%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 (Unsupervised Semantic Parsing of Video Collections) /Producer (PyPDF2) /Author (Ozan Sener\054 Amir R\056 Zamir\054 Silvio Savarese\054 Ashutosh Saxena) /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 /x12 17 0 R /x6 20 0 R /R32 23 0 R /R29 25 0 R /R30 26 0 R /R39 27 0 R /R34 29 0 R /R35 31 0 R /R25 32 0 R /R33 30 0 R /x10 44 0 R /R31 24 0 R >> /ExtGState << /s9 47 0 R /R7 50 0 R /s5 51 0 R /s7 54 0 R /s11 57 0 R /a0 << /CA 1 /ca 1 >> /R36 41 0 R /R23 42 0 R >> /Font << /F2 60 0 R /R8 61 0 R /R10 64 0 R /R12 68 0 R /F1 71 0 R /R14 72 0 R /R16 75 0 R /R37 33 0 R /R18 78 0 R /R20 81 0 R /R26 37 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] /ColorSpace << /R43 85 0 R /R24 28 0 R /R41 87 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 4 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 89 0 R /Resources << /ColorSpace << /R43 85 0 R /R41 87 0 R >> /ExtGState << /R7 50 0 R >> /Font << /R10 64 0 R /R20 81 0 R /F1 90 0 R /F2 91 0 R /R8 61 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 5 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 92 0 R /Resources << /XObject << /R58 93 0 R /R59 94 0 R /R72 95 0 R /R73 96 0 R /R70 97 0 R /R71 98 0 R /R74 99 0 R /R75 100 0 R /R94 114 0 R /R65 103 0 R /R64 104 0 R /R67 105 0 R /R66 106 0 R /R63 107 0 R /R69 108 0 R /R68 109 0 R >> /ExtGState << /R36 41 0 R /R7 50 0 R /R23 42 0 R /R57 101 0 R /R62 102 0 R >> /Font << /F2 115 0 R /R78 116 0 R /R8 61 0 R /F1 121 0 R /R76 122 0 R /R10 64 0 R /R88 126 0 R /R60 110 0 R /R20 81 0 R /R90 131 0 R /R82 134 0 R /R92 137 0 R /R80 140 0 R /R86 143 0 R /R84 148 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R43 85 0 R /R41 87 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 6 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 153 0 R /Resources << /XObject << /R105 154 0 R /R104 155 0 R /R126 156 0 R /R125 157 0 R /R124 158 0 R /R123 159 0 R /R122 160 0 R /R121 161 0 R /R120 162 0 R /R109 164 0 R /R108 165 0 R /R149 167 0 R /R148 168 0 R /R107 166 0 R /R106 169 0 R /R133 170 0 R /R134 171 0 R /R135 172 0 R /R136 173 0 R /R137 174 0 R /R118 178 0 R /R119 163 0 R /R150 195 0 R /R151 196 0 R /R112 179 0 R /R113 181 0 R /R110 182 0 R /R111 180 0 R /R116 175 0 R /R117 177 0 R /R114 183 0 R /R115 176 0 R >> /ExtGState << /R127 191 0 R /R132 193 0 R /R62 102 0 R /R7 50 0 R /R103 192 0 R /R57 101 0 R >> /Font << /R78 116 0 R /R76 122 0 R /F2 197 0 R /R90 131 0 R /R88 126 0 R /R138 198 0 R /R82 134 0 R /R92 137 0 R /R80 140 0 R /R86 143 0 R /R129 184 0 R /R84 148 0 R /R144 201 0 R /R146 204 0 R /R140 207 0 R /R8 61 0 R /R142 212 0 R /R10 64 0 R /F1 215 0 R /R20 81 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R43 85 0 R /R41 87 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 7 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 216 0 R /Resources << /ColorSpace << /R43 85 0 R /R41 87 0 R >> /ExtGState << /R7 50 0 R >> /Font << /R78 116 0 R /R76 122 0 R /F2 217 0 R /R90 131 0 R /R88 126 0 R /R138 198 0 R /R82 134 0 R /R92 137 0 R /R80 140 0 R /R86 143 0 R /R84 148 0 R /R163 218 0 R /R161 221 0 R /R8 61 0 R /R165 225 0 R /R140 207 0 R /R159 228 0 R /R10 64 0 R /R12 68 0 R /F1 232 0 R /R14 72 0 R /R20 81 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 8 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 233 0 R /Resources << /XObject << /R175 234 0 R >> /ExtGState << /R7 50 0 R >> /Font << /F2 235 0 R /R78 116 0 R /F1 236 0 R /R140 207 0 R /R8 61 0 R /R176 237 0 R /R76 122 0 R /R173 242 0 R /R10 64 0 R /R88 126 0 R /R20 81 0 R /R90 131 0 R /R82 134 0 R /R92 137 0 R /R80 140 0 R /R86 143 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R43 85 0 R /R41 87 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 9 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 245 0 R /Resources << /XObject << /R202 246 0 R /R203 247 0 R /R200 248 0 R /R201 250 0 R /R206 251 0 R /R207 253 0 R /R204 254 0 R /R205 252 0 R /R208 256 0 R /R209 258 0 R /R328 259 0 R /R191 262 0 R /R249 263 0 R /R246 264 0 R /R247 265 0 R /R244 267 0 R /R245 268 0 R /R242 269 0 R /R320 272 0 R /R323 275 0 R /R241 270 0 R /R280 278 0 R /R285 280 0 R /R189 255 0 R /R196 322 0 R /R275 308 0 R /R239 310 0 R /R238 300 0 R /R236 323 0 R /R235 345 0 R /R234 327 0 R /R233 341 0 R /R232 344 0 R /R231 340 0 R /R230 342 0 R /R197 350 0 R /R278 302 0 R /R339 351 0 R /R273 306 0 R /R272 305 0 R /R334 385 0 R /R270 309 0 R /R277 303 0 R /R333 386 0 R /R330 368 0 R /R331 367 0 R /R318 376 0 R /R194 396 0 R /R264 294 0 R /R265 295 0 R /R228 325 0 R /R267 296 0 R /R260 297 0 R /R262 299 0 R /R186 397 0 R /R220 331 0 R /R221 332 0 R /R222 333 0 R /R223 335 0 R /R224 336 0 R /R225 337 0 R /R226 338 0 R /R227 339 0 R /R325 377 0 R /R309 371 0 R /R212 399 0 R /R302 378 0 R /R300 379 0 R /R307 381 0 R /R327 260 0 R /R305 387 0 R /R304 373 0 R /R229 330 0 R /R312 372 0 R /R314 273 0 R /R322 276 0 R /R192 334 0 R /R269 301 0 R /R315 366 0 R /R190 400 0 R /R317 369 0 R /R310 370 0 R /R214 328 0 R /R217 329 0 R /R195 257 0 R /R219 343 0 R /R218 326 0 R /R198 401 0 R /R199 249 0 R /R211 402 0 R /R258 304 0 R /R255 298 0 R /R193 403 0 R /R256 311 0 R /R250 307 0 R /R253 312 0 R /R252 313 0 R /R298 382 0 R /R291 383 0 R /R210 404 0 R /R292 384 0 R /R295 374 0 R /R294 375 0 R /R297 380 0 R /R25 32 0 R >> /ExtGState << /R36 41 0 R /R7 50 0 R /R23 42 0 R >> /Font << /F2 405 0 R /R288 406 0 R /R215 346 0 R /R187 409 0 R /R8 61 0 R /R76 122 0 R /R281 318 0 R /R286 413 0 R /R78 116 0 R /R10 64 0 R /F1 416 0 R /R20 81 0 R /R337 388 0 R /R335 392 0 R /R283 314 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R266 281 0 R /R185 398 0 R /R261 282 0 R /R263 283 0 R /R43 85 0 R /R316 352 0 R /R41 87 0 R /R257 286 0 R /R308 355 0 R /R329 356 0 R /R248 288 0 R /R303 357 0 R /R324 358 0 R /R301 359 0 R /R326 261 0 R /R321 277 0 R /R243 266 0 R /R240 271 0 R /R251 291 0 R /R306 361 0 R /R213 324 0 R /R311 353 0 R /R313 354 0 R /R237 284 0 R /R268 285 0 R /R319 274 0 R /R279 279 0 R /R259 287 0 R /R254 289 0 R /R271 290 0 R /R332 360 0 R /R276 292 0 R /R274 293 0 R /R299 362 0 R /R290 363 0 R /R293 364 0 R /R296 365 0 R /R24 28 0 R >> >> /Group 43 0 R /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 10 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 417 0 R /Resources << /XObject << /R361 418 0 R /R360 419 0 R /R363 420 0 R /R362 421 0 R /R365 422 0 R /R364 423 0 R /R359 424 0 R /R347 426 0 R /R351 427 0 R /R352 428 0 R /R353 429 0 R /R354 430 0 R /R355 431 0 R /R366 432 0 R /R348 437 0 R /R358 425 0 R >> /ExtGState << /R7 50 0 R /R23 42 0 R >> /Font << /F2 438 0 R /R288 406 0 R /R8 61 0 R /R349 439 0 R /R76 122 0 R /R286 413 0 R /R78 116 0 R /R356 433 0 R /R10 64 0 R /F1 443 0 R /R20 81 0 R /R82 134 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R43 85 0 R /R41 87 0 R >> >> /Group 194 0 R /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 11 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 444 0 R /Resources << /ColorSpace << /R43 85 0 R /R41 87 0 R >> /ExtGState << /R7 50 0 R >> /Font << /R10 64 0 R /R8 61 0 R /F2 445 0 R /F1 446 0 R /R18 78 0 R /R20 81 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 12 0 obj << /Type /Catalog /Pages 1 0 R >> endobj 13 0 obj << /Length 17726 >> stream q q q 0.10000 0 0 0.10000 0 0 cm /R7 gs 0 g q 10 0 0 10 0 0 cm BT /R8 14.34620 Tf 1 0 0 1 137.62500 675.06700 Tm [ (Unsuper) 10 (vised) -249.99000 (Semantic) -250 (P) 10.01330 (arsing) -250.01200 (of) -250.01200 (V) 37.01370 (ideo) -250.00300 (Collections) ] TJ /R10 11.95520 Tf -16.41210 -37.85820 Td [ (Ozan) -250.01500 (Sener) ] TJ /R12 7.97010 Tf 55.43520 4.33906 Td (1) Tj /R14 7.97010 Tf 4.23398 0 Td (\054) Tj /R12 7.97010 Tf 2.35312 0 Td (2) Tj /R10 11.95520 Tf 17.34880 -4.33906 Td [ (Amir) -249.99500 (R\056) -249.98700 (Zamir) ] TJ /R12 7.97010 Tf 71.39610 4.33906 Td (1) Tj /R10 11.95520 Tf 17.34880 -4.33906 Td [ (Silvio) -250 (Sa) 20.00670 (v) 24.98340 (arese) ] TJ /R12 7.97010 Tf 73.51210 4.33906 Td (1) Tj /R10 11.95520 Tf 17.35000 -4.33906 Td [ (Ashutosh) -249.99300 (Sax) 14.99740 (ena) ] TJ /R12 7.97010 Tf 82.50200 4.33906 Td (2) Tj /R14 7.97010 Tf 4.23398 0 Td (\054) Tj /R12 7.97010 Tf 2.35195 0 Td (3) Tj -323.94800 -27.89610 Td (1) Tj /R10 11.95520 Tf 4.73203 -4.33789 Td [ (Stanford) -250 (Uni) 24.99570 (v) 14.98510 (ersity) 64.98870 (\054) ] TJ /R12 7.97010 Tf 99.35900 4.33789 Td (2) Tj /R10 11.95520 Tf 4.73203 -4.33789 Td [ (Cornell) -249.99100 (Uni) 24.99570 (v) 14.98510 (ersity) 64.98870 (\054) ] TJ /R12 7.97010 Tf 94.05120 4.33789 Td (3) Tj /R10 11.95520 Tf 4.73203 -4.33789 Td [ (Brain) -249.98900 (of) -250.01400 (Things) -250.00600 (Inc\056) ] TJ /R16 8.96640 Tf -186.35800 -13.94800 Td (\173) Tj /R18 8.96640 Tf 4.60703 0 Td (ozansener\054zamir\054ssilvio\054asaxena) Tj /R16 8.96640 Tf 166.77500 0 Td (\175) Tj /R18 8.96640 Tf 4.60703 0 Td (\100cs\056stanford\056edu) Tj /R8 11.95520 Tf -196.57300 -41.04610 Td (Abstract) Tj /R20 9.96260 Tf -83.92770 -19.25270 Td [ (Human) -647.98900 (communication) -646.98000 (typically) -647.98400 (has) -647.01200 (an) -648 (underlying) ] TJ -11.95510 -11.71600 Td [ (structur) 37.01280 (e) 15.01280 (\056) -303.98600 (This) -230.98700 (is) -231.98700 (r) 37.01960 (e\337ected) -232 (in) -231.00800 (the) -231.99300 (fact) -231.00600 (that) -232.01500 (in) -231.98800 (many) -231.01300 (user) -231.98800 (g) 10.00320 (en\055) ] TJ 11.71600 TL T* [ (er) 15.01890 (ated) -324.99200 (videos\054) -344.00400 (a) -325 (starting) -324.99500 (point\054) -343.98900 (ending) 9.98118 (\054) -344.01900 (and) -324.99000 (certain) -324.98300 (objec\055) ] TJ T* [ (tive) -231.01500 (steps) -231 (between) -231.00300 (these) -231.01000 (two) -230.99500 (can) -231.01400 (be) -231.01900 (identi\336ed\056) -303.99300 (In) -231.01800 (this) -230.99600 (paper) 111.01800 (\054) ] TJ T* [ (we) -283.98300 (pr) 44.98390 (opose) -284.01400 (a) -283.98700 (method) -284.01000 (for) -284.01300 (par) 10.01430 (sing) -284.00800 (a) -283.98700 (video) -285.00400 (into) -284.00700 (suc) 14.98520 (h) -283.98700 (seman\055) ] TJ T* [ (tic) -318.99600 (steps) -318.00500 (in) -318.99300 (an) -318.01600 (unsupervised) -319.00700 (way) 54.98470 (\056) -515.99000 (The) -318.01100 (pr) 44.98510 (oposed) -318.98400 (method) -318.98600 (is) ] TJ T* [ (capable) -387.99500 (of) -388.00100 (pr) 44.98390 (o) 10.00320 (viding) -387.98900 (a) -386.98900 (\322semantic) -388.01400 (storyline\323) -388.00400 (of) -388.00100 (the) -388.00600 (video) ] TJ T* [ (composed) -322.00400 (of) -322.01200 (its) -322.00400 (objective) -322.01000 (steps\056) -526.00800 (W) 91.98590 (e) -321.99000 (accomplish) -321.99000 (this) -322 (using) ] TJ T* [ (both) -284.98900 (visual) -284.98500 (and) -284.99700 (langua) 9.99098 (g) 10.00320 (e) -285.01600 (cues) -285.00900 (in) -285 (a) -285.98700 (joint) -285.01900 (g) 10.00320 (ener) 15.01960 (ative) -285.00100 (model\056) ] TJ T* [ (The) -221.98700 (pr) 44.98390 (oposed) -223 (method) -221.98100 (can) -223.01500 (also) -222.01800 (pr) 44.98390 (o) 10.00320 (vide) -223.01800 (a) -221.99700 (te) 20 (xtual) -223.00200 (description) ] TJ T* [ (for) -326.00600 (eac) 15.01340 (h) -326.01900 (of) -324.99200 (the) -326.01700 (identi\336ed) -325.99100 (semantic) -326 (steps\056) -536.98600 (W) 91.98710 (e) -325.99000 (e) 15.01220 (valuate) -326.00400 (this) ] TJ T* [ (method) -288.99000 (on) -287.98100 (a) -289.00600 (lar) 36.99260 (g) 10.00320 (e) -287.99600 (number) -288.99100 (of) -288.01800 (comple) 20.00770 (x) -289.01600 (Y) 92 (ouT) 54.99570 (ube) -287.98600 (videos) -288.99600 (and) ] TJ T* [ (show) -264.98500 (r) 37.01960 (esults) -264.98800 (of) -265.00200 (unpr) 36.99510 (ecedented) -265.01500 (quality) -264.98300 (for) -266.01700 (this) -264.99000 (intricate) -265.01700 (and) ] TJ T* [ (impactful) -249.99400 (pr) 44.98390 (oblem\056) ] TJ /R8 11.95520 Tf 25.16480 TL T* [ (1\056) -249.99000 (Intr) 18.01460 (oduction) ] TJ /R10 9.96260 Tf 11.95510 -16.34020 Td [ (Human) -195.00500 (communication) -195.00800 (tak) 10.00570 (es) -194.00600 (man) 14.99010 (y) -194.98200 (forms\054) -205.98500 (including) -194.98500 (lan\055) ] TJ -11.95510 -11.71600 Td [ (guage) -356.01300 (and) -356.01400 (vision\056) -628.99400 (F) 14.99260 (or) -356.01800 (instance\054) -383.01500 (e) 15.01220 (xplaining) -356.01700 (\322ho) 24.98110 (w\055to\323) -356.99200 (per) 19.99420 (\055) ] TJ 11.71600 TL T* [ (form) -213.99300 (a) -214.98800 (certain) -214 (task) -215.00900 (can) -214.00700 (be) -214.00200 (communicated) -214.98100 (via) -213.99600 (language) -215.00800 (\050) ] TJ /R20 9.96260 Tf 219.52300 0 Td [ (e) 15.01220 (\056g) ] TJ /R10 9.96260 Tf 11.74570 0 Td (\056\054) Tj -231.26900 -11.71600 Td [ (Do\055It\055Y) 109.99000 (ourself) -328.01100 (books\051) -328 (as) -327.98300 (well) -328.01100 (as) -327.00300 (visual) -327.99700 (\050) ] TJ /R20 9.96260 Tf 166.24900 0 Td [ (e) 15.01220 (\056g) ] TJ /R10 9.96260 Tf 11.74610 0 Td [ (\056\054) -346.99600 (instructional) ] TJ -177.99500 -11.71600 Td [ (Y) 109.98500 (ouT) 44.98700 (ube) -499.98900 (videos\051) -500.01300 (information\056) -1058.99000 (Re) 14.98650 (g) 4.98446 (ardless) -499.98300 (of) -499.99300 (the) -499.98800 (form\054) ] TJ T* [ (such) -409.99800 (human\055generated) -411.00900 (communication) -409.99000 (is) -409.99800 (generally) -411.01700 (struc\055) ] TJ T* [ (tured) -205.01500 (and) -204 (has) -205.01800 (a) -204.99000 (clear) -204.01400 (be) 15.01830 (ginning\054) -214.00100 (end\054) -213.99500 (and) -204 (a) -204.99000 (set) -205.01500 (of) -204.00300 (steps) -205.00500 (in) -205.01200 (be\055) ] TJ T* [ (tween\056) -366.98900 (P) 14.99200 (arsing) -268.01800 (such) -269.00200 (communication) -268.99600 (into) -268.98900 (its) -268.99400 (semantic) -268.98900 (steps) ] TJ T* [ (is) -249.98400 (the) -249.99000 (k) 10.00320 (e) 15.01280 (y) -249.99300 (to) -249.98500 (understanding) -249.99300 (structured) -250.01600 (human) -250.01500 (acti) 24.98110 (vities\056) ] TJ 11.95510 -12.96720 Td [ (Language) -390.98300 (and) -392.00700 (vision) -390.99700 (pro) 14.98520 (vide) -392 (dif) 24.98480 (ferent\054) -426.00700 (b) 20.00160 (ut) -392 (correlating) ] TJ -11.95510 -11.71600 Td [ (and) -213.01800 (complementary) -212.98500 (information\056) -297.98300 (Challenge) -212.99400 (lies) -213.00600 (in) -213.99100 (that) -213.00900 (both) ] TJ T* [ (video) -427.98900 (frames) -428.00300 (and) -428 (language) -427.99200 (\050from) -429.01900 (subtit) 1.01699 (les) -429.01700 (generated) -428.00500 (via) ] TJ T* [ (ASR\051) -391.00600 (are) -391.00500 (only) -391.99100 (a) -390.99800 (noisy) 64.99770 (\054) -426.98500 (partial) -391.01800 (observ) 24.99090 (ation) -391.01300 (of) -390.99100 (the) -392.00600 (actions) ] TJ T* [ (being) -363.02000 (performed\056) -649.01800 (Ho) 24.98360 (we) 25.01420 (v) 14.98280 (er) 39.98600 (\054) -390.99100 (the) -362.99100 (complementary) -362.99800 (nature) -362.99300 (of) ] TJ T* [ (language) -233.00600 (and) -233.99400 (vision) -232.98400 (gi) 24.98850 (v) 14.98280 (es) -233.01900 (the) -233.99300 (opportunity) -233.00500 (to) -233.00800 (understand) -234.01200 (the) ] TJ T* [ (acti) 24.98170 (vities) -284.98300 (from) -285 (these) -285.00100 (partial) -284.99700 (observ) 24.98970 (ations\056) -415 (In) -285.00900 (this) -284.98700 (paper) 39.99090 (\054) -293.98800 (we) ] TJ T* [ (present) -346.98900 (a) -347.00500 (uni\336ed) -347.00800 (model\054) -371.00900 (incorporating) -347.00300 (both) -347.01800 (of) -346.99800 (the) -346.99400 (modal\055) ] TJ T* [ (ities\054) -375.99000 (in) -350.98800 (order) -349.98200 (to) -350.98800 (parse) -351.00100 (human) -351.01600 (acti) 24.98240 (vities) -349.99100 (into) -351.01500 (acti) 24.98110 (vity) -351.01500 (steps) ] TJ T* [ (with) -302.00600 (no) -300.99900 (form) -302.01700 (of) -301.98800 (supervision) -301.98400 (other) -300.99900 (than) -302.01600 (requiring) -302.00400 (videos) -302.00400 (to) ] TJ T* [ (be) -418.98800 (of) -420.00700 (the) -418.98100 (same) -419.99900 (cate) 15.01100 (gory) -419.01600 (\050) ] TJ /R20 9.96260 Tf 108.71600 0 Td [ (e) 15.01220 (\056g) ] TJ /R10 9.96260 Tf 11.74570 0 Td [ (\056\054) -461.99600 (videos) -419.00400 (retrie) 25.01540 (v) 14.98280 (ed) -420.00800 (by) -419.01800 (query) ] TJ -120.46200 -11.71600 Td [ (cooking) -250.00900 (e) 15.01280 (ggs\054) -250.01800 (changing) -250.01300 (tires\054) -249.98200 (etc\056\051\056) ] TJ 11.95510 -12.96680 Td [ (The) -285.00700 (k) 10.00320 (e) 15.01280 (y) -283.98700 (idea) -285.00800 (in) -284.01900 (our) -285.00400 (approach) -283.98800 (is) -284.99800 (the) -285.00400 (observ) 24.99090 (ation) -284.01100 (that) -284.99700 (the) ] TJ -11.95510 -11.71600 Td [ (lar) 17.99640 (ge) -372.01600 (collection) -373.00500 (of) -372.01500 (videos\054) -403.00400 (pertaining) -373.00600 (to) -372.00400 (the) -372.00900 (same) -372.98700 (acti) 24.98110 (vity) ] TJ ET Q q 3080 4590 2390 1060 re W n /R39 Do Q q 10 0 0 10 0 0 cm BT /R20 8.96640 Tf 1 0 0 1 308.86200 447.93200 Tm [ (F) 44.99760 (igur) 37 (e) -391.01900 (1\072) ] TJ /R10 8.96640 Tf 39.94180 0 Td [ (Gi) 25.00660 (v) 14.98650 (en) -391.01900 (a) -390.01700 (lar) 18.01160 (ge) -391.01900 (video) -390.00900 (collection) -391.02100 (\050frames) -390.01900 (and) -391.02100 (subtitles\051) ] TJ -39.94180 -10.95900 Td [ (of) -384 (an) -385.00700 (structured) -383.98600 (cate) 14.98380 (gory) -384.00200 (\050) ] TJ /R20 8.96640 Tf 99.29570 0 Td [ (e) 15.01370 (\056g) ] TJ /R10 8.96640 Tf 10.57230 0 Td [ (\056\054) -418.01000 (Ho) 24.99570 (w) -383.99700 (to) -384.97700 (cook) -384.00700 (an) -384.00700 (omelette\077\051\054) -418 (we) ] TJ -109.86800 -10.95900 Td [ (disco) 14.98110 (v) 14.98650 (er) -364.98000 (acti) 24.98750 (vity) -364.01600 (steps) -364.99100 (\050) ] TJ /R20 8.96640 Tf 87.51090 0 Td [ (e) 15.01100 (\056g) ] TJ /R10 8.96640 Tf 10.57110 0 Td [ (\056\054) -393.00400 (crack) -365.01800 (the) -365.00200 (e) 15.01370 (ggs\051\056) -653.99300 (W) 80.00790 (e) -364.00500 (also) -364.99700 (parse) -364.99700 (the) ] TJ -98.08200 -10.95900 Td [ (videos) -249.98400 (based) -250.01100 (on) -249.97800 (the) -249.98900 (disco) 14.98110 (v) 14.98650 (ered) -249.98900 (steps\056) ] TJ /R10 9.96260 Tf 25.63790 TL T* [ (class\054) -330.01100 (typically) -313.99200 (include) -314.00600 (only) -314.00200 (a) -313.99200 (fe) 25.00560 (w) -315.00100 (objecti) 25 (v) 14.98280 (e) -313.99200 (acti) 24.98110 (vity) -314.00200 (steps\054) ] TJ 11.71600 TL T* [ (and) -259.99100 (the) -259.00800 (v) 24.98110 (ariability) -260 (is) -259.98400 (the) -259.00800 (result) -260.01500 (of) -259.01600 (e) 15.01220 (xponentially) -259.98600 (man) 14.99260 (y) -259.99100 (w) 10.00320 (ays) ] TJ T* [ (of) -293.98800 (generating) -294.01200 (videos) -294.00500 (from) -294.02000 (acti) 24.98110 (vity) -294.00500 (steps) -294.99000 (through) -294 (subset) -294.00500 (se\055) ] TJ T* [ (lection) -240.99700 (and) -239.99400 (time) -240.99700 (ordering\056) -306.00800 (W) 79.98660 (e) -240.98400 (study) -240.98900 (this) -240.01400 (construction) -241.01900 (based) ] TJ T* [ (on) -310.99700 (the) -312.01900 (lar) 17.99700 (ge\055scale) -311.00900 (information) -311.98200 (a) 19.99180 (v) 24.98110 (ailable) -310.99200 (in) -312.01700 (Y) 109.98700 (ouT) 44.98760 (ube) -311.00200 (in) -312.01700 (the) ] TJ T* [ (form) -201.99600 (of) -202.98300 (instructional) -201.99600 (videos) -201.98100 (\050) ] TJ /R20 9.96260 Tf 114.87400 0 Td [ (e) 15.01220 (\056g) ] TJ /R10 9.96260 Tf 11.74610 0 Td [ (\056\054) -211.99900 (\322Making) -201.98100 (pancak) 9.99833 (e\323\054) -212.00900 (\322Ho) 25.02030 (w) ] TJ -126.62000 -11.71600 Td [ (to) -271.00100 (tie) -271.00400 (a) -270 (bo) 24.98600 (w) -271.01100 (tie\323\051\056) -372.99200 (W) 79.98660 (e) -271.01800 (adopt) -270.01400 (Instructional) -271.01400 (videos) -270.98900 (since) -271.00400 (the) 14.98770 (y) ] TJ T* [ (ha) 19.99670 (v) 14.98280 (e) -226.98700 (man) 14.99010 (y) -227.01600 (desirabl) 1 (e) -226.98700 (properties) -227.00100 (lik) 10.01790 (e) -226.98700 (the) -225.99200 (v) 20.00160 (olume) -226.99600 (of) -227.02100 (the) -227.01100 (infor) 20.00650 (\055) ] TJ T* [ (mation) -301.01600 (\050) ] TJ /R20 9.96260 Tf 33.98910 0 Td [ (e) 15.01220 (\056g) ] TJ /R10 9.96260 Tf 11.74570 0 Td [ (\056\054) -313.00200 (Y) 109.98400 (ouT) 44.98760 (ube) -301.00400 (has) -301.00400 (281\056000) -299.99400 (videos) -300.98400 (for) ] TJ /R20 9.96260 Tf 142.45200 0 Td [ (\323How) -301.01300 (to) -300.01900 (tie) ] TJ -188.18700 -11.71600 Td [ (a) -231.99600 (bow) -232.99800 (tie\323) ] TJ /R10 9.96260 Tf 41.71880 0 Td [ (\051) -232.00300 (and) -233.01500 (a) -232.00500 (well) -231.98800 (de\336ned) -232.02000 (notion) -232.98600 (of) -232 (acti) 24.98110 (vity) -232.01500 (step\056) -304.01300 (Ho) 24.98600 (w\055) ] TJ -41.71880 -11.71600 Td [ (e) 25.01050 (v) 14.98280 (er) 39.98600 (\054) -255.01400 (the) -253.00900 (proposed) -254.01200 (parsing) -254.01600 (method) -254.00700 (is) -253.00200 (applicable) -254.01600 (to) -253.98700 (an) 15.01710 (y) -253.99200 (type) ] TJ T* [ (of) -313.98400 (structured) -314.00600 (videos) -314.00200 (as) -315.00400 (long) -314.00200 (as) -313.98700 (the) 14.98770 (y) -313.98200 (are) -313.99700 (composed) -313.99700 (of) -315.00600 (a) -313.99200 (set) ] TJ T* [ (of) -249.99500 (objecti) 24.99830 (v) 14.98280 (e) -250.00200 (steps\056) ] TJ 11.95510 -13.61600 Td [ (The) -374.01100 (output) -373.98200 (of) -374.01400 (our) -374.00900 (method) -373.98700 (can) -373.98200 (be) -374.99600 (seen) -374.01600 (as) -374.01600 (the) -374.00600 (\322seman\055) ] TJ -11.95510 -11.71600 Td [ (tic) -259.98600 (storyline\323) -260.98300 (of) -259.99300 (a) -261.02000 (rather) -259.99800 (long) -260.01100 (and) -261.01000 (comple) 14.99750 (x) -259.99100 (video) -260.99600 (collection) ] TJ T* [ (\050see) -339.00700 (Fig\056) ] TJ ET Q /R41 cs 1 0 0 scn q 10 0 0 10 0 0 cm BT /R10 9.96260 Tf 1 0 0 1 347.44700 211.77700 Tm (1) Tj ET Q /R43 cs 0 scn q 10 0 0 10 0 0 cm BT /R10 9.96260 Tf 1 0 0 1 352.42900 211.77700 Tm [ (\051\056) -577.98200 (This) -339.01900 (storyline) -339.98700 (pro) 14.98280 (vides) -338.98300 (what) -338.99200 (particular) -340.01200 (steps) ] TJ -43.56640 -11.71600 Td [ (are) -308 (taking) -308.00300 (place) -307.98800 (in) -308.01500 (the) -309 (video) -308.00700 (collection\054) -321.99000 (when) -308.01700 (the) 14.98770 (y) -309.00200 (are) -307.99800 (oc\055) ] TJ 11.71600 TL T* [ (curring\054) -296.99500 (and) -287.98600 (what) -287.98100 (their) -288.01800 (meaning) -288.00600 (is) -288.01600 (\050) ] TJ /R20 9.96260 Tf 144.85600 0 Td (what\055when\055how) Tj /R10 9.96260 Tf 63.65120 0 Td [ (\051\056) -424.00800 (This) ] TJ -208.50700 -11.71600 Td [ (method) -335.99300 (also) -336.98800 (puts) -336 (videos) -335.99800 (performing) -337.01700 (the) -336.01300 (same) -336.01300 (o) 14.98280 (v) 14.98280 (erall) -337.00300 (task) ] TJ T* [ (in) -249.98500 (common) -250.00200 (ground) -250.01500 (and) -249.99300 (capture) -249.99700 (their) -249.98800 (high\055le) 24.99580 (v) 14.98280 (el) -249.99300 (relations\056) ] TJ 11.95510 -13.61680 Td [ (In) -323.98300 (the) -324.99700 (proposed) -324 (approach\054) -342.99700 (gi) 24.98600 (v) 14.98280 (en) -325.00500 (a) -323.99000 (collection) -323.99500 (of) -325.00500 (videos\054) ] TJ -11.95510 -11.71600 Td [ (we) -265.99700 (\336rst) -264.99000 (generate) -265.99700 (a) -265.02000 (set) -265.98500 (of) -265.01200 (language) -265.98000 (and) -265.01000 (visual) -266.01000 (atoms\056) -357.00900 (These) ] TJ T* [ (atoms) -302.00100 (are) -302.00100 (the) -301.98100 (result) -302.01100 (of) -301.98600 (relating) -302.00800 (object) -302.99400 (proposals) -301.99900 (from) -302.01800 (each) ] TJ T* [ (frame) -408.01300 (as) -408.00800 (well) -409.01800 (as) -408.00800 (detecting) -407.98600 (the) -408 (frequent) -408.00500 (w) 10.00320 (ords) -409.01500 (fr) 1.01454 (o) -1.01454 (m) -408 (sub\055) ] TJ T* [ (titles\056) -646.01700 (W) 79.98660 (e) -361.98400 (then) -362.00600 (emplo) 9.98363 (y) -362.01300 (a) -361.98400 (generati) 24.98600 (v) 14.98280 (e) ] TJ /R20 9.96260 Tf 151.82800 0 Td [ (beta) -362.00300 (pr) 44.98270 (ocess) -362.00800 (mixtur) 36.99870 (e) ] TJ -151.82800 -11.71600 Td (model) Tj /R10 9.96260 Tf 24.34800 0 Td [ (\054) -273.99100 (which) -269 (identi\336es) -269.00900 (the) -269.00400 (acti) 24.98110 (vity) -268.98900 (steps) -270.01400 (sha) 0.98513 (red) -270 (among) -268.98900 (the) ] TJ -24.34800 -11.71600 Td [ (videos) -329.01900 (of) -327.98200 (the) -328.99700 (same) -328.01400 (cate) 15.01220 (gory) -328.99400 (based) -328.01400 (on) -328.99400 (a) -327.98900 (representation) -329.01900 (us\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 (4480) 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 <<
/x24 18 0 R
>>
>>
/Length 28
/Group <<
/Type /Group
/S /Transparency
/CS /DeviceRGB
/I true
>>
/BBox [ 133 751 479 772 ]
/Type /XObject
/Subtype /Form
>>
stream
x+O4PH/VЯ02Qp
0
endstream
endobj
18 0 obj
<<
/Filter /FlateDecode
/Resources 19 0 R
/Length 53008
/Type /XObject
/BBox [ 133 751 479 772 ]
/Subtype /Form
>>
stream
xlK8,8?DKs9mavd
{f-8*2Y@H >חk}y}uf`v)_s}1z#*Gw_gX jow\o'1c|Z^G<
A̺X}ay?IT|y~L.[ {Ȟb\3-3]_'X\竵0{+_۾oY-wj+ B;)Aa=/
1~a(>}m_K' >*Q1:?OPnGsQ_`?mzN
?WBB!lU n'S 1Ipzχ1xF_09,}?e){:̹~Gu7g {0mD`pdS)>;ys`Pµx'O
+혝 oI~!p7̵&m=dC+>5ALV"I1"S*CKha#03L4|@p灟vMRQ==*/m"r>@uqo9ͻ;Vz 'ȳ;IBzWR~ihwa8PlT>]fZ8y52)r빃W{B7G/q)}M]elS(k=5^A] сڸc'w!Ll'M=Ȥ뎓z;#կN>\q6r>|8}}Lr$Mo5:>syls+Ǵ`C?/9{Y||jD u#i^\rJ8a|oY+Rxl<{[؏s迨nYaյ[?/^u;+ ,bhL}wrj:ɴdM8^q|m'aQ6;0k_ar|E8HɃpN('&r`V;Rs볩Nk\&W=DDG#5]~M{VyCSRH9ܜXˏt7߾i9ADگ
USڝ|{(hbIRZXg @6Hɹ0Dq;si\Tv<1?mUŏV7bF
p7Y9sƨ*.uz_LբřsB݃ɹ^'_Yae91f+n%A̶d
Ċc#q
CF=Mt'\&ξ)z;L-4:n}41j:mӶ_)\݇ |T@uAƓxAN6Ql;}Ɲ{~_y"g%]nCi']I~,xa?feoɌr0u?V=phldx.Y;~nZL603> \`ęܓ