%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 (Visual7W\072 Grounded Question Answering in Images) /Producer (PyPDF2) /Author (Yuke Zhu\054 Oliver Groth\054 Michael Bernstein\054 Li Fei\055Fei) /Subject (2016 IEEE Conference on Computer Vision and Pattern Recognition) >> endobj 3 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 14 0 R /Resources << /XObject << /x8 15 0 R /R69 18 0 R /R70 23 0 R /R71 24 0 R /R76 25 0 R /R77 26 0 R /x6 27 0 R /R67 30 0 R /R60 31 0 R /R90 32 0 R /x12 33 0 R /R68 36 0 R /x10 37 0 R >> /ExtGState << /s9 40 0 R /R59 43 0 R /R47 44 0 R /R56 45 0 R /s5 46 0 R /s7 49 0 R /R66 21 0 R /s11 52 0 R /a0 << /CA 1 /ca 1 >> >> /Font << /R79 55 0 R /R48 62 0 R /R54 65 0 R /R73 69 0 R /R50 76 0 R /R52 80 0 R /F2 84 0 R /F1 85 0 R /R62 86 0 R /R83 93 0 R /R87 100 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R58 19 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ 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 143 0 R 144 0 R ] >> endobj 4 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 145 0 R /Resources << /ExtGState << /R47 44 0 R >> /Font << /R54 65 0 R /F2 146 0 R /F1 147 0 R /R50 76 0 R /R48 62 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 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 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 190 0 R 191 0 R 192 0 R 193 0 R 194 0 R 195 0 R 196 0 R 197 0 R 198 0 R 199 0 R 200 0 R 201 0 R 202 0 R 203 0 R 204 0 R 205 0 R 206 0 R 207 0 R 208 0 R 209 0 R 210 0 R ] >> endobj 5 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 211 0 R /Resources << /XObject << /R192 212 0 R /R193 213 0 R /R190 214 0 R /R191 215 0 R /R196 216 0 R /R184 219 0 R /R187 220 0 R /R205 221 0 R /R189 222 0 R /R188 223 0 R /R186 224 0 R /R183 225 0 R /R182 226 0 R /R185 227 0 R >> /ExtGState << /R56 45 0 R /R59 43 0 R /R173 228 0 R >> /Font << /R202 229 0 R /F2 236 0 R /F1 237 0 R /R179 238 0 R /R48 62 0 R /R54 65 0 R /R175 245 0 R /R198 252 0 R /R50 76 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R58 19 0 R /R195 217 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ 259 0 R 260 0 R 261 0 R 262 0 R ] >> endobj 6 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 263 0 R /Resources << /XObject << /R361 264 0 R /R360 265 0 R /R363 266 0 R /R362 267 0 R /R365 268 0 R /R364 269 0 R /R367 270 0 R /R366 271 0 R /R369 272 0 R /R368 273 0 R /R372 274 0 R /R373 275 0 R /R370 276 0 R /R371 277 0 R /R376 278 0 R /R377 279 0 R /R374 280 0 R /R375 281 0 R /R378 282 0 R /R379 283 0 R /R279 284 0 R /R278 285 0 R /R273 286 0 R /R272 287 0 R /R271 288 0 R /R270 289 0 R /R277 290 0 R /R276 291 0 R /R275 292 0 R /R274 293 0 R /R264 294 0 R /R265 295 0 R /R266 296 0 R /R267 297 0 R /R260 298 0 R /R261 299 0 R /R349 300 0 R /R348 301 0 R /R347 302 0 R /R346 303 0 R /R345 304 0 R /R344 305 0 R /R343 306 0 R /R342 307 0 R /R341 308 0 R /R340 309 0 R /R418 310 0 R /R413 311 0 R /R412 312 0 R /R411 313 0 R /R410 314 0 R /R417 315 0 R /R416 316 0 R /R415 317 0 R /R414 318 0 R /R358 319 0 R /R359 320 0 R /R350 321 0 R /R351 322 0 R /R352 323 0 R /R353 324 0 R /R354 325 0 R /R355 326 0 R /R356 327 0 R /R357 328 0 R /R404 329 0 R /R405 330 0 R /R406 331 0 R /R407 332 0 R /R400 333 0 R /R401 334 0 R /R402 335 0 R /R403 336 0 R /R408 337 0 R /R409 338 0 R /R299 339 0 R /R298 340 0 R /R291 341 0 R /R290 342 0 R /R293 343 0 R /R292 344 0 R /R295 345 0 R /R294 346 0 R /R297 347 0 R /R296 348 0 R /R263 349 0 R /R329 350 0 R /R328 351 0 R /R325 352 0 R /R324 353 0 R /R327 354 0 R /R326 355 0 R /R321 356 0 R /R320 357 0 R /R323 358 0 R /R322 359 0 R /R288 360 0 R /R289 361 0 R /R262 362 0 R /R282 363 0 R /R283 364 0 R /R280 365 0 R /R281 366 0 R /R286 367 0 R /R287 368 0 R /R284 369 0 R /R285 370 0 R /R268 371 0 R /R269 372 0 R /R338 373 0 R /R339 374 0 R /R336 375 0 R /R337 376 0 R /R334 377 0 R /R335 378 0 R /R332 379 0 R /R333 380 0 R /R330 381 0 R /R331 382 0 R /R309 383 0 R /R308 384 0 R /R303 385 0 R /R302 386 0 R /R301 387 0 R /R300 388 0 R /R307 389 0 R /R306 390 0 R /R305 391 0 R /R304 392 0 R /R383 393 0 R /R382 394 0 R /R381 395 0 R /R380 396 0 R /R387 397 0 R /R386 398 0 R /R385 399 0 R /R384 400 0 R /R389 401 0 R /R388 402 0 R /R314 403 0 R /R315 404 0 R /R316 405 0 R /R317 406 0 R /R310 407 0 R /R311 408 0 R /R312 409 0 R /R313 410 0 R /R318 411 0 R /R319 412 0 R /R259 413 0 R /R258 414 0 R /R255 415 0 R /R254 416 0 R /R257 417 0 R /R256 418 0 R /R253 419 0 R /R252 420 0 R /R398 421 0 R /R399 422 0 R /R394 423 0 R /R395 424 0 R /R396 425 0 R /R397 426 0 R /R390 427 0 R /R391 428 0 R /R392 429 0 R /R393 430 0 R >> /ExtGState << /R47 44 0 R /R56 45 0 R /R66 21 0 R /R59 43 0 R >> /Font << /R419 431 0 R /R48 62 0 R /R54 65 0 R /R425 435 0 R /R50 76 0 R /R423 439 0 R /R52 80 0 R /R421 443 0 R /F2 447 0 R /F1 448 0 R /R248 449 0 R /R246 453 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] /ColorSpace << /R251 457 0 R /R195 217 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ 459 0 R 460 0 R 461 0 R 462 0 R 463 0 R 464 0 R 465 0 R 466 0 R 467 0 R 468 0 R 469 0 R 470 0 R 471 0 R 472 0 R 473 0 R 474 0 R 475 0 R 476 0 R 477 0 R 478 0 R 479 0 R 480 0 R 481 0 R 482 0 R 483 0 R 484 0 R 485 0 R 486 0 R 487 0 R 488 0 R 489 0 R 490 0 R 491 0 R 492 0 R ] >> endobj 7 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 493 0 R /Resources << /XObject << /R459 494 0 R /R465 495 0 R /R471 496 0 R /R470 497 0 R /R473 498 0 R /R472 499 0 R /R475 500 0 R /R467 501 0 R /R464 502 0 R /R476 503 0 R /R457 504 0 R /R468 505 0 R /R469 506 0 R /R474 507 0 R /R458 508 0 R /R489 509 0 R /R484 510 0 R /R466 511 0 R /R481 512 0 R /R482 513 0 R /R483 514 0 R >> /ExtGState << /R56 45 0 R /R66 21 0 R /R59 43 0 R /R173 228 0 R >> /Font << /R490 515 0 R /R48 62 0 R /R54 65 0 R /F1 519 0 R /R425 435 0 R /R50 76 0 R /F2 520 0 R /R492 521 0 R /R461 524 0 R /R496 531 0 R /R494 536 0 R /R246 453 0 R /R478 540 0 R /R486 547 0 R /R498 554 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R58 19 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ 557 0 R 558 0 R 559 0 R 560 0 R 561 0 R 562 0 R 563 0 R 564 0 R 565 0 R 566 0 R 567 0 R 568 0 R 569 0 R 570 0 R 571 0 R 572 0 R 573 0 R 574 0 R 575 0 R 576 0 R 577 0 R 578 0 R 579 0 R 580 0 R ] >> endobj 8 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 581 0 R /Resources << /ColorSpace << /R58 19 0 R >> /ExtGState << /R47 44 0 R >> /Font << /R425 435 0 R /R48 62 0 R /R54 65 0 R /F1 582 0 R /R515 583 0 R /R50 76 0 R /R513 586 0 R /F2 589 0 R /R492 521 0 R /R490 515 0 R /R496 531 0 R /R494 536 0 R /R246 453 0 R /R498 554 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 590 0 R 591 0 R 592 0 R 593 0 R 594 0 R 595 0 R ] >> endobj 9 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 596 0 R /Resources << /XObject << /R541 597 0 R /R540 598 0 R /R543 599 0 R /R542 600 0 R /R545 601 0 R /R544 602 0 R >> /ExtGState << /R47 44 0 R /R56 45 0 R /R59 43 0 R >> /Font << /R529 603 0 R /R490 515 0 R /R48 62 0 R /R54 65 0 R /F1 610 0 R /R425 435 0 R /R50 76 0 R /F2 611 0 R /R537 612 0 R /R496 531 0 R /R533 619 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R58 19 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ 626 0 R 627 0 R 628 0 R 629 0 R 630 0 R 631 0 R ] >> endobj 10 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 632 0 R /Resources << /XObject << /R578 633 0 R /R579 634 0 R /R608 635 0 R /R573 636 0 R /R574 637 0 R /R575 638 0 R /R576 639 0 R /R577 640 0 R /R606 641 0 R /R607 642 0 R /R604 643 0 R /R605 644 0 R /R602 645 0 R /R603 646 0 R /R600 647 0 R /R601 648 0 R /R642 649 0 R /R643 650 0 R /R640 651 0 R /R641 652 0 R /R646 653 0 R /R647 654 0 R /R644 655 0 R /R645 656 0 R /R609 657 0 R /R633 658 0 R /R632 659 0 R /R631 660 0 R /R630 661 0 R /R637 662 0 R /R636 663 0 R /R635 664 0 R /R634 665 0 R /R639 666 0 R /R638 667 0 R /R585 668 0 R /R584 669 0 R /R587 670 0 R /R586 671 0 R /R581 672 0 R /R580 673 0 R /R583 674 0 R /R582 675 0 R /R589 676 0 R /R588 677 0 R /R559 678 0 R /R628 679 0 R /R629 680 0 R /R624 681 0 R /R625 682 0 R /R626 683 0 R /R627 684 0 R /R620 685 0 R /R621 686 0 R /R622 687 0 R /R623 688 0 R /R596 689 0 R /R597 690 0 R /R594 691 0 R /R595 692 0 R /R592 693 0 R /R593 694 0 R /R590 695 0 R /R591 696 0 R /R598 697 0 R /R599 698 0 R /R648 699 0 R /R564 700 0 R /R563 701 0 R /R562 702 0 R /R561 703 0 R /R560 704 0 R /R615 705 0 R /R614 706 0 R /R617 707 0 R /R616 708 0 R /R611 709 0 R /R610 710 0 R /R613 711 0 R /R612 712 0 R /R619 713 0 R /R618 714 0 R >> /ExtGState << /R56 45 0 R /R66 21 0 R /R59 43 0 R /R173 228 0 R >> /Font << /F2 715 0 R /R54 65 0 R /R649 716 0 R /R496 531 0 R /R48 62 0 R /R570 719 0 R /R566 726 0 R /F1 733 0 R /R50 76 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R58 19 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ 734 0 R 735 0 R 736 0 R 737 0 R 738 0 R ] >> endobj 11 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 739 0 R /Resources << /ExtGState << /R47 44 0 R >> /Font << /R54 65 0 R /F2 740 0 R /F1 741 0 R /R50 76 0 R /R48 62 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 742 0 R 743 0 R 744 0 R 745 0 R 746 0 R 747 0 R 748 0 R 749 0 R 750 0 R 751 0 R 752 0 R 753 0 R 754 0 R 755 0 R 756 0 R 757 0 R 758 0 R 759 0 R 760 0 R 761 0 R 762 0 R 763 0 R 764 0 R 765 0 R 766 0 R 767 0 R 768 0 R 769 0 R 770 0 R 771 0 R 772 0 R 773 0 R 774 0 R 775 0 R 776 0 R 777 0 R 778 0 R 779 0 R 780 0 R 781 0 R 782 0 R 783 0 R 784 0 R 785 0 R 786 0 R 787 0 R 788 0 R 789 0 R 790 0 R 791 0 R 792 0 R 793 0 R 794 0 R 795 0 R 796 0 R 797 0 R ] >> endobj 12 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 798 0 R /Resources << /ExtGState << /R47 44 0 R >> /Font << /R54 65 0 R /F2 799 0 R /F1 800 0 R /R50 76 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 801 0 R 802 0 R 803 0 R 804 0 R 805 0 R 806 0 R 807 0 R 808 0 R 809 0 R 810 0 R 811 0 R 812 0 R 813 0 R 814 0 R 815 0 R 816 0 R 817 0 R 818 0 R 819 0 R 820 0 R 821 0 R 822 0 R 823 0 R 824 0 R 825 0 R 826 0 R 827 0 R 828 0 R 829 0 R 830 0 R 831 0 R 832 0 R 833 0 R 834 0 R 835 0 R 836 0 R ] >> endobj 13 0 obj << /Type /Catalog /Pages 1 0 R >> endobj 14 0 obj << /Length 34552 >> stream q q q 0.10000 0 0 0.10000 0 0 cm /R47 gs 0 g q 10 0 0 10 0 0 cm BT /R48 14.34620 Tf 1 0 0 1 136.79300 675.06700 Tm [ (V) 37.01370 (isual7W) 54.99300 (\072) -249.98900 (Gr) 18.01160 (ounded) -249.99000 (Question) -249.99600 (Answering) -249.98800 (in) -249.99100 (Images) ] TJ /R50 11.95520 Tf -9.94688 -37.85820 Td [ (Y) 110.99500 (uk) 10.00850 (e) -249.99700 (Zhu) ] TJ /R52 7.97010 Tf 46.69690 4.33906 Td [ (y) -0.19911 ] TJ /R50 11.95520 Tf 31.16090 -4.33906 Td [ (Oli) 25.01400 (v) 14.98510 (er) -250.01200 (Groth) ] TJ /R52 7.97010 Tf 60.94730 4.33906 Td [ (z) -0.19911 ] TJ /R50 11.95520 Tf 31.16090 -4.33906 Td [ (Michael) -250.00200 (Bernstein) ] TJ /R52 7.97010 Tf 87.98790 4.33906 Td [ (y) -0.19911 ] TJ /R50 11.95520 Tf 31.16090 -4.33906 Td [ (Li) -249.98500 (Fei\055Fei) ] TJ /R52 7.97010 Tf 48.15510 4.33906 Td [ (y) -0.20217 ] TJ -292.41200 -14.05820 Td [ (y) -0.19911 ] TJ /R50 11.95520 Tf 4.26211 -4.33789 Td [ (Computer) -250.01400 (Science) -250.00600 (Department\054) -249.98900 (Stanford) -250 (Uni) 24.99570 (v) 14.98510 (ersity) ] TJ /R52 7.97010 Tf -39.69690 -9.71992 Td [ (z) -0.19911 ] TJ /R50 11.95520 Tf 4.26172 -4.33789 Td [ (Computer) -250.01400 (Science) -250.00600 (Department\054) -249.98900 (Dresden) -249.99300 (Uni) 24.99570 (v) 14.98510 (ersity) -249.98900 (of) -250.01400 (T) 70.01640 (echnology) ] TJ /R48 11.95520 Tf 5.46406 -41.04610 Td (Abstract) Tj /R54 9.96260 Tf -83.92770 -24.22190 Td [ (W) 91.98650 (e) -382.01900 (have) -380.99400 (seen) -382.01200 (gr) 36.98520 (eat) -380.98800 (pr) 44.98390 (o) 10.00320 (gr) 36.98520 (ess) -382.00700 (in) -382.00200 (basic) -381.01500 (per) 36.98160 (ceptual) -381.99500 (tasks) ] TJ -11.95510 -11.95510 Td [ (suc) 14.98460 (h) -221.01700 (as) -219.99100 (object) -221.00600 (r) 37.01960 (eco) 9.99466 (gnition) -220 (and) -221.00700 (detection\056) -300.00100 (Howe) 14.99500 (ver) 110.99900 (\054) -226 (AI) -220.98300 (mod\055) ] TJ 11.95510 TL T* [ (els) -285.00200 (still) -285.99600 (fail) -284.98500 (to) -286.01800 (matc) 14.98890 (h) -285.00600 (humans) -286.00100 (in) -285 (high\055le) 15 (vel) -285.99400 (vision) -284.98400 (tasks) -286.01100 (due) ] TJ T* [ (to) -265.98300 (the) -265.98700 (lac) 20.00460 (k) -265.01900 (of) -265.98300 (capacities) -266.00500 (for) -266.01600 (deeper) -265.99100 (r) 37.01960 (easoning) 15.01220 (\056) -356.99700 (Recently) -265.99000 (the) ] TJ T* [ (ne) 15.01770 (w) -314.99300 (task) -314.01200 (of) -314.99400 (visual) -314 (question) -315.01000 (answering) -314.01700 (\050QA\051) -315.00600 (has) -314.00900 (been) -315.00400 (pr) 44.98510 (o\055) ] TJ 11.95590 TL T* [ (posed) -231.98400 (to) -233.00800 (e) 15.01280 (valuate) -232.01900 (a) -231.99600 (model\047) 40.00560 (s) -233.01400 (capaci) 0.99371 (ty) -233.01700 (for) -231.98300 (deep) -231.99800 (ima) 10.01300 (g) 10.00320 (e) -232.98600 (under) 20.01380 (\055) ] TJ 11.95510 TL T* [ (standing) 14.98400 (\056) -483.99500 (Pr) 36.98400 (e) 15.01220 (vious) -308.00900 (works) -308.00100 (have) -307.98700 (established) -307.98300 (a) -307.98300 (loose) 10.01790 (\054) -323.00200 (global) ] TJ T* [ (association) -356.98400 (between) -358.00200 (QA) -356.99200 (sentences) -356.98700 (and) -358.00400 (ima) 10.01300 (g) 10.00320 (es\056) -632.00700 (Howe) 14.99500 (ver) 110.99900 (\054) ] TJ T* [ (many) -414 (questions) -414.98700 (and) -413.99500 (answer) 9.98975 (s\054) -455.99400 (in) -413.99700 (pr) 14.98890 (actice) 10.00570 (\054) -456 (r) 37.01830 (elate) -414 (to) -415.01700 (local) ] TJ T* [ (r) 37.01960 (e) 39.98900 (gions) -418.02000 (in) -417.01600 (the) -418.00100 (ima) 10.01300 (g) 10.00320 (es\056) -812.99800 (W) 91.98590 (e) -418.01400 (establish) -418.00400 (a) -416.98400 (semantic) -417.98400 (link) -417.99400 (be\055) ] TJ T* [ (tween) -193.01000 (te) 20 (xtual) -191.98700 (descriptions) -193.01400 (and) -193.01200 (ima) 10.01300 (g) 10.00320 (e) -192.01200 (r) 37.01960 (e) 39.98840 (gions) -192.99700 (by) -192.98700 (object\055le) 14.99260 (vel) ] TJ 11.95590 TL T* [ (gr) 44.98390 (ounding) 15.01530 (\056) -297.00700 (It) -210 (enables) -209.98500 (a) -210 (ne) 15.01830 (w) -209.99100 (type) -210 (of) -209.99200 (QA) -211.01600 (wi) 1 (th) -211.01100 (visual) -210.01600 (answer) 9.99098 (s\054) ] TJ 11.95510 TL T* [ (in) -350.98800 (additi) 1.01699 (on) -350.99000 (to) -350.98800 (te) 20 (xtual) -350 (answer) 9.98975 (s) -350.99400 (used) -350.00800 (in) -350.98800 (pr) 36.98650 (e) 15.01220 (vious) -350 (work\056) -611.99300 (W) 91.98710 (e) ] TJ T* [ (study) -240.01900 (the) -239.99100 (visual) -240.01300 (QA) -239.99200 (tasks) -239 (in) -239.98700 (a) -239.99400 (gr) 44.98390 (ounded) -240.01900 (setting) -240.00400 (with) -240.00700 (a) -239.99400 (lar) 36.99140 (g) 10.00320 (e) ] TJ T* [ (collection) -361.98700 (of) -362.00600 (7W) -360.99100 (multiple\055c) 15.00730 (hoice) -362.00900 (QA) -362.01100 (pair) 9.98118 (s\056) -646 (Furthermor) 37.01830 (e) 9.99343 (\054) ] TJ T* [ (we) -235.99100 (e) 15.01280 (valuate) -236.01800 (human) -236.01600 (performance) -236.01000 (and) -235.98500 (se) 15.01960 (ver) 15.01470 (al) -235.98800 (baseline) -236.01200 (mod\055) ] TJ T* [ (els) -398.00300 (on) -396.98300 (the) -398.00400 (QA) -398.00500 (tasks\056) -752.99000 (F) 45.01580 (inally) 54.99820 (\054) -434.98400 (we) -396.98400 (pr) 44.98390 (opose) -397.99500 (a) -398.00700 (no) 10.00810 (vel) -398.01400 (LSTM) ] TJ 11.95590 TL T* [ (model) -249.98200 (with) -250.00400 (spatial) -249.99400 (attention) -249.99300 (to) -249.98500 (tac) 20.00400 (kle) -249.99900 (the) -249.99000 (7W) -249.99000 (QA) -249.99000 (tasks\056) ] TJ /R48 11.95520 Tf 36.79920 TL T* [ (1\056) -249.99000 (Intr) 18.01460 (oduction) ] TJ /R50 9.96260 Tf 11.95510 -19.23980 Td [ (The) -388.99000 (recent) -390.01700 (de) 25.01670 (v) 14.98280 (elopment) -389 (of) -390.01200 (deep) -388.99300 (learning) -389.99600 (technologies) ] TJ -11.95510 -11.95510 Td [ (has) -207.99800 (achie) 25.01480 (v) 14.98280 (ed) -208.00400 (successes) -207.99200 (i) 0.98758 (n) -208 (man) 14.99010 (y) -208 (perceptual) -207.99500 (visual) -208.01700 (tasks) -207.98500 (such) ] TJ 11.95510 TL T* [ (as) -249.01500 (object) -249.00100 (recognition\054) -250 (image) -249.00100 <636c6173736902636174696f6e> -249.00700 (and) -249.01200 (pose) -249.98500 (estima\055) ] TJ 11.95590 TL T* [ (tion) -801.01800 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 76.91170 248.68300 Tm (15) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 86.87420 248.68300 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 93.35000 248.68300 Tm (21) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 103.31300 248.68300 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 109.78800 248.68300 Tm (25) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 119.75100 248.68300 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 126.23600 248.68300 Tm (38) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 136.19900 248.68300 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 142.67500 248.68300 Tm (39) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 152.63700 248.68300 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 159.11300 248.68300 Tm (42) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 169.07500 248.68300 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 175.56100 248.68300 Tm (43) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 185.52300 248.68300 Tm [ (\135\056) -759.99200 (Y) 99.98580 (et) -400.98900 (the) -400.00400 (status) -399.98500 (quo) -401.01600 (of) ] TJ -135.41100 -11.95510 Td [ (computer) -326.98200 (vision) -327.00800 (is) -326.99100 (st) 0.99371 (ill) -326.98200 (f) 9.99466 (ar) -327.01200 (from) -326.99300 (matching) -327.01200 (human) -326.98200 (capabil\055) ] TJ 11.95510 TL T* [ (ities\054) -353.99400 (especially) -334.00200 (when) -332.99500 (it) -334.00800 (comes) -332.98900 (to) -334.01000 (understanding) -332.99800 (an) -333.98300 (image) ] TJ T* [ (in) -282.01900 (all) -283.00200 (its) -282.01100 (details\056) -407.00300 (Recently) 64.99410 (\054) -291.00800 (visual) -282.00600 (question) -281.99500 (answering) -282.99200 (\050QA\051) ] TJ T* [ (has) -342.99500 (been) -343.98100 (proposed) -343.01800 (as) -344.02000 (a) -343.00700 (proxy) -344.00400 (task) -342.98800 (for) -343.98400 (e) 25.01050 (v) 24.98110 (aluating) -343.01400 (a) -343.98700 (vision) ] TJ T* [ (system\047) 55 (s) -277.98700 (capacity) -279.00500 (for) -277.99400 (deeper) -277.99900 (image) -278.99600 (understanding\056) -394.98600 (Se) 24.98600 (v) 14.98280 (eral) ] TJ 11.95590 TL T* [ (QA) -271.00800 (datasets) -270.01500 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 104.74700 176.95200 Tm (1) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 109.72900 176.95200 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 114.91900 176.95200 Tm (7) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 119.90000 176.95200 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 125.08100 176.95200 Tm (27) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 135.04300 176.95200 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 140.23400 176.95200 Tm (36) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 150.19600 176.95200 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 155.38700 176.95200 Tm (49) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 165.35000 176.95200 Tm [ (\135) -269.99800 (ha) 19.99790 (v) 14.98280 (e) -271.01800 (been) -271.01100 (releas) 1.01699 (ed) -271.01400 (since) -271.00400 (last) ] TJ -115.23700 -11.95510 Td [ (year) 55.00370 (\056) -317.00300 (The) 14.98220 (y) -253.01200 (contrib) 20.01870 (uted) -251.98500 (v) 24.98110 (aluable) -251.98600 (data) -253.01300 (for) -252 (training) -252.00700 (visual) -252.99000 (QA) ] TJ 11.95510 TL T* [ (systems) -275.99400 (and) -275.00700 (introduced) -276.00500 (v) 24.98110 (arious) -276.01600 (tasks\054) -282.00900 (from) -275.98300 (picking) -276.00300 (correct) ] TJ T* [ (multiple\055choice) -249.98900 (answers) -249.98900 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 153.03600 141.08700 Tm (1) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 158.01700 141.08700 Tm [ (\135) -250.00100 (to) -249.98500 <026c6c696e67> -249.99100 (in) -249.98500 (blanks) -250.01200 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 242.43000 141.08700 Tm (49) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 252.39200 141.08700 Tm (\135\056) Tj -190.32500 -12.26680 Td [ (Pioneer) -449.98400 (w) 10 (ork) -451.01500 (in) -449.99100 (image) -451.00700 (captioning) -450.00600 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 212.79100 128.82000 Tm (4) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 217.77300 128.82000 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 224.74600 128.82000 Tm (5) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 229.72800 128.82000 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 236.71100 128.82000 Tm (13) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 246.67400 128.82000 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 253.64800 128.82000 Tm (45) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 263.61000 128.82000 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 270.59400 128.82000 Tm (47) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 280.55600 128.82000 Tm (\135\054) Tj -230.44400 -11.95510 Td [ (sentence\055based) -211.99400 (image) -211.98800 (retrie) 25.01420 (v) 24.98110 (al) -212.00100 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 177.11500 116.86500 Tm (14) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 187.07800 116.86500 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 191.68000 116.86500 Tm (40) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 201.64300 116.86500 Tm [ (\135) -212.00600 (and) -211.99900 (visual) -212.99600 (QA) -211.99900 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 269.30900 116.86500 Tm (1) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 274.29000 116.86500 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 278.89300 116.86500 Tm (7) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 283.87400 116.86500 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 50.11210 104.91000 Tm (36) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 60.07460 104.91000 Tm [ (\135) -305.01100 (sho) 24.99220 (ws) -306.01800 (promising) -304.99800 (results\056) -476.01100 (These) -305.00200 (w) 10 (orks) -304.99600 (aimed) -304.99300 (at) -305.98600 (estab\055) ] TJ -9.96250 -11.95510 Td [ (lishing) -273.01000 (a) -273.01800 (global) -272.98900 (association) -273.01700 (between) -273.98600 (s) 0.98635 (entences) -273.98900 (and) -273.00800 (images\056) ] TJ 11.95510 TL T* [ (Ho) 24.98420 (we) 25.01420 (v) 14.98280 (er) 39.98600 (\054) -416.98700 (as) -384.01300 (Flickr30K) -384.01400 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 152.44800 81 Tm (34) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 162.41100 81 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 168.72700 81 Tm (48) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 178.69000 81 Tm [ (\135) -384.01800 (and) -384.00800 (V) 59.99240 (isual) -383.99200 (Madlibs) -384.01700 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 273.08600 81 Tm (49) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 283.04800 81 Tm (\135) Tj ET Q /R56 gs q 3088 4590 2364 1191 re W n /R58 cs 0 0 0 scn 3079.20000 4583.93000 2373.61000 1201.36000 re f 3479.82000 4687.79000 1315.80000 1086.85000 re f Q /R59 gs q 3479.82000 4687.79000 1315.80000 1086.85000 re W n q 1315.80000 0 0 1086.85000 3479.82000 4687.79000 cm /R60 Do Q Q /R56 gs q 3088 4590 2364 1191 re W n /R58 cs 0 0 0 scn 3479.82000 4914.54000 1223.96000 129.55100 re f /R59 gs 10.52640 w 14.33560 M 1 j /R58 CS 0 1 1 SCN 3479.82000 4914.54000 1223.96000 129.55100 re S /R56 gs 3670.93000 5017.98000 560.46900 484.71100 re f /R59 gs 1 0.59961 0 SCN 3670.93000 5017.98000 560.46900 484.71100 re S /R56 gs 3680.90000 5030.44000 102.85900 79.65230 re f /R59 gs 1 0 0 SCN 3680.90000 5030.44000 102.85900 79.65230 re S /R56 gs 3800.36000 5030.44000 155.98800 79.65230 re f /R59 gs 3800.36000 5030.44000 155.98800 79.65230 re S /R56 gs 4064.79000 5030.44000 155.98800 79.65230 re f /R59 gs 4064.79000 5030.44000 155.98800 79.65230 re S /R56 gs 3956.34000 5064.66000 102.85900 79.65230 re f /R59 gs 0 1 0 SCN 3956.34000 5064.66000 102.85900 79.65230 re S /R56 gs 3630.44000 4584.52000 696.39800 107.08600 re f /R59 gs q 10 0 0 10 0 0 cm BT /R62 6.31584 Tf 1 0 0 1 367.51200 460.72900 Tm [ <003a> 0.99731 <004b> 0.98957 <004c> 0.98184 <0046004b> 0.98957 <0003> ] TJ ET Q 0 1 0 scn q 10 0 0 10 0 0 cm BT /R62 6.31584 Tf 1 0 0 1 388.21100 460.72900 Tm [ <0053> 0.98957 <0044005a> ] TJ ET Q 0 0 0 scn q 10 0 0 10 0 0 cm BT /R62 6.31584 Tf 1 0 0 1 400.49500 460.72900 Tm [ <0003> 0.98184 <004c> 0.98184 <00560003> 0.98184 <004f> 0.98184 <004c> 0.98184 <0049005700480047> 0.98957 <0022> ] TJ ET Q /R56 gs 3073.70000 4973.68000 297.64100 246.08600 re f /R59 gs q 10 0 0 10 0 0 cm BT /R62 6.31584 Tf 1 0 0 1 309.73800 513.54500 Tm [ <003a> 0.99731 <004b> 0.98957 <004400570003> 0.98184 <004c> 0.98184 <00560003> ] TJ 7.63164 TL T* [ <0057004b> 0.98957 <00480003> ] TJ ET Q 1 0.59961 0 scn q 10 0 0 10 0 0 cm BT -0.00632 Tc /R62 6.31584 Tf 1 0 0 1 320.96600 505.91300 Tm <00470052004a> Tj ET Q 0 0 0 scn q 10 0 0 10 0 0 cm BT /R62 6.31584 Tf 1 0 0 1 332.54000 505.91300 Tm <0003> Tj -0.00632 Tc -22.80230 -7.63164 Td <00470052004c0051004a0022> Tj ET Q /R56 gs 3978.64000 4691.61000 m 3978.64000 4776.31000 3917.05000 4818.67000 3855.47000 4861.02000 c 3793.89000 4903.38000 3732.30000 4945.73000 3732.30000 5030.44000 c f /R59 gs 7.89480 w 1 0 0 SCN 3978.64000 4691.61000 m 3978.64000 4776.31000 3917.05000 4818.67000 3855.47000 4861.02000 c 3824.68000 4882.20000 3793.89000 4903.38000 3770.80000 4929.85000 c 3759.25000 4943.09000 3749.63000 4957.64000 3742.89000 4974.19000 c 3742.05000 4976.26000 3741.25000 4978.36000 3740.50000 4980.49000 c 3740.31000 4981.02000 3740.13000 4981.55000 3739.95000 4982.09000 c 3739.45000 4983.61000 l S /R67 Do /R56 gs 3978.64000 4691.61000 m 3978.64000 4776.31000 3953.57000 4818.67000 3928.50000 4861.02000 c 3903.42000 4903.38000 3878.35000 4945.73000 3878.35000 5030.44000 c f /R59 gs 3978.64000 4691.61000 m 3978.64000 4776.31000 3953.57000 4818.67000 3928.50000 4861.02000 c 3915.96000 4882.20000 3903.42000 4903.38000 3894.02000 4929.85000 c 3889.32000 4943.09000 3885.40000 4957.64000 3882.66000 4974.19000 c 3882.32000 4976.26000 3881.99000 4978.36000 3881.69000 4980.49000 c 3881.32000 4983.16000 l S /R68 Do /R56 gs 3978.64000 4691.61000 m 3978.64000 4784.87000 3985.93000 4831.50000 3993.23000 4878.13000 c 4000.52000 4924.77000 4007.82000 4971.40000 4007.82000 5064.67000 c f /R59 gs 0 1 0 SCN 3978.64000 4691.61000 m 3978.64000 4784.87000 3985.93000 4831.50000 3993.23000 4878.13000 c 3996.88000 4901.45000 4000.52000 4924.77000 4003.26000 4953.91000 c 4004.63000 4968.48000 4005.77000 4984.52000 4006.56000 5002.73000 c 4006.76000 5007.29000 4006.94000 5011.98000 4007.09000 5016.82000 c 4007.11000 5017.31000 l S /R69 Do /R56 gs 3978.64000 4691.61000 m 3978.64000 4776.31000 4019.67000 4818.67000 4060.70000 4861.02000 c 4101.72000 4903.38000 4142.75000 4945.73000 4142.75000 5030.44000 c f /R59 gs 1 0 0 SCN 3978.64000 4691.61000 m 3978.64000 4776.31000 4019.67000 4818.67000 4060.70000 4861.02000 c 4081.21000 4882.20000 4101.72000 4903.38000 4117.11000 4929.85000 c 4124.80000 4943.09000 4131.21000 4957.64000 4135.70000 4974.19000 c 4136.26000 4976.26000 4136.79000 4978.36000 4137.29000 4980.49000 c 4137.42000 4981.02000 4137.54000 4981.55000 4137.66000 4982.09000 c 4137.93000 4983.32000 l S /R70 Do /R56 gs 3371.34000 5096.72000 m 3446.23000 5096.72000 3483.68000 5137.63000 3521.13000 5178.53000 c 3558.59000 5219.43000 3596.05000 5260.34000 3670.96000 5260.34000 c f /R59 gs 1 0.59961 0 SCN 3371.34000 5096.72000 m 3446.23000 5096.72000 3483.68000 5137.63000 3521.13000 5178.53000 c 3539.86000 5198.98000 3558.59000 5219.43000 3582 5234.77000 c 3593.70000 5242.44000 3606.58000 5248.83000 3621.21000 5253.30000 c 3622.13000 5253.59000 3623.05000 5253.86000 3623.98000 5254.12000 c 3624 5254.13000 l S /R71 Do /R56 gs 4874.81000 5541.32000 502.86300 162.94900 re f /R59 gs q 10 0 0 10 0 0 cm BT /R62 6.31584 Tf 1 0 0 1 489.85000 561.99500 Tm [ <003a> 0.99731 <004b> 0.98957 <005c0003> 0.98184 <004c> 0.98184 <00560003> 0.98184 <0057004b> 0.98957 <004800550048> ] TJ /R73 6.31584 Tf 37.55230 0 Td <0003> Tj ET Q 1 0 1 scn q 10 0 0 10 0 0 cm BT /R62 6.31584 Tf 1 0 0 1 489.85000 554.36300 Tm [ <00490052> 0.98957 <00440050> ] TJ ET Q 0 0 0 scn q 10 0 0 10 0 0 cm BT /R62 6.31584 Tf 1 0 0 1 504.93900 554.36300 Tm <0022> Tj ET Q /R56 gs 3479.82000 5526.48000 1315.80000 211.35500 re f /R59 gs 10.52640 w 1 0 1 SCN 3479.82000 5526.48000 1315.80000 211.35500 re S /R56 gs 4874.81000 5622.79000 m 4855.02000 5622.79000 4845.11000 5625.13000 4835.22000 5627.47000 c 4825.32000 5629.82000 4815.44000 5632.16000 4795.66000 5632.16000 c f /R59 gs 7.89480 w 4874.81000 5622.79000 m 4864.91000 5622.79000 4857.49000 5623.38000 4851.30000 5624.25000 c 4848.21000 5624.70000 4845.43000 5625.21000 4842.80000 5625.75000 c 4842.60000 5625.80000 l S /R76 Do /R56 gs 4857.95000 4923.16000 594.45300 162.94900 re f /R59 gs q 10 0 0 10 0 0 cm BT /R62 6.31584 Tf 1 0 0 1 488.16300 500.17900 Tm [ <003a> 0.99731 <004b> 0.98957 <004400570003> 0.98184 <004c> 0.98184 <00560003> 0.98184 <0057004b> 0.98957 <00480003> ] TJ ET Q 1 0.59961 0 scn q 10 0 0 10 0 0 cm BT -0.00632 Tc /R62 6.31584 Tf 1 0 0 1 523.60400 500.17900 Tm <00470052004a> Tj ET Q 0 0 0 scn q 10 0 0 10 0 0 cm BT /R62 6.31584 Tf 1 0 0 1 535.17700 500.17900 Tm <0003> Tj -47.01410 -7.63164 Td [ <0056005700440051> 0.98957 <0047> 0.98957 <004c> 0.98184 <0051> 0.98957 <004a> 0.98957 <0003> 0.98184 <0052> 0.98957 <0051> 0.98957 <0022> ] TJ ET Q /R56 gs 5155.17000 5086.11000 m 5155.17000 5173.22000 4693.30000 5260.34000 4231.42000 5260.34000 c f /R59 gs 1 0.59961 0 SCN 5155.17000 5086.11000 m 5155.17000 5129.67000 5039.70000 5173.22000 4866.50000 5205.89000 c 4779.90000 5222.22000 4678.86000 5235.84000 4570.61000 5245.36000 c 4516.48000 5250.13000 4460.55000 5253.87000 4403.72000 5256.42000 c 4375.30000 5257.70000 4346.66000 5258.68000 4317.91000 5259.34000 c 4310.72000 5259.50000 4303.52000 5259.64000 4296.32000 5259.77000 c 4292.72000 5259.83000 4289.12000 5259.89000 4285.52000 5259.94000 c 4278.79000 5260.03000 l S /R77 Do /R56 gs 3073.70000 5314.63000 386.74200 465.31600 re f /R59 gs q 10 0 0 10 0 0 cm BT /R62 6.31584 Tf 1 0 0 1 309.73800 569.56300 Tm [ <003a> 0.99731 <004b> 0.98957 <0048005500480003> ] TJ 7.63164 TL T* [ <0047> 0.98957 <0052> 0.98957 <004800560003> 0.98184 <0057004b> 0.98957 <004c> 0.98184 <00560003> ] TJ T* [ <0056004600480051> 0.98957 <00480003> 0.98184 <00570044004e00480003> ] TJ T* [ <0053> 0.98957 <004f> 0.98184 <0044004600480022> ] TJ /R79 4.56144 Tf 5.94727 TL T* [ <0024> 1.00623 <000c0003> 0.96876 <002c> 0.96876 <00510003> 0.96876 <0057> 0.96876 <004b00480003> 0.96876 <0056004800440011> 0.96876 <0003> ] TJ ET Q 0 1 0 scn q 10 0 0 10 0 0 cm BT /R83 4.56144 Tf 1 0 0 1 338.12700 540.72000 Tm (F ) Tj ET Q 0 0 0 scn q 10 0 0 10 0 0 cm BT /R79 4.56144 Tf 1 0 0 1 309.73800 535.19400 Tm [ <0025> 1.00623 <000c0003> 0.96876 <002c> 0.96876 <00510003> 0.96876 <0057> 0.96876 <004b00480003> 0.96876 <004700480056004800550057> 0.96876 <0011> ] TJ 5.52617 TL T* [ <0026000c0003> 0.96876 <002c> 0.96876 <00510003> 0.96876 <0057> 0.96876 <004b00480003> 0.96876 <0049> 0.96876 <00520055004800560057> 0.96876 <0011> ] TJ T* [ <0027000c0003> 0.96876 <0032> 1.02228 <00510003> 0.96876 <00440003> 0.96876 <004f0044005a00510011> ] TJ ET Q /R56 gs 3073.70000 4715.64000 386.74200 264.82000 re f /R59 gs q 10 0 0 10 0 0 cm BT /R79 4.91232 Tf 1 0 0 1 309.73800 490.96200 Tm [ <0024> 1.02381 <000c0003> 1.03375 <0036> 1.02381 <005800550049> 1.03375 <004c0051004a0011> 1.03375 <0003> ] TJ ET Q 0 1 0 scn q 10 0 0 10 0 0 cm BT /R83 4.91232 Tf 1 0 0 1 334.29500 490.96200 Tm (F ) Tj ET Q 0 0 0 scn q 10 0 0 10 0 0 cm BT /R79 4.91232 Tf 1 0 0 1 309.73800 485.17200 Tm [ <0025> 1.02381 <000c0003> 1.03375 <0036> 1.02381 <004f004800480053004c0051004a0011> ] TJ 5.78945 TL T* [ <0026000c0003> 1.03375 <0035005800510051004c0051004a0011> ] TJ T* [ <0027000c0003> 1.03375 <0028> 1.02381 <00440057> 1.03375 <004c0051004a0011> ] TJ ET Q /R56 gs 4874.81000 5313.57000 607.30100 246.08600 re f /R59 gs q 10 0 0 10 0 0 cm BT /R79 4.91232 Tf 1 0 0 1 489.85000 548.88100 Tm [ <0024> 1.02381 <000c0003> 1.03375 <0025> 1.02381 <0048004600440058005600480003> 1.03375 <00520049> 1.03375 <0003> 1.03375 <00440003> 1.03375 <005a0044005900480011> 1.03375 <0003> ] TJ ET Q 0 1 0 scn q 10 0 0 10 0 0 cm BT /R83 4.91232 Tf 1 0 0 1 540.33600 548.88100 Tm (F ) Tj ET Q 0 0 0 scn q 10 0 0 10 0 0 cm BT /R79 4.91232 Tf 1 0 0 1 489.85000 543.09200 Tm [ <0025> 1.02381 <000c0003> 1.03375 <0025> 1.02381 <0048004600440058005600480003> 1.03375 <00520049> 1.03375 <0003> 1.03375 <00440003> 1.03375 <0045005200440057> 1.03375 <0011> ] TJ 5.78984 TL T* [ <0026000c0003> 1.03375 <0025> 1.02381 <0048004600440058005600480003> 1.03375 <00520049> 1.03375 <0003> 1.03375 <00440003> 1.03375 <0049> 1.03375 <004c005500480011> ] TJ T* [ <0027000c0003> 1.03375 <0025> 1.02381 <0048004600440058005600480003> 1.03375 <00520049> 1.03375 <0003> 1.03375 <00440003> 1.03375 <004f00480044004e0011> ] TJ ET Q /R56 gs 4853.76000 4660.05000 577.54300 264.81600 re f /R59 gs q 10 0 0 10 0 0 cm BT /R73 4.91232 Tf 1 0 0 1 487.74500 485.40300 Tm [ <0024> 1.02381 <000c0003> ] TJ /R79 4.91232 Tf 6.27344 0 Td [ <0032> 1.01387 <00510003> 1.03375 <00440003> ] TJ ET Q 0 1 1 scn q 10 0 0 10 0 0 cm BT /R87 4.91232 Tf 1 0 0 1 506.02600 485.40300 Tm [ <00560058> 1.00393 <005500490045> 1.00393 <0052> 1.00393 <004400550047> ] TJ ET Q 0 0 0 scn q 10 0 0 10 0 0 cm BT -0.00491 Tc /R79 4.91232 Tf 1 0 0 1 528.93900 485.40300 Tm <00110003> Tj ET Q 0 1 0 scn q 10 0 0 10 0 0 cm BT /R83 4.91232 Tf 1 0 0 1 531.66700 485.40300 Tm (F ) Tj ET Q 0 0 0 scn q 10 0 0 10 0 0 cm BT /R79 4.91232 Tf 1 0 0 1 487.74500 479.61300 Tm [ <0025> 1.02381 <000c0003> 1.03375 <0032> 1.01387 <00510003> 1.03375 <00440003> 1.03375 <0057> 1.03375 <00440045004f00480011> ] TJ 5.78945 TL T* [ <0026000c0003> 1.03375 <0032> 1.01387 <00510003> 1.03375 <00440003> 1.03375 <004a004400550044004a00480011> ] TJ T* [ <0027000c0003> 1.03375 <0032> 1.01387 <00510003> 1.03375 <00440003> 1.03375 <00450044004f004f0011> ] TJ ET Q /R56 gs 4853.76000 4792.46000 m 4816.27000 4792.46000 4797.52000 4839.17000 4778.77000 4885.88000 c 4760.03000 4932.58000 4741.30000 4979.29000 4703.82000 4979.29000 c f /R59 gs 0 1 1 SCN 4853.76000 4792.46000 m 4816.27000 4792.46000 4797.52000 4839.17000 4778.77000 4885.88000 c 4769.40000 4909.23000 4760.04000 4932.58000 4748.32000 4950.09000 c 4747.59000 4951.19000 4746.85000 4952.26000 4746.10000 4953.31000 c 4745.72000 4953.83000 4745.35000 4954.35000 4744.96000 4954.86000 c 4744.65000 4955.27000 l S /R90 Do Q /R59 gs q 10 0 0 10 0 0 cm BT /R48 8.96640 Tf 1 0 0 1 308.86200 449.25900 Tm [ (Figur) 17.99520 (e) -392.02100 (1\072) ] TJ /R50 8.96640 Tf 41.55780 0 Td [ (Deep) -392.00400 (image) -392.97900 (underst) 0.98022 (anding) -393.01500 (relies) -391.98500 (on) -392.00200 (detailed) -392.98700 (kno) 25.00660 (wl\055) ] TJ -41.55780 -10.95900 Td [ (edge) -265.00300 (about) -266.02100 (dif) 25.00110 (ferent) -264.97900 (image) -265.98600 (parts\056) -356.01700 (W) 80.00790 (e) -265.98300 (emplo) 10.02000 (y) -265.00800 (di) 25.01740 (v) 14.98650 (erse) -266.01600 (questions) -265.00300 (to) ] TJ 10.95900 TL T* [ (acquire) -219.98100 (detailed) -219.99200 (information) -220.01600 (on) -220.00500 (images\054) -226.01200 (ground) -220.02200 (objects) -220.02200 (mentioned) ] TJ T* [ (in) -192.98700 (te) 15.01920 (xt) -191.98500 (with) -192.98700 (their) -192.00900 (visual) -192.99200 (appearances\054) -203.99000 (and) -192.01500 (pro) 15.01370 (vide) -193.00600 (a) -193.01100 (mult) 0.98567 (iple\055choice) ] TJ T* [ (setting) -321.98100 (for) -322.97800 (e) 24.98750 (v) 25.00660 (aluati) 0.99111 (ng) -322.99400 (a) -322.01100 (visual) -321.98700 (question) -322.99400 (answering) -321.99700 (task) -321.99700 (with) -321.98700 (both) ] TJ T* [ (te) 15.01920 (xtual) -249.97800 (and) -249.99700 (visual) -250.01900 (answers\056) ] TJ /R50 9.96260 Tf 25.45820 TL T* [ (suggest\054) -472.00700 (a) -428.01200 (tighter) -427.00200 (semantic) -427.98300 (link) -427.00200 (between) -428.00200 (te) 14.98280 (xtual) -427.98700 (descrip\055) ] TJ 11.95590 TL T* [ (tions) -337.99300 (and) -336.99800 (corresponding) -337.98800 (visual) -337.01700 (re) 15.00730 (gions) -337.99300 (is) -338.00700 (a) -337.00700 (k) 10.00320 (e) 15.01220 (y) -338.01700 (ingredient) ] TJ 11.95510 TL T* [ (for) -299.01100 (better) -299.00200 (models\056) -456.00700 (As) -299 (Fig\056) ] TJ ET Q 1 0 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 417.74400 345.09500 Tm (1) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 425.69400 345.09500 Tm [ (sho) 24.99090 (ws\054) -310.99700 (the) -299 (localization) -299.01400 (of) -299.00900 (ob\055) ] TJ -116.83200 -11.95470 Td [ (jects) -300.01600 (can) -299.01100 (be) -299.98900 (a) -299.99400 (critical) -298.99700 (step) -300.01400 (to) -300.01900 (understand) -298.98400 (images) -300.00900 (better) -299.97900 (and) ] TJ 11.95510 TL T* [ (solv) 15 (e) -325.99000 (image\055related) -326.99000 (questions\056) -539.00300 (Pro) 14.99260 (viding) -325.99000 (these) -326.99400 (image\055te) 14.99260 (xt) ] TJ T* [ (correspondences) -362.00300 (is) -362.00600 (called) ] TJ /R54 9.96260 Tf 107.65200 0 Td [ (gr) 44.98270 (ounding) ] TJ /R50 9.96260 Tf 41.06600 0 Td [ (\056) -646.98500 (Inspired) -361.99800 (by) -362.00800 (Geman) ] TJ -148.71800 -11.95590 Td [ (et) -414.00700 (al\056) 69.99320 (\047) 55.00800 (s) -412.98500 (prototype) -414.01700 (of) -414.00700 (a) -412.99500 (visual) -413.98500 (T) 45.01700 (uring) -413.99000 (test) -413.01400 (based) -414 (on) -414 (image) ] TJ T* [ (re) 15.00980 (gions) -381.00500 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 345.15600 285.31900 Tm (8) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 350.13700 285.31900 Tm [ (\135) -382.01700 (and) -380.99000 (the) -382.00700 (comprehensi) 25.00560 (v) 14.98280 (e) -381 (data) -382.01000 (collection) -381.00500 (of) -382.01500 (QA) ] TJ -41.27500 -11.95510 Td [ (pairs) -203.99500 (on) -203.99500 (COCO) -204.01300 (images) -205.00300 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 405.54900 273.36400 Tm (25) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 415.51200 273.36400 Tm [ (\135) -204.01000 (such) -203.99500 (as) -204.00500 (VQA) -203.99500 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 478.41600 273.36400 Tm (1) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 483.39700 273.36400 Tm [ (\135) -204.01000 (and) -205.02000 (Baidu) -204.02000 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 534.32600 273.36400 Tm (7) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 539.30700 273.36400 Tm (\135\054) Tj -230.44500 -11.95510 Td [ (we) -305.01000 (fuse) -305.00500 (visual) -304.98100 (QA) -305.00300 (and) -305.00300 (grounding) -305.00800 (in) -304.99800 (order) -305.00800 (to) -306.01800 (create) -304.98300 (a) -305.01300 (ne) 25.01540 (w) ] TJ 11.95470 TL T* [ (QA) -487.01200 (dataset) -487.00500 (with) -488.01400 (dense) -487.00900 (annotations) -487.99200 (and) -487.01200 (a) -486.98200 (more) -487.98700 <0365> 14.98770 (xible) ] TJ T* [ (e) 25.01050 (v) 24.98110 (aluation) -387.98700 (en) 39.99330 (vironment\056) -725.01400 (Object\055le) 24.99090 (v) 14.98280 (el) -388.00900 (grounding) -388.99400 (pro) 14.98280 (vides) ] TJ 11.95630 TL T* [ (a) -357.98400 (stronger) -356.98200 (link) -357.99400 (between) -356.99200 (QA) -358.01400 (pairs) -356.98900 (and) -358.01400 (images) -357.01900 (than) -358.00400 (global) ] TJ 11.95470 TL T* [ (image\055le) 24.99090 (v) 14.98280 (el) -354.99700 (associations\056) -624.98900 (Furthermore\054) -381 (it) -356.00400 (allo) 24.99090 (ws) -354.99000 (us) -354.99000 (to) -354.99000 (re\055) ] TJ T* [ (solv) 15 (e) -254.98200 (coreference) -255.00400 (ambiguity) -255.00600 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 427.54600 201.63300 Tm (19) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 437.50900 201.63300 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 442.54000 201.63300 Tm (35) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 452.50300 201.63300 Tm [ (\135) -255.02100 (and) -255.01100 (to) -255.00600 (understand) -254.99200 (ob\055) ] TJ -143.64100 -11.95510 Td [ (ject) -269.00900 (distrib) 19.99180 (utions) -269.01900 (in) -269.00200 (QA\054) -269.00700 (and) -269.00900 (enables) -269.98400 (vis) 1 (ually) -270.01400 (grounded) -268.98900 (an\055) ] TJ 11.95510 TL T* [ (swers) -249.99000 (that) -249.98300 (consist) -250.00500 (of) -249.99500 (object) -249.98300 (bounding) -249.99300 (box) 14.99260 (es\056) ] TJ 11.95510 -13.03710 Td [ (Moti) 25.00560 (v) 24.98110 (ated) -450 (by) -449.99400 (the) -449.99600 (goal) -449.98900 (of) -450.00300 (de) 25.01540 (v) 14.98280 (eloping) -450.01300 (a) -450.98800 (model) -450.01800 (for) -450.00300 (vi\055) ] TJ -11.95510 -11.95470 Td [ (sual) -336.00800 (QA) -334.99800 (based) -336.01500 (on) -334.99300 (grounded) -335.99800 (re) 15.00730 (gions\054) -357.00900 (our) -334.99800 (paper) -335.98300 (introduces) ] TJ T* [ (a) -381 (dataset) -380.00300 (that) -381.02000 (e) 15.01220 (xtends) -381.01000 (pre) 25.01050 (vi) 0.99003 (ous) -381.02000 (approaches) -380.99000 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 491.12800 140.77600 Tm (1) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 496.10900 140.77600 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 502.38600 140.77600 Tm (7) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 507.36700 140.77600 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 513.65400 140.77600 Tm (36) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 523.61600 140.77600 Tm [ (\135) -381 (and) ] TJ -214.75400 -11.95510 Td [ (proposes) -240.99400 (an) -240 (attention\055based) -241.01900 (model) -240.99400 (to) -239.98900 (perform) -241.00900 (this) -240.99400 (task\056) -306.99300 (W) 79.98660 (e) ] TJ 11.95590 TL T* [ (collected) -274.01800 (327\054939) -272.98100 (QA) -273.98900 (pairs) -273.00300 (on) -273.98400 (47\054300) -274.00300 (COCO) -272.97900 (images) -274.01300 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 529.34400 116.86500 Tm (25) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R50 9.96260 Tf 1 0 0 1 539.30700 116.86500 Tm (\135\054) Tj -230.44500 -11.95510 Td [ (together) -262.99800 (with) -262.01300 (1\054311\054756) -263.01500 (human\055generated) -262.99500 (multiple\055choices) ] TJ 11.95510 TL T* [ (and) -207.01900 (561\054459) -206.99200 (object) -207.01000 (groundings) -207.01900 (from) -207.01400 (36\054579) -206.99500 (cate) 15.01220 (gories\056) -296.00900 (Our) ] TJ T* [ (data) -356.01700 (collection) -355.98700 (w) 10 (as) -356.01700 (inspired) -357.00900 (by) -356.00900 (the) -356.00900 (age\055old) -356.00900 (idea) -356.01400 (of) -357 (the) ] TJ /R54 9.96260 Tf 227.95100 0 Td (W) Tj /R50 9.96260 Tf -241.69200 -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 (4995) 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 /BBox [ 3988.38000 5013.09000 4025.72000 5064.68000 ] /Resources << /ColorSpace << /R58 19 0 R >> /ExtGState << /R66 21 0 R >> >> /Length 130 /Group 22 0 R /Type /XObject /FormType 1 /Matrix [ 1 0 0 1 0 0 ] /Subtype /Form >> stream x;@ާ±> -8@ģHdJfaFo2`Z4̞4(mY)(MW/\X ORK7@s R,Su:?ҙl) endstream endobj 19 0 obj [ /ICCBased 20 0 R ] endobj 20 0 obj << /Length 3144 /N 3 >> stream HLino mntrRGB XYZ 1 acspMSFT IEC sRGB -HP cprt P 3desc lwtpt bkpt rXYZ gXYZ , bXYZ @ dmnd T pdmdd vued L view $lumi meas $tech 0 rTRC <