%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 ] /Type /Pages /Count 9 >> endobj 2 0 obj << /Title (VQA\072 Visual Question Answering) /Producer (PyPDF2) /Author (Stanislaw Antol\054 Aishwarya Agrawal\054 Jiasen Lu\054 Margaret Mitchell\054 Dhruv Batra\054 C\056 Lawrence Zitnick\054 Devi Parikh) /Subject (2015 IEEE International Conference on Computer Vision) >> endobj 3 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 13 0 R /Resources << /XObject << /x8 14 0 R /x6 17 0 R /R50 20 0 R /R51 21 0 R /R52 22 0 R /R53 23 0 R /x12 24 0 R /x10 27 0 R >> /ExtGState << /s9 30 0 R /s11 33 0 R /a0 << /CA 1 /ca 1 >> /R25 36 0 R /s5 37 0 R /s7 40 0 R >> /Font << /R48 43 0 R /R54 47 0 R /R46 51 0 R /R44 55 0 R /R42 58 0 R /R40 62 0 R /F2 66 0 R /R28 67 0 R /F1 71 0 R /R38 72 0 R /R36 75 0 R /R34 79 0 R /R32 82 0 R /R30 85 0 R /R26 89 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 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 ] >> endobj 4 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 110 0 R /Resources << /ExtGState << /R25 36 0 R >> /Font << /R46 51 0 R /R40 62 0 R /F2 111 0 R /R28 67 0 R /F1 112 0 R /R26 89 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 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 145 0 R 146 0 R 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 ] >> endobj 5 0 obj << /Parent 1 0 R /Rotate 0 /Contents 168 0 R /Resources << /XObject << /R157 169 0 R /R154 171 0 R /R155 172 0 R /R152 173 0 R /R156 170 0 R /R144 174 0 R /R147 175 0 R /R146 176 0 R /R141 177 0 R /R153 178 0 R /R150 179 0 R /R151 180 0 R /R142 181 0 R /R149 182 0 R /R148 183 0 R /R158 184 0 R /R143 186 0 R >> /ExtGState << /R138 185 0 R /R25 36 0 R >> /Font << /R40 62 0 R /F2 192 0 R /R28 67 0 R /F1 193 0 R /R38 72 0 R /R139 187 0 R /R26 89 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /Group 191 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 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 ] /Type /Page >> endobj 6 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 209 0 R /Resources << /ExtGState << /R25 36 0 R >> /Font << /F2 210 0 R /R175 211 0 R /R42 58 0 R /R173 215 0 R /R171 218 0 R /R28 67 0 R /F1 221 0 R /R46 51 0 R /R34 79 0 R /R40 62 0 R /R26 89 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 222 0 R 223 0 R 224 0 R 225 0 R 226 0 R 227 0 R ] >> endobj 7 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 228 0 R /Resources << /XObject << /R200 229 0 R /R201 230 0 R /R198 233 0 R /R199 234 0 R >> /ExtGState << /R193 235 0 R /R190 236 0 R >> /Font << /R202 237 0 R /R28 67 0 R /F1 241 0 R /R196 242 0 R /F2 246 0 R /R46 51 0 R /R171 218 0 R /R26 89 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R192 247 0 R /R195 231 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ 249 0 R 250 0 R 251 0 R 252 0 R 253 0 R 254 0 R ] >> endobj 8 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 255 0 R /Resources << /XObject << /R215 256 0 R /R214 257 0 R /R217 258 0 R /R216 259 0 R /R220 260 0 R /R221 261 0 R /R219 262 0 R /R218 263 0 R >> /ExtGState << /R25 36 0 R >> /Font << /R202 237 0 R /R28 67 0 R /F1 264 0 R /R171 218 0 R /F2 265 0 R /R46 51 0 R /R40 62 0 R /R42 58 0 R /R26 89 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 266 0 R 267 0 R 268 0 R ] >> endobj 9 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 269 0 R /Resources << /ExtGState << /R25 36 0 R >> /Font << /F2 270 0 R /R28 67 0 R /F1 271 0 R /R40 62 0 R /R46 51 0 R /R234 272 0 R /R171 218 0 R /R26 89 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 275 0 R 276 0 R 277 0 R 278 0 R 279 0 R 280 0 R 281 0 R ] >> endobj 10 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 282 0 R /Resources << /ExtGState << /R25 36 0 R >> /Font << /F2 283 0 R /R28 67 0 R /F1 284 0 R /R46 51 0 R /R40 62 0 R /R42 58 0 R /R171 218 0 R /R26 89 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 285 0 R 286 0 R 287 0 R 288 0 R 289 0 R ] >> endobj 11 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 290 0 R /Resources << /ExtGState << /R25 36 0 R >> /Font << /F2 291 0 R /R28 67 0 R /F1 292 0 R /R40 62 0 R /R26 89 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 293 0 R 294 0 R 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 308 0 R 309 0 R 310 0 R 311 0 R 312 0 R 313 0 R 314 0 R 315 0 R 316 0 R 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 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 341 0 R 342 0 R 343 0 R 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 ] >> endobj 12 0 obj << /Type /Catalog /Pages 1 0 R >> endobj 13 0 obj << /Length 22009 >> stream q q q 0.10000 0 0 0.10000 0 0 cm /R25 gs 0 g q 10 0 0 10 0 0 cm BT /R26 14.34620 Tf 1 0 0 1 194.06200 675.78400 Tm [ (VQA\072) -249.99500 (V) 37.01370 (isual) -250.00500 (Question) -249.99600 (Answering) ] TJ /R28 11.95520 Tf -84.86020 -37.30000 Td [ (Stanisla) 14.99840 (w) -249.98800 (Antol) ] TJ /R30 7.97010 Tf 76.53590 4.33906 Td [ <03> -0.29866 ] TJ /R32 7.97010 Tf 4.23398 0 Td [ (1) -0.29866 ] TJ /R28 11.95520 Tf 19.67700 -4.33906 Td [ (Aishw) 10.01460 (arya) -249.99700 (Agra) 14.99740 (w) 10 (al) ] TJ /R30 7.97010 Tf 95.52110 4.33906 Td [ <03> -0.30019 ] TJ /R32 7.97010 Tf 4.23398 0 Td [ (1) -0.30019 ] TJ /R28 11.95520 Tf 19.67620 -4.33906 Td [ (Jiasen) -250.01000 (Lu) ] TJ /R32 7.97010 Tf 45.48910 4.33906 Td [ (1) -0.30019 ] TJ /R28 11.95520 Tf 19.67580 -4.33906 Td [ (Mar) 17.98710 (g) 4.98688 (aret) -250.01000 (Mitchell) ] TJ /R32 7.97010 Tf 87.04530 4.33906 Td [ (2) -0.30019 ] TJ /R28 11.95520 Tf -313.13100 -17.83090 Td [ (Dhruv) -250.00400 (Batra) ] TJ /R32 7.97010 Tf 59.42890 4.33867 Td [ (1) -0.30019 ] TJ /R28 11.95520 Tf 19.67620 -4.33867 Td [ (C\056) -249.98700 (La) 14.98510 (wrence) -250.00400 (Zitnick) ] TJ /R32 7.97010 Tf 98.42620 4.33867 Td [ (2) -0.30019 ] TJ /R28 11.95520 Tf 19.67580 -4.33867 Td [ (De) 24.98950 (vi) -249.99100 (P) 15.01580 (arikh) ] TJ /R32 7.97010 Tf 56.96600 4.33867 Td [ (1) -0.30019 ] TJ -216.38300 -13.49180 Td [ (1) -0.30019 ] TJ /R28 11.95520 Tf 4.73164 -4.33789 Td [ (V) 59.99160 (ir) 18 (ginia) -250.00200 (T) 70.01640 (ech) ] TJ /R32 7.97010 Tf 79.90820 4.33789 Td [ (2) -0.30019 ] TJ /R28 11.95520 Tf 4.73203 -4.33789 Td [ (Microsoft) -250.01000 (Research) ] TJ /R34 5.97760 Tf -214.91100 -9.58125 Td [ (1) -0.10006 ] TJ /R36 8.96640 Tf 4.15078 -3.80898 Td [ (f) -0.89990 ] TJ /R38 8.96640 Tf 4.60820 0 Td [ (santol\054) -599.99400 (aish\054) -599.98600 (jiasenlu\054) -600.00500 (dbatra\054) -599.99400 (parikh) ] TJ /R36 8.96640 Tf 204.43300 0 Td [ (g) -0.90126 ] TJ /R38 8.96640 Tf 4.60820 0 Td (\100vt\056edu) Tj /R34 5.97760 Tf 52.60200 3.80898 Td [ (2) -0.09802 ] TJ /R36 8.96640 Tf 4.15078 -3.80898 Td [ (f) -0.90126 ] TJ /R38 8.96640 Tf 4.60820 0 Td [ (memitc\054) -599.99400 (larryz) ] TJ /R36 8.96640 Tf 75.31800 0 Td [ (g) -0.89854 ] TJ /R38 8.96640 Tf 4.60703 0 Td (\100microsoft\056com) Tj /R26 11.95520 Tf -293.50100 -40.87890 Td (Abstract) Tj /R40 9.96260 Tf -95.88280 -26.42110 Td [ (W) 91.98650 (e) -386.01800 (pr) 44.98390 (opose) -385.99700 (the) -386.98700 (task) -386 (of) ] TJ /R28 9.96260 Tf 100.12500 0 Td (free\055form) Tj /R40 9.96260 Tf 42.01410 0 Td (and) Tj /R28 9.96260 Tf 19.06090 0 Td (open\055ended) Tj /R40 9.96260 Tf 50.32300 0 Td [ (V) 73.99260 (isual) ] TJ -211.52300 -11.47700 Td [ (Question) -277.99800 (Answering) -277.98400 (\050VQA\051\056) -277.00600 (Given) -277.98700 (an) -277.98300 (ima) 10.01300 (g) 10.00320 (e) -277.99800 (and) -278.01700 (a) -277.98800 (natur) 15.00730 (al) ] TJ 11.47700 TL T* [ (langua) 9.99159 (g) 10.00320 (e) -216.98800 (question) -218.00700 (about) -216.99600 (the) -217.99500 (ima) 10.01300 (g) 10.00320 (e) 9.99343 (\054) -224 (the) -217.01500 (task) -217.00800 (is) -217.99100 (to) -217.01000 (pr) 44.98510 (o) 10.00320 (vide) -218 (an) ] TJ T* [ (accur) 14.98520 (ate) -394.00500 (natur) 15.00610 (al) -394 (langua) 9.99098 (g) 10.00320 (e) -394.01800 (answer) 110.99300 (\056) -741.99100 (Mirr) 44.99490 (oring) -393.98300 (r) 37.01830 (eal\055world) ] TJ T* [ (scenarios\054) -363.00300 (suc) 14.98520 (h) -340.01700 (as) -340.01100 (helping) -339.99100 (the) -340.99500 (visually) -339.99200 (impair) 37.00850 (ed\054) -362.01600 (both) -341.01900 (the) ] TJ T* [ (questions) -206.00200 (and) -205.99000 (answer) 9.98975 (s) -205.99900 (ar) 36.98520 (e) -206.01000 (open\055ended\056) -294.99200 (V) 73.99260 (isual) -206.01500 (questions) -206.00200 (se\055) ] TJ T* [ (lectively) -261.98900 (tar) 36.99260 (g) 10.00320 (et) -261.99300 (dif) 18.01660 (fer) 36.98280 (ent) -261.98800 (ar) 36.98520 (eas) -261.98800 (of) -261.98300 (an) -261.98600 (ima) 10.01300 (g) 10.00320 (e) 9.99343 (\054) -265.99200 (including) -261.99300 (bac) 20.00160 (k\055) ] TJ T* [ (gr) 44.98390 (ound) -223.98200 (details) -222.98900 (and) -223.98700 (underlying) -222.98500 (conte) 20.00520 (xt\056) -301.00400 (As) -223.99200 (a) -223.01700 (r) 37.01830 (esult\054) -228.99600 (a) -223.99700 (system) ] TJ T* [ (that) -343.01600 (succeeds) -342.99700 (at) -342.98900 (VQA) -342.99100 (typically) -343.01500 (needs) -342.99400 (a) -344.01600 (mor) 36.98890 (e) -343.00700 (detailed) -343.01400 (un\055) ] TJ T* [ (der) 10.00510 (standing) -213.01700 (of) -211.99200 (the) -213.01600 (ima) 10.01300 (g) 10.00320 (e) -212.98900 (and) -211.98900 (comple) 20.00770 (x) -212.98900 (r) 37.01830 (easoning) -212.00900 (than) -213.00100 (a) -213.01800 (sys\055) ] TJ T* [ (tem) -325 (pr) 44.98390 (oducing) -325.01600 (g) 10.00320 (eneric) -326.01900 (ima) 10.01300 (g) 10.00320 (e) -325.00900 (captions\056) -535.98900 (Mor) 36.98650 (eo) 9.99833 (ver) 110.99900 (\054) -344.01900 (VQA) -324.99500 (is) ] TJ T* [ (amenable) -282.99000 (to) -283 (automat) 1.01699 (ic) -283.00900 (e) 15.01220 (valuation\054) -291.00400 (since) -283.00200 (many) -283.00400 (open\055ended) ] TJ T* [ (answer) 9.99037 (s) -213.01700 (contain) -212.99300 (only) -211.99200 (a) -213.01800 (fe) 14.98030 (w) -213.01000 (wor) 36.99870 (ds) -213.01200 (or) -213.01200 (a) -211.99900 (closed) -213.00900 (set) -213.01300 (of) -213.01100 (answer) 9.99098 (s) ] TJ T* [ (that) -312.98200 (can) -313.98100 (be) -313.00600 (pr) 44.98390 (o) 10.00320 (vided) -312.99800 (in) -314.01400 (a) -313.00200 (multiple\055c) 15.00730 (hoice) -313 (format\056) -499.99700 (W) 91.98710 (e) -313.99200 (pr) 44.98510 (o\055) ] TJ T* [ (vide) -344.01700 (a) -344.01600 (dataset) -343.01000 (containing) ] TJ /R42 9.96260 Tf 106.68300 0 Td [ (\030) -0.80011 ] TJ /R40 9.96260 Tf 7.74805 0 Td [ (0\05625M) -344.00100 (ima) 10.01300 (g) 10.00320 (es\054) ] TJ /R42 9.96260 Tf 63.33320 0 Td [ (\030) -0.79889 ] TJ /R40 9.96260 Tf 7.74883 0 Td [ (0\05676M) -344.00100 (ques\055) ] TJ -185.51300 -11.47700 Td [ (tions\054) -436.01200 (and) ] TJ /R42 9.96260 Tf 45.12300 0 Td [ (\030) -0.80011 ] TJ /R40 9.96260 Tf 7.74883 0 Td [ (10M) -398 (answer) 9.98975 (s) -398.98600 (\050) ] TJ ET Q 0 1 0 0 k q 10 0 0 10 0 0 cm BT /R44 9.96260 Tf 1 0 0 1 165.06000 358.65600 Tm (www\056visualqa\056org) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R40 9.96260 Tf 1 0 0 1 261.26900 358.65600 Tm [ (\051\054) -436.00600 (and) ] TJ -211.15700 -11.47730 Td [ (discuss) -236.98800 (the) -237.99200 (information) -236.99600 (it) -237.00500 (pr) 44.98390 (o) 10.00320 (vides\056) -306.01500 (Numer) 44.99490 (ous) -237.00200 (baselines) -238.00700 (for) ] TJ 11.47700 TL T* [ (VQA) -245.98800 (ar) 36.98520 (e) -246.00200 (pr) 44.98390 (o) 10.00320 (vided) -245.99000 (and) -245.98300 (compar) 36.99380 (ed) -245.99700 (with) -246.00400 (human) -246.01500 (performance) 15.00730 (\056) ] TJ /R26 11.95520 Tf 23.68200 TL T* [ (1\056) -249.99000 (Intr) 18.01460 (oduction) ] TJ /R28 9.96260 Tf 15.67890 TL T* [ (W) 79.98730 (e) -230.98500 (are) -230.99300 (witnessing) -229.99200 (a) -230.98600 (rene) 25.00810 (wed) -231.01700 (e) 15.01220 (xcitement) -230.98800 (in) -231.00800 (multi\055discipline) ] TJ 11.47700 TL T* [ <41727469026369616c> -234.98600 (Intelligence) -236.01600 (\050AI\051) -236.00700 (res) 1.01821 (earch) -236.01100 (problems\056) -305.01300 (In) -235.99700 (particular) 40.00560 (\054) ] TJ T* [ (research) -208.01700 (in) -207.01200 (image) -207.98900 (and) -207.99800 (video) -207.98600 (captioning) -206.98600 (that) -207.99000 (combines) -208.01200 (Com\055) ] TJ T* [ (puter) -443.99500 (V) 59.99310 (ision) -444.99300 (\050CV\051\054) -443.99200 (Natural) -445.01400 (Language) -443.99500 (Processing) -444.98700 (\050NLP\051\054) ] TJ T* [ (and) -241.01300 (Kno) 24.98910 (wledge) -241.00800 (Representation) -241.01700 (\046) -241.00600 (Reasoning) -240.98200 (\050KR\051) -241.98900 (has) -241.01100 (dra\055) ] TJ T* [ (matically) -206.01100 (increased) -204.98300 (in) -205.99300 (the) -205.01700 (past) -205.99100 (year) -206.01200 (\133) ] TJ ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 193.54400 238.95600 Tm (13) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 203.50600 238.95600 Tm (\054) Tj ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 208.03900 238.95600 Tm (7) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 213.02000 238.95600 Tm (\054) Tj ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 217.56300 238.95600 Tm (9) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 222.54500 238.95600 Tm (\054) Tj ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 227.08800 238.95600 Tm (32) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 237.05000 238.95600 Tm (\054) Tj ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 241.58300 238.95600 Tm (21) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 251.54500 238.95600 Tm (\054) Tj ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 256.08800 238.95600 Tm (19) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 266.05100 238.95600 Tm (\054) Tj ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 270.59400 238.95600 Tm (45) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 280.55600 238.95600 Tm (\135\056) Tj -230.44400 -11.47810 Td [ (P) 14.99200 (art) -509.01400 (of) -509.01100 (this) -508.00700 (e) 15.01220 (xcitement) -508.98200 (stems) -508.98600 (from) -507.98300 (a) -509.01800 (belief) -509.00600 (that) -508.99800 (multi\055) ] TJ 11.47700 TL T* [ (discipline) -408.00900 (tasks) -407.01000 (lik) 10.01790 (e) -408.01500 (image) -407.99400 (captioning) -408.01300 (are) -407.00300 (a) -408.01500 (step) -407.99600 (to) 24.98850 (w) 10 (ards) ] TJ T* [ (solving) -396.01100 (AI\056) -396.01000 (Ho) 24.98360 (we) 25.01420 (v) 14.98280 (er) 39.98600 (\054) -432.98400 (the) -396.00500 (current) -396.01000 (state) -396 (of) -396.01000 (the) -396.98500 (art) -396.01200 (demon\055) ] TJ T* [ (strates) -288.99600 (that) -290.01500 (a) -289.01500 (coarse) -289.99600 (scene\055le) 25.01300 (v) 14.98280 (el) -289.00800 (understanding) -289.98600 (of) -289.00800 (an) -289.99100 (image) ] TJ T* [ (paired) -275.98800 (with) -276.01000 (w) 10 (ord) ] TJ /R46 9.96260 Tf 71.23790 0 Td [ (n) -0.19972 ] TJ /R28 9.96260 Tf 5.98008 0 Td [ (\055gram) -275.98700 (statistics) -275.98700 (suf) 24.98360 <02636573> -277 (to) -275.98100 (generate) -275.99600 (rea\055) ] TJ -77.21800 -11.47700 Td [ (sonable) -350.98100 (image) -350.98400 (captions\054) -377.00600 (which) -350.98400 (suggests) -351.00300 (image) -352.00500 (captioning) ] TJ T* [ (may) -249.98500 (not) -250.02000 (be) -249.99700 (as) -249.99600 (\223) 79.98180 (AI\055complete\224) -249.98200 (as) -249.99600 (desired\056) ] TJ 14.46600 TL T* [ (What) -455.99400 (mak) 10.01120 (es) -456.98100 (for) -456.00400 (a) -456.98700 (compelling) -455.99600 (\223) 79.98180 (AI\055complete\224) -455.98800 (task\077) -929.99200 (W) 79.98660 (e) ] TJ 11.47700 TL T* [ (belie) 24.98660 (v) 14.98280 (e) -412.01400 (that) -411.99400 (in) -411.99700 (order) -412.99100 (to) -411.99700 (spa) 14.98520 (wn) -412.00200 (the) -412.00200 (ne) 15.01710 (xt) -411.99700 (generation) -413.01200 (of) -412.00700 (AI) ] TJ T* [ (algorithms\054) -455.01200 (an) -414.00800 (ideal) -415.01800 (task) -413.99600 (should) -827.98600 (\050i\051) -414.98800 (requir) 0.98758 (e) ] TJ /R40 9.96260 Tf 187.54300 0 Td (multi\055modal) Tj -187.54300 -11.47700 Td [ (knowledg) 9.99098 (e) ] TJ /R28 9.96260 Tf 45.99370 0 Td [ (be) 15.01830 (yond) -350 (a) -349.98600 (single) -349.01400 (sub\055domain) -350.01500 (\050such) -350.01300 (as) -348.99800 (CV\051) -350.00800 (and) ] TJ -45.99370 -11.47700 Td [ (\050ii\051) -410.98200 (ha) 19.99730 (v) 14.98280 (e) -410.99500 (a) -410.99500 <77656c6c2d6465026e6564> ] TJ /R40 9.96260 Tf 101.24400 0 Td [ (quantitative) -411.01600 (e) 15.01220 (valuation) -410.98800 (metric) ] TJ /R28 9.96260 Tf 127.25500 0 Td (to) Tj ET Q 3.98000 w 0 G 501.12100 904.14800 m 1446.11000 904.14800 l S q 10 0 0 10 0 0 cm BT /R48 5.97760 Tf 1 0 0 1 64.45780 83.81290 Tm [ <03> -0.90058 ] TJ /R28 7.97010 Tf 6.31016 -2.81289 Td [ (The) -249.98800 <02727374> -250.01700 (tw) 10.02130 (o) -250.02000 (authors) -250.01700 (contrib) 19.99660 (uted) -250.01900 (equally) 64.97830 (\056) ] TJ ET Q q 3088.62000 3806.89000 2362.40000 1865.05000 re W* n 1 g 3088.62000 3806.89000 2362.40000 1865.05000 re f* q 1176.64000 0 0 755.55300 3103.96000 3960.65000 cm /R50 Do Q 3.31565 w 1 j 3102.30000 3959 1179.96000 758.86700 re S q 1134.78000 0 0 755.13900 3104.78000 4907.69000 cm /R51 Do Q 3103.13000 4906.03000 1138.10000 758.45300 re S q 1007.13000 0 0 755.13900 4410.73000 4907.69000 cm /R52 Do Q 4409.08000 4906.03000 1010.44000 758.45300 re S q 1002.57000 0 0 755.55300 4397.06000 3960.65000 cm /R53 Do Q 4395.40000 3959 1005.89000 758.86700 re S 0 g q 10 0 0 10 0 0 cm BT /R54 5.52055 Tf 1 0 0 1 439.83000 390.32100 Tm [ (Does i) -2.98069 (t appear) 4.99730 ( ) -6.02330 (to ) -5.97908 (be ) -4.99730 (ra) 2.96300 (i) -2.01661 (ny) 12.02890 (\077) ] TJ /R54 5.51227 Tf 6.63984 TL T* [ (Does) -2.96746 ( thi) -2.99403 (s) -2.99403 ( person) -1.99307 ( ) -7.01560 (hav) -3.98614 (e 20\05720 v) -4.97824 (i) -2.99403 (s) -2.99403 (i) -2.99403 (on\077) ] TJ -129.25200 6.51523 Td [ (Is) -2.99846 ( ) -7.01560 (thi) -2.99846 (s) -2.99846 ( ) 7.01560 (person) -2.01521 ( ) -7.01560 (ex) 3.02061 (pec) -4 (ti) -2.99403 (ng ) -7.02446 (c) -2.99846 (ompan) -2.01521 (y) 11.03270 (\077) ] TJ 6.63125 TL T* [ (W) -1.97092 (hat ) -8.01657 (i) -2.99403 (s) -2.99846 ( ) 7.01560 (j) -2.99403 (us) -3.00289 (t under ) -5.96149 (the ) -8.01657 (tree\077) ] TJ 130.67900 101.02100 Td [ (How) 7.98999 ( ) -7.01560 (many) 9.03525 ( ) -21.96810 (s) -2.99403 (l) -2.99403 (i) -2.99403 (c) -2.99403 (es) -2.99403 ( ) 21.96810 (of ) -7.01560 (pi) -2.99403 (z) -2.99403 (z) -2.99403 (a ) 7.01560 (are ) -12.95940 (there\077) ] TJ T* [ (Is) -2.99403 ( ) -7.01560 (thi) -2.99403 (s) -2.99403 ( ) 7.01560 (a ) -7.01560 (v) -2.99403 (egetari) -1.99307 (an pi) -3.98614 (z) -2.99403 (z) -2.99403 (a\077) ] TJ -130.30600 6.44805 Td [ (W) -1.97092 (hat ) -8.01657 (c) -2.99846 (ol) -2.99846 (or ar) 2.98518 (e ) -7.02003 (her ) -4.96053 (ey) 11.02830 (es) -3.00289 (\077) ] TJ T* [ (W) -1.97092 (hat ) -8.01657 (i) -2.99403 (s) -2.99846 ( ) 7.01560 (the ) -8.01657 (mus) -5 (tac) -3.00289 (he ) -7.02446 (made ) -9.02639 (of\077) ] TJ ET Q Q q 10 0 0 10 0 0 cm BT /R28 8.96640 Tf 1 0 0 1 308.86200 371.33200 Tm [ (Figure) -278.99800 (1\072) -367.98900 (Examples) -279.01500 (of) -279.00700 (free\055form\054) -285.99600 (open\055ended) -278.97700 (questions) -277.98600 (collected) ] TJ 10.52110 TL T* [ (for) -409.01900 (images) -409.00900 (via) -409.00300 (Amazon) -407.98500 (Mechanical) -409.01900 (T) 44.99760 (urk\056) -787.00400 (Note) -409.01400 (that) -407.99000 (common\055) ] TJ T* [ (sense) -331.99800 (kno) 25.00660 (wledge) -332.00700 (is) -331.99800 (needed) -331.98200 (along) -331.97900 (with) -332.00700 (a) -331.98500 (visual) -332.00700 (understanding) -332.00100 (of) ] TJ 10.51990 TL T* [ (the) -249.98900 (scene) -249.98700 (to) -250.01400 (answer) -249.99200 (man) 15.01920 (y) -249.97800 (questions\056) ] TJ /R28 9.96260 Tf 20.26290 TL T* [ (track) -317.98600 (progress\056) -514.99200 (F) 14.99260 (or) -317.98400 (some) -318.98600 (tasks\054) -335.01800 (such) -318.01600 (as) -317.98600 (image) -318.99100 (captioning\054) ] TJ 11.47700 TL T* [ (automatic) -397.98000 (e) 25.01050 (v) 24.98110 (aluation) -397.00500 (is) -398 (still) -396.99700 (a) -398.01700 (dif) 24.98600 <0263756c74> -397.00200 (and) -398.00700 (open) -398.00200 (research) ] TJ T* [ (problem) -250.01000 (\133) ] TJ ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 347.87600 296.55300 Tm (43) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 357.83800 296.55300 Tm (\054) Tj ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 362.82000 296.55300 Tm (10) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 372.78200 296.55300 Tm (\054) Tj ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 377.76300 296.55300 Tm (18) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 387.72600 296.55300 Tm (\135\056) Tj -78.86370 -14.46600 Td [ (In) -291.98800 (this) -292.98500 (paper) 39.99090 (\054) -303.00600 (we) -291.99300 (introduce) -292.98500 (the) -291.98100 (task) -292.01500 (of) ] TJ /R40 9.96260 Tf 156.09300 0 Td [ (fr) 36.98890 (ee\055form) ] TJ /R28 9.96260 Tf 40.17300 0 Td (and) Tj /R40 9.96260 Tf 17.29880 0 Td (open\055) Tj -213.56500 -11.47700 Td (ended) Tj /R28 9.96260 Tf 28.56880 0 Td [ (V) 59.99240 (isual) -453 (Question) -452.00800 (Answering) -452.99800 (\050VQA\051\056) -451.99800 (A) -453.01800 (VQA) -451.99300 (sys\055) ] TJ -28.56880 -11.47700 Td [ (tem) -332.00800 (tak) 10.00570 (es) -331.98100 (as) -331.98100 (input) -331.99400 (an) -331.98400 (image) -330.98900 (and) -332.01800 (a) -331.98900 (free\055form\054) -353.00500 (open\055ended\054) ] TJ 11.47700 TL T* [ (natural\055language) -316.98100 (question) -317.01100 (about) -316.98600 (the) -316.99600 (image) -316.99100 (and) -317.00100 (produces) -316.99100 (a) ] TJ T* [ (natural\055language) -423.98300 (answer) -424.00500 (as) -424.00800 (the) -423.99800 (output\056) -832.99800 (This) -423.98300 (goal\055dri) 24.99580 (v) 14.98280 (en) ] TJ T* [ (task) -324.99000 (is) -324.99200 (applicable) -324.00500 (to) -324.99200 (scenarios) -324.99000 (encountered) -324.99500 (when) -324.99500 (visually\055) ] TJ T* [ (impaired) -246.99300 (users) -247.98800 (\133) ] TJ ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 373.00100 213.22500 Tm (2) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 377.98200 213.22500 Tm [ (\135) -246.98100 (or) -247.99500 (intelligence) -246.98300 (analysts) -248.00800 (acti) 24.98110 (v) 14.98280 (ely) -247.00800 (elicit) -248.01700 (vi\055) ] TJ -69.12030 -11.47700 Td [ (sual) -249.98300 (information\056) -309.98000 (Example) -250.01200 (questions) -249.99300 (are) -250.00700 (sho) 24.99090 (wn) -249.98800 (in) -249.98800 (Fig\056) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 529.48400 201.74800 Tm (1) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 534.46500 201.74800 Tm (\056) Tj -225.60300 -14.46600 Td [ (Open\055ended) -566.01100 (questions) -567 (require) -565.98400 (a) -567.00800 (potentially) -566.01300 (v) 24.98110 (ast) -566.01300 (set) -566.99400 (of) ] TJ 11.47700 TL T* [ (AI) -370.99700 (capabilities) -369.98200 (to) -370.98500 (answer) -370.99700 (\226) -370.01200 <026e652d677261696e6564> -371.01700 (recognition) -371.00200 (\050) ] TJ /R40 9.96260 Tf 219.52300 0 Td [ (e) 15.01220 (\056g) ] TJ /R28 9.96260 Tf 11.74570 0 Td (\056\054) Tj -231.26900 -11.47700 Td [ (\223What) -358.01400 (kind) -357.99700 (of) -358.01600 (cheese) -357.00700 (is) -358.00600 (on) -358.00900 (the) -358.00900 (pizza\077\224\051\054) -385.00400 (object) -358.00400 (detection) ] TJ (\050) ' /R40 9.96260 Tf 3.31797 0 Td [ (e) 15.01220 (\056g) ] TJ /R28 9.96260 Tf 11.74610 0 Td [ (\056\054) -426.98300 (\223Ho) 25.01790 (w) -390.99100 (man) 14.99010 (y) -390.98900 (bik) 10.01550 (es) -390.99100 (are) -391.98400 (there\077\224\051\054) -426.01700 (acti) 24.98110 (vity) -391.98800 (recognition) ] TJ -15.06410 -11.47730 Td (\050) Tj /R40 9.96260 Tf 3.31797 0 Td [ (e) 15.01220 (\056g) ] TJ /R28 9.96260 Tf 11.74610 0 Td [ (\056\054) -511.00800 (\223Is) -458.00400 (this) -458.99700 (man) -459.00900 (crying\077\224\051\054) -509.98800 (kno) 24.99090 (wledge) -459.01200 (base) -459.02100 (reasoning) ] TJ -15.06410 -11.47700 Td (\050) Tj /R40 9.96260 Tf 3.31797 0 Td [ (e) 15.01220 (\056g) ] TJ /R28 9.96260 Tf 11.74610 0 Td [ (\056\054) -289.00600 (\223Is) -281.99500 (this) -280.98700 (a) -281.01700 (v) 14.98280 (e) 15.01220 (getarian) -281.98200 (pizza\077\224\051\054) -288.98100 (and) -281.00700 (commonsense) -281.98700 (rea\055) ] TJ -15.06410 -11.47700 Td [ (soning) -232.98600 (\050) ] TJ /R40 9.96260 Tf 32.20390 0 Td [ (e) 15.01220 (\056g) ] TJ /R28 9.96260 Tf 11.74610 0 Td [ (\056\054) -235.99500 (\223Does) -233.01500 (this) -232.01500 (person) -233.00500 (ha) 19.99670 (v) 14.98280 (e) -232.00500 (20\05720) -232.99100 (vision\077\224\054) -236.01000 (\223Is) -232 (this) ] TJ -43.95000 -11.47700 Td [ (person) -249.98300 (e) 15.01220 (xpecting) -250.01200 (compan) 14.99750 (y\077\224\051\056) ] TJ 14.46600 TL T* [ (VQA) -352.01200 (\133) ] TJ ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 337.26600 92.47700 Tm (16) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 347.22800 92.47700 Tm (\054) Tj ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 353.23600 92.47700 Tm (30) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 363.19800 92.47700 Tm (\054) Tj ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 369.19600 92.47700 Tm (42) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 379.15800 92.47700 Tm (\054) Tj ET Q 0 0 1 rg q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 385.15600 92.47700 Tm (2) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R28 9.96260 Tf 1 0 0 1 390.13700 92.47700 Tm [ (\135) -351.98300 (is) -352.98800 (also) -352.00500 (amenable) -352.01000 (to) -352.01000 (automatic) -353.01000 (quan\055) ] TJ -81.27500 -11.47700 Td [ (titati) 25.00810 (v) 14.98280 (e) -392.99800 (e) 25.01050 (v) 24.98110 (aluation\054) -428.99700 (making) -393.00300 (it) -393.01800 (possible) -392.99300 (to) -394.00300 (ef) 25.00560 (fecti) 25.01540 (v) 14.98280 (ely) -392.98300 (track) ] 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 32 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 35 Tm (2425) Tj T* ET Q endstream endobj 14 0 obj << /Filter /FlateDecode /Resources << /ExtGState << /a0 << /CA 1 /ca 1 >> >> /XObject << /x18 15 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 15 0 obj << /Filter /FlateDecode /Resources 16 0 R /Length 107 /Type /XObject /BBox [ 78 746 96 765 ] /Subtype /Form >> stream xe AC̬wʠ =p,?]%+H-
Jc "82w8VSnGW;"
endstream
endobj
16 0 obj
<<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
>>
endobj
17 0 obj
<<
/Filter /FlateDecode
/Resources <<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
/XObject <<
/x15 18 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
18 0 obj
<<
/Filter /FlateDecode
/Resources 19 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
19 0 obj
<<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
>>
endobj
20 0 obj
<<
/Filter /DCTDecode
/BitsPerComponent 8
/Height 400
/Length 23003
/ColorSpace /DeviceRGB
/Width 623
/Subtype /Image
>>
stream
Adobe d C
$, !$4.763.22:ASF:=N>22HbINVX]^]8EfmeZlS[]Y C**Y;2;YYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYY o"
} !1AQa"q2#BR$3br
%&'()*456789:CDEFGHIJSTUVWXYZcdefghijstuvwxyz
w !1AQaq"2B #3Rbr
$4%&'()*56789:CDEFGHIJSTUVWXYZcdefghijstuvwxyz ?
)(NQhJ(hJ.#
U
"Ն&Ub# t*S ]qԶP 呂)f'E'Hp0G֘QIE -P!hE% RQ@E% RQ@E% RQ@E% RQ@E% RQ@-PIE -PIE s3fI$Uɍ0y{f./R}kHo嵸]JXtcUYQUFK@*y/muŕK牀4idFyǛ+n`=aVBXd`jp[3&eHDpܠe*0
Kآ9Aʲ*zai=SY81B!u4晟S[JaI!@'ڤqOAFw1's7^1R
ө6&3ڔsQ&<}㡩)1FU 5! u4P.ҀyjjS0xE#LYHE`tkrLe+9 F( ^C Ss֍֔HdyZz6cշ͖L,;U`ÊZܒ# rqJ'FTG0NBy[t$F>ʹaA^[ӣ*li\jZiP6U8'SN;bK* H$hQ(tYa^wTV}vEm9rO_aWkY5ߧIs <17~z\FbǑAc^Nÿc)C}nⴋ|$2݀CgFb%a[}=ϩ-,KY.b_zүWQQE(JZ ( ( ( ( ( ( ( ( ( ( ()(h ( ( ݧ"z
?F&s̳a
6]GG Gq0AxFGCXԛ&hĩ%R5$MgQL$ڼzmݖ>@ RԀQE!8R} -0Fih &B rM# W$0lsJ
ҩE
9OSJX%!8p iZ+gGܺcfU9}s]
c'67˿5ިMֻX>a_kÚ&a4e=N
fXP@,@7zf!mbx|P\k˯n,.
^\]K$G_#Lz[Ӳvb:VPx ^MG YƑ¬0^k~Q'Q|"Fi`Fy9ު&s*"C* }ki%&* JAf%zr;HvD鎕:A uW >HQaX
(p( (+.e ib
42ۈ2;GVW e#?B?)>TQ\#.>zn4!OG+O &V
ЯTslTTUQETG@v|_CE]9rRR${OrβTD8#?Gȿyz/m4}3ɅhB\R]OKQEQ!EPEPEPEPEPEPEPEPEPEPEPEPEPEPEPEPEP/>'جBdFN;fwgv,Ř䜓Z.ķ- ;n6@?\d۳6JJ)m*ZaC+4P`ڬJqU4݆<-; ,2m,y9)epA*ӄa8U*E^枈5ڀ=(,B$9P@k>AJFTT})B'<E*,SJðt㓁ZoJ[s)ЎGE7du92SRAQW%E8cyӌ=nVkyRXۣ!ȩkɬn&x\+ѾME%~j:ڐxw*@Ao,wF!$Cw0\Cs
,mѐ
IUȯs.v t.c[\ jrw[v2ctWt<#b (8HRƦVV eވEh,ޅ%cB\kxFWoB4%HrAs4kQEvxQEIFbe9?Ig E!ݙ ض?ɿ WQa 0qjQZ\LøP }1P8 ː{qS8EǹQ(鍉qҮ~wde*/d@r
|x8ST:(QP^\}1Gi쁻j&w1_ dKXdW 4 boŖP,sHɽ'AiK$s\Dq q䂏dxҗ4CN |ӢҬad4u9vn( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ('ĖvZ=^GAH8d ussױ^%p!CW_\W&28W5Z7F}h
#6s:) =3yrzӰMAjYR2p3֚ұb#
@IJ`njP;dc
+<=>K'3D `$XiYV5iUE,O).;4"iqޮ[-Q]2_ŹVdpN=5cRr02io!?CWv.Xٜm>4dgsE'(A:OT:(ii4
n?ߗ#W?E\jJ;39SG\