 obj << /Length 19862 >> stream q q q 0.10000 0 0 0.10000 0 0 cm /R13 gs 0 g q 10 0 0 10 0 0 cm BT /R14 14.34620 Tf 1 0 0 1 106.17100 675.06700 Tm [ (Lear) 14.98930 (ning) -249.99400 (to) -250.01200 (Disambiguate) -250.01200 (by) -250.01300 (Asking) -249.99600 (Discriminati) 10 (v) 9.99625 (e) -249.99800 (Questions) ] TJ /R16 11.95520 Tf 34.08200 -37.85820 Td [ (Y) 54.99150 (ining) -250.00500 (Li) ] TJ /R16 7.97010 Tf 46.16990 4.33906 Td (1) Tj /R16 11.95520 Tf 22.41600 -4.33906 Td [ (Chen) -249.99500 (Huang) ] TJ /R16 7.97010 Tf 60.09770 4.33906 Td (2) Tj /R16 11.95520 Tf 22.41600 -4.33906 Td [ (Xiaoou) -249.98700 (T) 80.01470 (ang) ] TJ /R16 7.97010 Tf 61.79610 4.33906 Td (1) Tj /R16 11.95520 Tf 22.41600 -4.33906 Td [ (Chen) -249.99500 (Change) -250.01400 (Lo) 10.01460 (y) ] TJ /R16 7.97010 Tf 86.87810 4.33906 Td (1) Tj -351.89800 -13.95900 Td (1) Tj /R16 11.95520 Tf 4.48320 -4.33789 Td [ (Department) -250 (of) -250.01500 (Information) -250.00400 (Engineering\054) -249.98500 (The) -250.01400 (Chinese) -250.01200 (Uni) 24.99570 (v) 14.98510 (ersity) -249.98900 (of) -250.01400 (Hong) -249.98900 (K) 34.99390 (ong) ] TJ /R16 7.97010 Tf 74.24180 -9.62031 Td (2) Tj /R16 11.95520 Tf 4.48320 -4.33789 Td [ (Robotics) -250.00600 (Institute\054) -250.00400 (Carne) 15.01370 (gie) -250 (Mellon) -249.99700 (Uni) 24.99570 (v) 14.98510 (ersity) ] TJ /R18 8.96640 Tf -48.09610 -13.94800 Td (f) Tj /R20 8.96640 Tf 4.60781 0 Td [ (ly015\054) -600.01200 (xtang\054) -600.01000 (ccloy) ] TJ /R18 8.96640 Tf 102.21600 0 Td (g) Tj /R20 8.96640 Tf 4.60820 0 Td [ (\100ie\056cuhk\056edu\056hk\054) -600.01600 (chenh2\100andrew\056cmu\056edu) ] TJ /R14 11.95520 Tf -111.09400 -41.04610 Td (Abstract) Tj /R22 9.96260 Tf -83.92810 -23.91680 Td [ (The) -290.01500 (ability) -289.99600 (to) -290.01800 (ask) -289.98300 (questions) -289.98800 (is) -290.01700 (a) -291.00500 (powe) 1 (rful) -290.98600 (tool) -290.00600 (to) -290.01800 (gather) ] TJ -11.95470 -11.95510 Td [ (information) -328 (in) -328.01100 (or) 36.98520 (der) -328.99700 (to) -328.01100 (learn) -328.00500 (about) -327.99700 (the) -328.01600 (world) -329.00600 (and) -328.00900 (r) 37.01830 (esolve) ] TJ 11.95590 TL T* [ (ambiguities\056) -566.00700 (In) -335.98200 (this) -335.01600 (paper) 111.01800 (\054) -356.99700 (we) -336.01400 (e) 19.99180 (xplor) 36.99260 (e) -335.00800 (a) -336.01700 (no) 10.00810 (vel) -335.00500 (pr) 44.98510 (oblem) -336.00800 (of) ] TJ 11.95510 TL T* [ (g) 10.00320 (ener) 15.01960 (ating) -403.00100 (discriminative) -403.99800 (questions) -402.98900 (to) -403.01900 (help) -403.99900 (disambiguate) ] TJ T* [ (visual) -490.99000 (instances\056) -1035.01000 (Our) -491.98300 (work) -490.99600 (can) -491.99000 (be) -491.01500 (seen) -491.99400 (as) -491.00400 (a) -491.99100 (comple\055) ] TJ T* [ (ment) -459.01100 (and) -459.00600 (ne) 15.01830 (w) -459.98800 (e) 19.99180 (xtension) -458.99400 (to) -459.00900 (the) -459.01400 (ric) 14.98650 (h) -459.01700 (r) 37.01830 (esear) 36.98160 (c) 15.01220 (h) -459.99700 (st) 0.99248 (ud) -1.00964 (i) 0.98758 (es) -460 (on) ] TJ T* [ (ima) 10.01360 (g) 10.00320 (e) -398.01600 (captioning) -398.98400 (and) -397.99700 (question) -398.99600 (answering) 14.98770 (\056) -755.00900 (W) 91.98710 (e) -398.99700 (intr) 44.99740 (oduce) ] TJ T* [ (the) -314.01800 <0272> 10.00940 (st) -314.01300 (lar) 36.99260 (g) 10.00320 (e\055scale) -315.00400 (dataset) -313.99500 (with) -313.99300 (o) 10.00320 (ver) -313.98900 (10\054000) -314 (car) 36.98160 (efully) -314.98900 (an\055) ] TJ 11.95590 TL T* [ (notated) -303.01700 (ima) 10.01360 (g) 10.00320 (es\055question) -303.01300 (tuples) -302.98300 (to) -302.99600 (facilitate) -303.00300 (benc) 15.01960 (hmarking) 15.00980 (\056) ] TJ 11.95510 TL T* [ (In) -358.01700 (particular) 111.01100 (\054) -386.01200 (eac) 15.01340 (h) -358.01400 (tuple) -358 (consists) -358.02000 (of) -358.98700 (a) -358.01400 (pair) -357.99700 (of) -358.00600 (ima) 10.01300 (g) 10.00320 (es) -358.99600 (and) ] TJ T* [ (4\0566) -332.01100 (discriminative) -332.01000 (questions) -333 (\050as) -332.01600 (positive) -331.99400 (samples\051) -331.99900 (and) -332.98900 (5\0569) ] TJ T* [ (non\055discriminative) -329.99800 (questions) -329.98200 (\050as) -328.99700 (ne) 39.99460 (gative) -330.00900 (samples\051) -329.99900 (on) -330.01400 (av\055) ] TJ T* [ (er) 15.01890 (a) 10.00320 (g) 10.00320 (e) 15.01280 (\056) -580.01800 (In) -339.98100 (addition\054) -362.00800 (we) -340.01300 (pr) 36.98520 (esent) -340.01200 (an) -340.01200 (ef) 18 (fective) -340.01700 (method) -340.00200 (for) -340.00500 (vi\055) ] TJ T* [ (sual) -207.98100 (discriminative) -207.99100 (question) -208.98800 (g) 10.00320 (ener) 15.01960 (ation\056) -295.99700 (The) -207.99000 (method) -207.98500 (can) -209.01900 (be) ] TJ 11.95630 TL T* [ (tr) 14.99140 (ained) -262.99700 (in) -263.00300 (a) -263.01000 (weakly) -263.00300 (supervised) -262.98800 (manner) -262.99500 (without) -263.00500 (discrimina\055) ] TJ 11.95470 TL T* [ (tive) -253.01100 (ima) 10.01360 (g) 10.00320 (es\055question) -254.00200 (tuples) -252.99100 (b) 20.00160 (ut) -253.98500 (just) -252.99100 (e) 19.99180 (xisting) -252.98200 (visual) -254.00900 (question) ] TJ T* [ (answering) -364.98900 (datasets\056) -654.98900 (Pr) 44.98270 (omising) -365 (r) 37.01960 (esults) -365.01000 (ar) 36.98650 (e) -365.00300 (shown) -365.00300 (a) 10.00320 (gainst) ] TJ T* [ (r) 37.01960 (epr) 36.98100 (esentative) -461.98000 (baselines) -462.00700 (thr) 44.99130 (ough) -463.00100 (quantitativ) 0.99738 (e) -462.98600 (e) 15.01220 (valuations) ] TJ T* [ (and) -249.98200 (user) -249.98500 (studies\056) ] TJ /R14 11.95520 Tf 35.88590 TL T* [ (1\056) -249.99000 (Intr) 18.01460 (oduction) ] TJ /R16 9.96260 Tf 11.95470 -18.93590 Td [ (Imagine) -336.00500 (a) -335.98800 (natural) -336.01000 (language) -336.00800 (dialog) -337.01700 (bet) 0.98758 (ween) -337.00300 (a) -335.98800 (computer) ] TJ -11.95470 -11.95510 Td [ (and) -249.99200 (a) -250.00200 (human) -250.01400 (\050see) -250.00400 (Fig\056) ] TJ ET Q 1 0 0 rg q 10 0 0 10 0 0 cm BT /R16 9.96260 Tf 1 0 0 1 140.31400 248.41200 Tm (1) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R16 9.96260 Tf 1 0 0 1 145.29500 248.41200 Tm (\051\072) Tj -83.22810 -11.96210 Td [ (Kid) -2702.98000 (\072) ] TJ /R22 9.96260 Tf 47.73130 0 Td [ (\223What) -250.00400 (sport) -250.00700 (is) -249.98400 (the) -249.99000 (man) -249.98500 (playing\077\224\056) ] TJ /R16 9.96260 Tf -47.73130 -11.96090 Td [ (Computer) -228 (\072) ] TJ /R22 9.96260 Tf 47.97620 0 Td [ (\223What) -250.00400 (is) -249.98400 (the) -249.99000 (color) -250.01800 (of) -249.98500 (his) -250.02000 (shirt\077\224) ] TJ /R16 9.96260 Tf -47.97620 -11.96170 Td [ (Kid) -2702.98000 (\072) ] TJ /R22 9.96260 Tf 47.73130 0 Td [ (\223Blue) 15.00360 (\056) 140.01800 (\224) ] TJ /R16 9.96260 Tf -47.73130 -11.96210 Td [ (Computer) -228 (\072) ] TJ /R22 9.96260 Tf 47.97620 0 Td [ (\223He) -249.98500 (is) -249.98400 (playing) -250.00600 (tennis\056) 140.00500 (\224) ] TJ /R16 9.96260 Tf -59.93090 -11.96210 Td [ (In) -329.98200 (the) -330.99600 (con) 39.99880 (v) 14.98280 (ersation\054) -350 (the) -330.99600 (kid) -330.00600 (refers) -331.01600 (to) -330.01100 (an) -329.98400 (object) -330.98900 (b) 20.00160 (ut) -330.01100 (the) -330.99600 (de\055) ] TJ 11.95510 TL T* [ (scription) -258.98100 (is) -259.00200 (linguistically) -258.98700 <756e64657273706563690265642e> -337.00700 (The) -259.99100 (computer) -258.99400 (an\055) ] TJ T* [ (alyzes) -202 (the) -201.99800 (visual) -202.01900 (dif) 24.98480 (ferences) -202.01800 (of) -202.00400 (plausible) -202.01000 (tar) 17.99700 (gets) -203.01000 (and) -202 (then) -201.99300 (re\055) ] TJ T* [ (acts) -263.01100 (by) -263.00500 (asking) -262.99000 (a) -262 (discriminati) 24.98970 (v) 14.98280 (e) -263.02000 (question) ] TJ /R22 9.96260 Tf 161.97300 0 Td [ (\223What) -262.98100 (is) -263.00300 (the) -263.00800 (color) ] TJ -161.97300 -11.95630 Td [ (of) -249.98500 (his) -250.01800 (shirt\077\224) ] TJ /R16 9.96260 Tf 56.23790 0 Td [ (to) -249.98500 (resolv) 14.99380 (e) -250.00200 (the) -249.99000 (reference\056) ] TJ -44.28320 -11.96090 Td [ (W) 79.98730 (e) -412.99400 <6465026e65> -413.99300 (the) -412.98300 (aforementioned) -413.98100 (problem) -413.00200 (as) ] TJ /R22 9.96260 Tf 178.29400 0 Td [ (V) 73.99260 (isual) -413 (Dis\055) ] TJ -190.24900 -11.95590 Td [ (criminative) -212.98800 (Question) -214.00800 (Gener) 14.98280 (ation) ] TJ /R16 9.96260 Tf 132.96500 0 Td [ (\050VDQG\051\056) -213.99600 (Spec) 0.99003 <690263616c6c79> 65.00140 (\054) -221.01900 (the) ] TJ -132.96500 -11.95510 Td [ (computer) -246.01500 (is) -245.98500 (gi) 24.98850 (v) 14.98280 (en) -245.99700 (with) -246.01500 (tw) 10.00810 (o) -245.01300 (visual) -246.01200 (instances\054) -246.99100 (and) -245.99300 (the) -245.99100 (goal) -245.98600 (is) ] TJ T* [ (to) -243.00600 (ask) -243.01200 (a) -242.98300 (good) -242.01800 (question) -242.98300 (to) -243.00600 (distinguish) -242.99100 (or) -243.01600 (disambiguate) -243.01100 (them\056) ] TJ T* [ (In) -366.99600 (this) -367.01100 (study) 64.99770 (\054) -396.01000 (we) -366.99900 (call) -366.01100 (the) -366.99000 (pair) -366.99300 (images) -367.01500 (as) -366.99500 (ambiguous) -366.99000 (pairs) ] TJ ET Q q 3089.58000 4141.33000 2359.59000 1500.29000 re W* n q 1163.06000 0 0 957.16900 3090.53000 4684.45000 cm /R24 Do Q 1 1 0 rg 3501.81000 5612.93000 m 4074.73000 5612.93000 l 4074.73000 4695.93000 l 3501.81000 4695.93000 l h 3509.46000 4699.75000 m 3505.64000 4703.57000 l 4070.91000 4703.57000 l 4067.08000 4699.75000 l 4067.08000 5609.11000 l 4070.91000 5605.29000 l 3505.64000 5605.29000 l 3509.46000 5609.11000 l f q 1159.23000 0 0 957.16900 4288.03000 4684.45000 cm /R25 Do Q 4406.63000 5597.64000 m 5415.70000 5597.64000 l 5415.70000 4695.93000 l 4406.63000 4695.93000 l h 4414.28000 4699.75000 m 4410.45000 4703.57000 l 5411.88000 4703.57000 l 5408.05000 4699.75000 l 5408.05000 5593.81000 l 5411.88000 5589.98000 l 4410.45000 5589.98000 l 4414.28000 5593.81000 l f 0.57227 0.81641 0.31396 rg 3090.39000 4548.81000 305.25000 119.33600 re f* 3395.64000 4548.81000 2051.68000 119.33600 re f* 3090.39000 4429.47000 305.25000 119.33600 re f* 3395.64000 4429.47000 2051.68000 119.33600 re f* 0.95703 0.69336 0.51367 rg 3090.39000 4310.14000 305.25000 119.33600 re f* 3395.64000 4310.14000 2051.68000 119.33600 re f* 3090.39000 4190.80000 305.25000 119.33600 re f* 3395.64000 4190.80000 2051.68000 119.33600 re f* 1.27512 w 1 j 0 G q 1.00013 0 0 0.99987 0 0 cm 3395.20000 4669.39000 m 3395.20000 4190.72000 l S Q q 1.00013 0 0 0.99987 0 0 cm 3089.35000 4549.41000 m 5447.25000 4549.41000 l S Q q 1.00013 0 0 0.99987 0 0 cm 3089.35000 4430.05000 m 5447.25000 4430.05000 l S Q q 1.00013 0 0 0.99987 0 0 cm 3089.35000 4310.70000 m 5447.25000 4310.70000 l S Q q 1.00013 0 0 0.99987 0 0 cm 3089.99000 4669.39000 m 3089.99000 4190.72000 l S Q q 1.00013 0 0 0.99987 0 0 cm 5446.61000 4669.39000 m 5446.61000 4190.72000 l S Q q 1.00013 0 0 0.99987 0 0 cm 3089.35000 4668.76000 m 5447.25000 4668.76000 l S Q q 1.00013 0 0 0.99987 0 0 cm 3089.35000 4191.36000 m 5447.25000 4191.36000 l S Q Q q 3089.58000 4222.60000 307.02000 771.66800 re W* n q 10 0 0 10 0 0 cm BT /R26 9.18602 Tf 1.00026 0 0 1 309.95900 457.39400 Tm [ (G) -1.98216 (o) 4.01480 (o) 4.01480 (d) ] TJ ET Q Q q 3394.69000 4222.60000 2053.53000 771.66800 re W* n q 10 0 0 10 0 0 cm BT /R29 9.18602 Tf 1.00026 0 0 1 340.49600 457.39400 Tm [ <003a> -13.17630 <004b> 4.18485 <0044> 4.18485 <0057> -2.16018 <0003> 6.81002 <00460052> 4.18485 <004f> 3.18580 <0052> 4.18485 <00550003> -2.16018 <004c> 3.18580 <00560003> -2.16018 <0057> -2.16018 <004b> 4.18485 <0048> 4.18485 <0003> -2.15752 <0050> 9.99582 <0044> 4.18220 <0051> 4.18220 <00b6> 45.19110 <00560003> -44.16020 <0056004b> 4.18220 <004c> 3.18846 <00550057> -2.15752 <0022> ] TJ ET Q Q q 3089.58000 4141.33000 307.02000 733.40200 re W* n q 10 0 0 10 0 0 cm BT /R26 9.18602 Tf 1.00026 0 0 1 309.95900 445.45400 Tm [ (G) -1.98216 (o) 4.01480 (o) 4.01480 (d) ] TJ ET Q Q q 3394.69000 4141.33000 2053.53000 733.40200 re W* n q 10 0 0 10 0 0 cm BT /R32 9.18602 Tf 1.00026 0 0 1 340.49600 445.45400 Tm [ (W) -13.00630 (h) 4.01480 (a) 4.01480 (t) -1.99013 ( ) 6.98007 (sp) 3.97760 (o) 4.01480 (rt) -1.99279 ( ) -1.99013 (i) 3.01575 (s ) -1.98216 (t) -1.99013 (h) 4.01480 (e) 4.01480 ( ) -1.98747 (m) 9.99582 (a) 4.01215 (n) 4.01215 ( ) -11.97800 (p) 4.01215 (l) 3.01841 (a) 4.01215 (yi) 2.98121 (n) 4.01215 (g) 4.01215 (\077) ] TJ ET Q Q q 3089.58000 4141.33000 307.02000 614.83600 re W* n q 10 0 0 10 0 0 cm BT /R26 9.18602 Tf 1.00026 0 0 1 309.95900 433.51100 Tm [ (Ba) 5.00854 (d) ] TJ ET Q Q q 3394.69000 4141.33000 2053.53000 614.83600 re W* n q 10 0 0 10 0 0 cm BT /R32 9.18602 Tf 1.00026 0 0 1 340.49600 433.51100 Tm [ (H) 2.98121 (o) 4.01480 (w) 2.98121 ( ) -11.98060 (m) 9.99582 (a) 4.01480 (n) 4.01480 (y ) -12.01520 (p) 4.01480 (e) 4.01480 (o) 4.01480 (p) 4.01480 (l) 3.01575 (e) 4.01215 ( ) -22.98880 (a) 4.01215 (re) 4.01215 ( ) -1.98747 (t) -1.98747 (h) 4.01215 (e) 4.01215 (re) 4.01215 (\077) ] TJ ET Q Q q 3089.58000 4141.33000 307.02000 495.30900 re W* n q 10 0 0 10 0 0 cm BT /R26 9.18602 Tf 1.00026 0 0 1 309.95900 421.57100 Tm [ (Ba) 5.00854 (d) ] TJ ET Q Q q 3394.69000 4141.33000 2053.53000 495.30900 re W* n q 10 0 0 10 0 0 cm BT /R32 9.18602 Tf 1.00026 0 0 1 340.49600 421.57100 Tm [ (W) -13.00630 (h) 4.01480 (e) 4.01480 (re) 4.01215 ( ) 6.98007 (i) 3.01575 (s ) -1.98216 (t) -1.99013 (h) 4.01480 (e) 4.01480 ( ) -1.99013 (m) 9.99582 (a) 4.01480 (n) 4.01215 (\077) ] TJ ET Q Q q 10 0 0 10 0 0 cm BT /R16 9.96260 Tf 1 0 0 1 308.86200 401.28000 Tm [ (Figure) -371.99200 (1\072) -554.99300 (Example) -371.99200 (ambiguous) -372.01100 (image) -372.00200 (pair) -372.01100 (and) -372.01100 (both) -373.01100 (good) ] TJ 11.95510 TL T* [ (and) -249.99300 (bad) -249.99300 (discriminati) 24.98850 (v) 14.98280 (e) -250.00200 (questions\056) ] TJ 31.50590 TL T* [ (\226) -309.98300 (the) -311 (ambiguity) -310.01700 (may) -310.99500 (not) -310.01000 (necessarily) -310.99200 (be) -309.98700 (due) -311.00200 (to) -310.01700 (their) -310.99700 (subtle) ] TJ 11.95510 TL T* [ (visual) -343.99400 (dif) 24.98600 (ferences\056) -591.00900 (The) 14.98280 (y) -342.99700 (may) -344.00600 (just) -343.99700 (belong) -343.99700 (to) -342.99200 (the) -344.01100 (same) -344.01100 (ob\055) ] TJ T* [ (ject) -267.00900 (class) -265.98500 (with) -266.99200 (close) -267.00400 (proximity) -267.00400 (in) -265.98500 (their) -267.00400 (deep) -267.01400 (representation\056) ] TJ T* [ (Although) -255.01400 (such) -253.98500 (ambiguity) -255.00600 (can) -254.00200 (be) -255.01600 (easily) -255.00200 (r) 1.01454 (esolv) 14.99750 (ed) -255.01600 (by) -255.00600 (human\054) ] TJ 11.95590 TL T* [ (the) 14.98520 (y) -370.99200 (can) -371.98000 (be) -370.99700 (dif) 24.98600 <0263756c74> -371.00700 (to) -372.00400 (a) -371.00200 (machine\056) -673.98600 (Distinguishing) -372.01100 (dif) 24.98600 (fer) 19.98690 (\055) ] TJ 11.95510 TL T* [ (ent) -283.98400 (image) -284.01600 (pairs) -283.98200 (require) -283.00900 (asking) -284.00700 (dif) 24.98600 (ferent) -283.99700 (types) -284.01100 (of) -283.99200 (questions\054) ] TJ T* [ (ranging) -320.00500 (from) -320.01500 (color) 40.00310 (\054) -337 (action\054) -338.00200 (location\054) -337.01200 (and) -320.02000 (number) 54.99080 (\056) -519.98700 (Akin) -320.00500 (to) ] TJ T* [ (the) -243.01100 (classic) -243.00400 (\223T) 79.98660 (wenty) -241.98400 (Questions\224) -242.98400 (g) 4.98446 (ame\054) -243.99400 (a) -242.98400 (careful) -242.98400 (selection) -242.99400 (of) ] TJ T* [ (questions) -261.99300 (can) -261.99800 (greatly) -261.98300 (impro) 15.00240 (v) 14.98280 (e) -262 (the) -261.98600 (odds) -262.01500 (of) -261.99500 (the) -261.98600 (questioner) -262.00500 (to) ] TJ T* [ (narro) 25.01540 (w) -212.00100 (do) 24.98600 (wn) -211.99600 (the) -211.99600 (ans) 0.98758 (wer) 55.00550 (\056) -297.98700 (A) -211.99900 (bad) -211.01900 (question) -212.00900 (w) 10.00320 (ould) -211.97900 (f) 9.99343 (ail) -211.99400 (to) -211.99400 (elim\055) ] TJ 11.95590 TL T* [ (inate) -235.01000 (ambiguities\056) -304.99100 (Figure) ] TJ ET Q 1 0 0 rg q 10 0 0 10 0 0 cm BT /R16 9.96260 Tf 1 0 0 1 410.95900 238.26700 Tm (1) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R16 9.96260 Tf 1 0 0 1 418.28200 238.26700 Tm [ (gi) 24.98600 (v) 14.98280 (es) -235.02000 (good) -236.01900 (and) -235.01500 (bad) -235.01500 (e) 15.01220 (xamples) -236.01500 (of) ] TJ -109.42000 -11.95510 Td [ (questions\056) -378.99100 (This) -272.01100 (questioning) -272.99900 (capability) -272.98400 (can) -273.01800 (subsequently) -273.00300 (be) ] TJ 11.95550 TL T* [ (e) 15.01220 (xtended) -332.98400 (to) -332.99100 (generating) -332.00600 (a) -333.00800 (sequence) -332.99800 (of) -333.00300 (discriminati) 24.99090 (v) 14.98280 (e) -333.00800 (ques\055) ] TJ T* [ (tions) -335.99300 (and) -336.99800 (prompting) -336.01300 (a) -337.00700 (human\055in\055the\055loop) -336.00800 (to) -336.01300 (answer) -337.00300 (them\056) ] TJ T* [ (In) -303.00600 (the) -303.98100 (process\054) -316.01300 (the) -303.98100 (machine) -302.99400 (accumulates) -302.98900 (e) 25.01050 (vidence) -302.99800 (that) -304.01300 (can) ] TJ T* [ (gradually) -294.98800 <7265026e65> -295.00200 (the) -295.98200 (language) -294.99500 (e) 15.01220 (xpression) -295.01900 (from) -295 (humans) -296 (and) ] TJ 11.95630 TL T* [ <026e616c6c79> -250.00200 (distinguish) -250.01000 (the) -249.99000 (object) -249.98300 (of) -249.99700 (interest\056) ] TJ 11.95510 -13.80470 Td [ (Such) -405.99100 (VDQG) -405.99800 (ability) -404.99600 (allo) 24.99090 (ws) -406 (a) -406.01600 (machine) -405.99600 (to) -406 (play) -405.99600 (a) -406.01600 (more) ] TJ -11.95510 -11.95510 Td [ (natural) -227.98900 (and) -229.01600 (interacti) 24.99340 (v) 14.98280 (e) -228.00600 (role) -227.99600 (in) -227.99100 (Human\055Computer) -229.00100 (Interaction) ] TJ 11.95510 TL T* [ (\050HCI\051\054) -263.00500 (or) -261.99300 (impro) 15.00480 (v) 14.98280 (e) -263.02000 (a) -262 (robot) -262.99500 (to) -263.00500 (bind) -262.01000 (the) -263.00500 (references) -262.01500 (made) -263.00500 (by) -263.00500 (a) ] TJ 11.95590 TL T* [ (speak) 10.00570 (er) -291.01800 (more) -291 (accurately) -289.99100 (to) -291 (objects) -290.99100 (in) -291 (a) -291.01500 (scene\056) -432.00600 (While) -290.99600 (there) ] TJ 11.95510 TL T* [ (ha) 19.99670 (v) 14.98280 (e) -307.01200 (been) -307.98500 (v) 24.98110 (arious) -306.99300 (attempts) -307.99300 (to) -306.99800 (b) 20.00160 (uild) -308.00300 (a) -307.01200 (system) -307.99300 (that) -306.99300 (can) -307.99300 (pro\055) ] TJ T* [ (vide) -350.00800 (e) 15.01220 (xplanations) -350.01500 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R16 9.96260 Tf 1 0 0 1 386.52100 92.95510 Tm (14) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R16 9.96260 Tf 1 0 0 1 396.48300 92.95510 Tm [ (\135) -349.98300 (or) -349 (ask) -350.01500 (questions) -350.01500 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R16 9.96260 Tf 1 0 0 1 476.27400 92.95510 Tm (30) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R16 9.96260 Tf 1 0 0 1 486.23600 92.95510 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R16 9.96260 Tf 1 0 0 1 492.21400 92.95510 Tm (33) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R16 9.96260 Tf 1 0 0 1 502.17700 92.95510 Tm [ (\135) -349.98600 (based) -350.01000 (on) ] TJ -193.31400 -11.95510 Td [ (visual) -371.00900 (instances\054) -401.98400 (the) -370.99000 (problem) -371.01200 (of) -372.01600 (VDQG) -370.98700 (has) -370.99200 (not) -371.01700 (been) -372.01600 (e) 15.01220 (x\055) ] TJ -13.74100 -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 1 297 50 Tm (3419) 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 /DCTDecode
/Interpolate true
/Height 331
/Length 27112
/ColorSpace /DeviceRGB
/Width 401
/BitsPerComponent 8
/Subtype /Image
>>
stream
Adobe d C
$, !$4.763.22:ASF:=N>22HbINVX]^]8EfmeZlS[]Y C**Y;2;YYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYY K"
} !1AQa"q2#BR$3br
%&'()*456789:CDEFGHIJSTUVWXYZcdefghijstuvwxyz
w !1AQaq"2B #3Rbr
$4%&'()*56789:CDEFGHIJSTUVWXYZcdefghijstuvwxyz ? !P|}}5j+J