%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 (Hierarchical Recurrent Neural Encoder for Video Representation With Application to Captioning) /Producer (PyPDF2) /Author (Pingbo Pan\054 Zhongwen Xu\054 Yi Yang\054 Fei Wu\054 Yueting Zhuang) /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 /x6 18 0 R /x12 21 0 R /x10 24 0 R >> /ExtGState << /s9 27 0 R /s11 30 0 R /a0 << /CA 1 /ca 1 >> /R25 33 0 R /s5 34 0 R /s7 37 0 R >> /Font << /F2 40 0 R /R28 41 0 R /F1 45 0 R /R36 46 0 R /R34 50 0 R /R32 53 0 R /R30 57 0 R /R26 61 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 64 0 R 65 0 R 66 0 R 67 0 R 68 0 R 69 0 R 70 0 R 71 0 R 72 0 R 73 0 R 74 0 R 75 0 R 76 0 R 77 0 R 78 0 R 79 0 R 80 0 R ] >> endobj 4 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 81 0 R /Resources << /ExtGState << /R25 33 0 R >> /Font << /R36 46 0 R /R28 41 0 R /F1 82 0 R /F2 83 0 R /R26 61 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 84 0 R 85 0 R 86 0 R 87 0 R 88 0 R 89 0 R 90 0 R 91 0 R 92 0 R 93 0 R 94 0 R 95 0 R 96 0 R 97 0 R 98 0 R 99 0 R 100 0 R 101 0 R 102 0 R 103 0 R 104 0 R 105 0 R 106 0 R 107 0 R 108 0 R 109 0 R 110 0 R 111 0 R 112 0 R 113 0 R 114 0 R 115 0 R 116 0 R 117 0 R ] >> endobj 5 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 118 0 R /Resources << /ExtGState << /R25 33 0 R >> /Font << /R28 41 0 R /R101 119 0 R /R103 123 0 R /R105 128 0 R /R107 133 0 R /F2 138 0 R /R99 139 0 R /F1 142 0 R /R36 46 0 R /R97 143 0 R /R93 147 0 R /R95 151 0 R /R26 61 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 154 0 R 155 0 R 156 0 R 157 0 R 158 0 R 159 0 R 160 0 R 161 0 R ] >> endobj 6 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 162 0 R /Resources << /ExtGState << /R25 33 0 R >> /Font << /R28 41 0 R /R101 119 0 R /R103 123 0 R /R105 128 0 R /R107 133 0 R /F2 163 0 R /R99 139 0 R /F1 164 0 R /R36 46 0 R /R93 147 0 R /R95 151 0 R /R97 143 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 165 0 R 166 0 R 167 0 R ] >> endobj 7 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 168 0 R /Resources << /ExtGState << /R25 33 0 R >> /Font << /R28 41 0 R /R101 119 0 R /R103 123 0 R /R105 128 0 R /R107 133 0 R /F2 169 0 R /R99 139 0 R /F1 170 0 R /R137 171 0 R /R36 46 0 R /R139 175 0 R /R97 143 0 R /R93 147 0 R /R95 151 0 R /R26 61 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 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 ] >> endobj 8 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 195 0 R /Resources << /ExtGState << /R25 33 0 R >> /Font << /R28 41 0 R /R103 123 0 R /R105 128 0 R /F2 196 0 R /R99 139 0 R /F1 197 0 R /R36 46 0 R /R93 147 0 R /R95 151 0 R /R26 61 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 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 211 0 R 212 0 R 213 0 R 214 0 R 215 0 R 216 0 R 217 0 R 218 0 R 219 0 R ] >> endobj 9 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 220 0 R /Resources << /ExtGState << /R25 33 0 R >> /Font << /R36 46 0 R /R28 41 0 R /F1 221 0 R /F2 222 0 R /R26 61 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 223 0 R 224 0 R 225 0 R 226 0 R 227 0 R 228 0 R 229 0 R 230 0 R 231 0 R 232 0 R 233 0 R 234 0 R 235 0 R 236 0 R 237 0 R 238 0 R 239 0 R 240 0 R 241 0 R 242 0 R 243 0 R 244 0 R 245 0 R 246 0 R 247 0 R 248 0 R 249 0 R 250 0 R 251 0 R 252 0 R 253 0 R 254 0 R 255 0 R 256 0 R 257 0 R 258 0 R 259 0 R 260 0 R 261 0 R ] >> endobj 10 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 262 0 R /Resources << /XObject << /R228 263 0 R /R229 264 0 R /R239 265 0 R /R238 266 0 R /R237 267 0 R /R236 268 0 R /R235 269 0 R /R234 270 0 R /R233 271 0 R /R232 272 0 R /R231 273 0 R /R230 274 0 R /R226 275 0 R /R227 276 0 R /R242 277 0 R /R243 278 0 R /R240 279 0 R /R241 280 0 R >> /ExtGState << /R25 33 0 R >> /Font << /R36 46 0 R /R28 41 0 R /F1 281 0 R /F2 282 0 R /R26 61 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 283 0 R 284 0 R 285 0 R 286 0 R 287 0 R 288 0 R 289 0 R 290 0 R 291 0 R 292 0 R ] >> endobj 11 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 293 0 R /Resources << /ExtGState << /R25 33 0 R >> /Font << /R36 46 0 R /R28 41 0 R /F1 294 0 R /F2 295 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 12 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 296 0 R /Resources << /ExtGState << /R25 33 0 R >> /Font << /R36 46 0 R /R28 41 0 R /F1 297 0 R /F2 298 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 13 0 obj << /Type /Catalog /Pages 1 0 R >> endobj 14 0 obj << /Length 18432 >> 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 14.34600 Tf 1 0 0 1 81.50000 675.07000 Tm [ (Hierar) 17.99830 (chical) -248.99700 (Recurr) 17.00270 (ent) -248.99600 (Neural) -249.99800 (Encoder) -249.99700 (f) 24.00730 (or) -249.00400 (V) 35.99650 (ideo) -248.99800 (Repr) 16.99080 (esentation) -248.98700 (with) ] TJ 136.58000 -17.93980 Td [ (A) 24.99610 (pplication) -248.99400 (to) -250 (Captioning) ] TJ /R28 11.95500 Tf -96.50000 -37.85000 Td [ (Pingbo) -249.01300 (P) 13.99390 (an) ] TJ /R30 7.97000 Tf 54.61990 4.33008 Td [ (x) -0.20064 ] TJ /R28 11.95500 Tf 16.21990 -4.33008 Td [ (Zhongwen) -248.98900 (Xu) ] TJ /R30 7.97000 Tf 68.73010 4.33008 Td [ (y) -0.19911 ] TJ /R28 11.95500 Tf 19.20000 -4.33008 Td [ (Y) 54 (i) -249.00900 (Y) 98.99390 (ang) ] TJ /R30 7.97000 Tf 38.98980 4.33008 Td [ (y) -0.19911 ] TJ /R28 11.95500 Tf 19.20000 -4.33008 Td [ (Fei) -249.99600 (W) 49.00990 (u) ] TJ /R30 7.97000 Tf 34.94020 4.33008 Td [ (x) -0.19911 ] TJ /R28 11.95500 Tf 19.20000 -4.33008 Td [ (Y) 110.00500 (ueting) -249.01300 (Zhuang) ] TJ /R30 7.97000 Tf 76.71020 4.33008 Td [ (x) -0.20217 ] TJ -313.19000 -14.05000 Td [ (x) -0.20064 ] TJ /R28 11.95500 Tf 4.26016 -4.34023 Td [ (Zhejiang) -249.99100 (Uni) 23.98930 (v) 15.00990 (ersity) ] TJ /R30 7.97000 Tf 113.42000 4.34023 Td [ (y) -0.19911 ] TJ /R28 11.95500 Tf 4.25977 -4.34023 Td [ (Uni) 23.98930 (v) 15.00990 (ersity) -248.98500 (of) -250 (T) 68.99450 (echnology) -248.99300 (Sydne) 13.98470 (y) ] TJ /R32 8.96600 Tf -130.83000 -13.95000 Td [ (f) -0.89994 ] TJ /R34 8.96600 Tf 4.60000 0 Td (lighnt001\054zhongwen\056s\056xu) Tj /R32 8.96600 Tf 123.74000 0 Td [ (g) -0.89858 ] TJ /R34 8.96600 Tf 4.61016 0 Td [ (\100gmail\056com) -982.01900 (yi\056yang\100uts\056edu\056au) ] TJ /R32 8.96600 Tf -59.88010 -13.95000 Td [ (f) -0.89858 ] TJ /R34 8.96600 Tf 4.60000 0 Td (wufei\054yzhuang) Tj /R32 8.96600 Tf 69.93980 0 Td [ (g) -0.89858 ] TJ /R34 8.96600 Tf 4.61016 0 Td (\100cs\056zju\056edu\056cn) Tj /R26 11.95500 Tf -153.54000 -41.03980 Td (Abstract) Tj /R36 9.96300 Tf -83.91990 -25.11020 Td [ (Recently) 54.99600 (\054) -286.00700 (deep) -279.00200 (learning) -278.98500 (appr) 44.00930 (oac) 15.00180 (h\054) -285.00100 (especially) -279.01600 (deep) -279.00100 (Con\055) ] TJ -11.95980 -11.95000 Td [ (volutional) -571.98300 (Neur) 14.01920 (al) -569.99600 (Networks) -572.00200 (\050Con) 38.99440 (vNets\051\054) -652.01000 (have) -572 (ac) 13.99710 (hie) 14.99440 (ved) ] TJ 11.95980 TL T* [ (o) 9.00365 (verwhelming) -308.98400 (accur) 14.01790 (acy) -310 (with) -311.01000 (fast) -310.00900 (pr) 43.99820 (ocessing) -310.00700 (speed) -311.01400 (for) -311.00700 (im\055) ] TJ 11.95000 TL T* [ (a) 9.00365 (g) 9.98384 (e) -385.98500 <636c6173736902636174696f6e2e> -723.01900 (Incorpor) 14.00810 (ating) -386.00500 (tempor) 14.01430 (al) -387.01300 (structur) 36.00230 (e) -387.00300 (with) ] TJ 11.96020 TL T* [ (deep) -332.99100 (Con) 38.98400 (vNets) -331.98200 (for) -333.00300 (video) -333.01200 (r) 36.01460 (epr) 36.99360 (esentation) -332.00200 (becomes) -333.00200 (a) -333.00500 (funda\055) ] TJ 11.95000 TL T* [ (mental) -359.00900 (pr) 43.99820 (oblem) -356.98900 (for) -358.99800 (video) -359.00700 (content) -359.00300 (analysis\056) -636.98100 (In) -358.98900 (this) -359.01900 (paper) 110.00900 (\054) ] TJ 11.96020 TL T* [ (we) -191.98500 (pr) 43.99820 (opose) -190.98700 (a) -192.99400 (ne) 13.99710 (w) -191.01800 (appr) 44.00930 (oac) 15.00180 (h\054) -203.01800 (namely) -191.99000 (Hier) 14.00450 (ar) 37.01930 (c) 14.99200 (hical) -192.01700 (Recurr) 36 (ent) ] TJ 11.95000 TL T* [ (Neur) 14.01920 (al) -207.01200 (Encoder) -208.98000 (\050HRNE\051\054) -207.99100 (to) -207.99200 (e) 18.98990 (xploit) -207.99500 (tempor) 14.01430 (al) -207.01100 (information) -208.99800 (of) ] TJ 11.95980 TL T* [ (videos\056) -428.01700 (Compar) 36.00480 (ed) -288.01900 (to) -288.99500 (r) 36.01460 (ecent) -288.99300 (video) -288.98200 (r) 36.01460 (epr) 36.99480 (esentation) -288.99200 (infer) 36.01210 (ence) ] TJ 11.95000 TL T* [ (appr) 44.00930 (oac) 15.00180 (hes\054) -269.00900 (this) -266.99900 (paper) -265.99100 (mak) 9.01039 (es) -265.98800 (the) -266.00600 (following) -266.98600 (thr) 36.01950 (ee) -265.99600 (contrib) 19.00590 (u\055) ] TJ 11.96020 TL T* [ (tions\056) -619.98600 (F) 44.01910 (ir) 9.99854 (st\054) -377.98700 (our) -354.01100 (HRNE) -353.01400 (is) -354.00400 (able) -352.98200 (to) -354.00100 (ef) 17.00870 <026369656e746c79> -352.01500 (e) 18.99120 (xploit) -352.98600 (video) ] TJ 11.95000 TL T* [ (tempor) 14.01490 (al) -239.98500 (structur) 36.00110 (e) -240.99500 (in) -241.00500 (a) -240.98400 (long) 8.99324 (er) -241.01300 (r) 14.01920 (ang) 9.99425 (e) -240.01500 (by) -241.98900 (r) 36.01460 (educing) -240 (the) -242.01100 (length) ] TJ 11.95980 TL T* [ (of) -227.00800 (input) -225.99800 (information) -226.99500 <036f77> 73.01670 (\054) -230.99900 (and) -227.01500 (compositing) -227 (multiple) -227 (consec\055) ] TJ 11.95000 TL T* [ (utive) -343.99700 (inputs) -345.01100 (at) -344.00400 (a) -345.00200 (higher) -343.99600 (le) 13.98970 (vel\056) -592.99600 (Second\054) -368.01900 (computation) -345.01200 (oper) 18.98380 (\055) ] TJ 11.96020 TL T* [ (ations) -294.98200 (ar) 36 (e) -294.00400 <7369676e690263616e746c79> -294.98500 (lessened) -295.00100 (while) -295 (attaining) -294.99500 (mor) 35.99250 (e) -294.98300 (non\055) ] TJ 11.95000 TL T* [ (linearity) 54.00410 (\056) -620.01600 (Thir) 36.01580 (d\054) -378.98200 (HRNE) -353.99400 (is) -352.98400 (able) -354.00100 (to) -354.00100 (unco) 8.98833 (ver) -353.01500 (tempor) 14.01430 (al) -352.98100 (tr) 13.99710 (an\055) ] TJ 11.96020 TL T* [ (sitions) -370.99000 (between) -372.00900 (fr) 13.99830 (ame) -370.98900 (c) 14.01060 (hunks) -370.00900 (with) -372.01700 (dif) 16.99890 (fer) 36.98740 (ent) -371.00400 (gr) 14.00450 (anularities\054) ] TJ /R28 9.96300 Tf 11.95000 TL (i\056e) ' /R36 9.96300 Tf 9.68984 0 Td [ (\056) -371.98900 (it) -372.98400 (can) -372.01600 (model) -372.98500 (the) -373.00400 (tempor) 14.01550 (al) -371.01800 (tr) 13.99830 (ansitions) -371.99900 (between) -372.98900 (fr) 13.99710 (ames) ] TJ -9.68984 -11.95980 Td [ (as) -312.00600 (well) -312.98200 (as) -312.98700 (the) -313.01600 (tr) 13.99830 (ansitions) -312.01100 (between) -313.00100 (se) 39.00790 (gments\056) -497.00500 (W) 90.99590 (e) -312 (apply) -313.00400 (the) ] TJ T* [ (ne) 13.99650 (w) -348.98600 (method) -349.00400 (to) -350.00200 (video) -349.98900 (captioning) -349.98300 (wher) 36.01460 (e) -349.01100 (tempor) 14.01430 (al) -348.98200 (informa\055) ] TJ 11.96020 TL T* [ (tion) -247 (plays) -246.98300 (a) -246.98300 (crucial) -248.00900 (r) 44.01290 (ole) 14.99440 (\056) -308.00200 (Experiments) -247.01100 (demonstr) 13.98240 (ate) -246.01000 (that) -248.02000 (our) ] TJ 11.95000 TL T* [ (method) -192.99700 (outperforms) -193.00700 (the) -193.98200 (state\055of\055the\055art) -193 (on) -193.00900 (video) -193.98200 (captioning) ] TJ 11.95980 TL T* [ (benc) 13.99590 (hmarks\056) ] TJ /R26 11.95500 Tf 39.44020 TL T* [ (1\056) -250.00700 (Intr) 17.00610 (oduction) ] TJ /R28 9.96300 Tf 11.95980 -20.11990 Td [ (Incorporating) -258.01500 (temporal) -260.01400 (information) -259.00800 (into) -258.99800 (video) -259.01500 (represen\055) ] TJ -11.95980 -11.96020 Td [ (tation) -387.99700 (has) -388.01600 (long) -387.98300 (been) -389.01900 (a) -387.98500 (fundamental) -388.00200 (problem) -387.99300 (in) -389.01300 (computer) ] TJ 11.95000 TL T* [ (vision\056) -546.01500 (Earlier) -328.99600 (w) 9.01039 (orks) -328.00700 (such) -328.00300 (as) -328.99400 (Dense) -329.00300 (T) 33.98320 (rajectories) -327.99100 (\133) 0.79395 (39) 0.28425 (\135) -328.98100 (and) ] TJ 11.95980 TL T* [ (impro) 14.00140 (v) 15.00240 (ed) -197.99900 (Dense) -198.99100 (T) 33.98190 (rajectories) -198.99800 (\050iDT\051) -198.99700 (\133) 0.55871 (40) 0.28548 (\135) -199.00800 (typically) -198.99800 (utilize) -199.99000 (op\055) ] TJ 11.95000 TL T* [ (tical) -282.99700 <036f> 24.01770 (w) -281.98800 (to) -282.99600 (e) 14.01060 (xtract) -282 (temporal) -282.98900 (information) -283.00300 (and) -283.01500 (hand\055crafted) ] TJ 11.96020 TL T* [ (features) -361.01500 (to) -362 (model) -361.01600 (appearances) -362.01100 (and) -362.01800 (motions\056) -643.99200 (W) 39.01150 (ith) -361.00100 (the) -361.98600 (re\055) ] TJ 11.95000 TL T* [ (cent) -247.01600 (success) -248.01400 (of) -247.01100 (deep) -247.98900 (Con) 38.98460 (v) 19.98120 (olutional) -245.99700 (Neural) -247.99300 (Netw) 8.99569 (orks) -247.00500 (\050Con\055) ] TJ 11.96020 TL T* [ (vNets\051) -269 (both) -268.98800 (in) -269 (ef) 24.01830 <026369656e63> 15.02020 (y) -269.01800 (and) -269.01800 (ef) 24.01830 <02636163> 15.01530 (y) 64.99140 (\054) -272.99000 (we) -268.99600 (ha) 19.01570 (v) 15.00180 (e) -268.00800 (witnessed) -269.99100 (a) ] TJ 11.95000 TL T* [ (ne) 23.99440 (w) -220.00100 (trend) -220.99900 (in) -221.00900 (le) 23.98770 (v) 15.00180 (eraging) -219.98000 (Con) 38.98460 (vNets) -220.00100 (to) -221.00900 (infer) -221.01300 (video) -222.00300 (representa\055) ] TJ 11.95980 TL T* [ (tion\056) -307.99400 (Xu) ] TJ /R36 9.96300 Tf 35.70000 0 Td [ (et) -245.99600 (al) ] TJ /R28 9.96300 Tf 17.40980 0 Td [ (\056) -244.99600 (\133) -1.08801 (41) 0.32469 (\135) -246.01800 (propose) -246 (to) -245.98500 (utilize) -247 (V) 110.00900 (ector) -245.00600 (of) -245.99300 (Locally) -246.98300 (Ag\055) ] TJ 205.64000 455.28000 Td [ (gre) 14.00690 (g) 5.00387 (ated) -239.99700 (Descriptors) -242.01400 (\050VLAD\051) -240.98400 (\133) -0.38227 (13) 0.32346 (\135) -241 (to) -241.98400 (aggre) 14.00690 (g) 5.00387 (ate) -239.98500 (frame) -242.01400 (le) 23.99010 (v) 15.00180 (el) ] TJ T* [ (Con) 38.98460 (vNet) -316.98300 (for) -318.00500 (video) -319.00300 (representation\054) -335 (which) -318.01700 (is) -318.99300 (unable) -318.00800 (to) -318.98800 (cap\055) ] TJ 11.95000 TL T* [ (ture) -198.01000 (temporal) -197.98800 (structure\056) -292.98600 (Simon) 13.98730 (yan) -198.01300 (and) -198.01300 (Zisserman) -197.99800 (\133) 0.55871 (26) -0.69593 (\135) -197.00800 (com\055) ] TJ 11.95980 TL T* [ (bine) -292.99600 (stack) 8.99324 (ed) -292.99800 (optical) -292.98300 <036f> 24.01710 (w) -291.98600 (frames) -293.98300 (and) -293.01300 (RGB) -293.98300 (streams) -293.01300 (to) -294.01300 (train) ] TJ 11.95000 TL T* [ (Con) 38.98460 (vNets) -208.98400 (for) -211.00800 (video) -210.00600 <636c6173736902636174696f6e2c> -219.01800 (which) -210 (achie) 24 (v) 15.00180 (es) -210 (compara\055) ] TJ 11.96020 TL T* [ (ble) -284.98300 (performance) -284.99200 (to) -284.99500 (iDT) -285 (in) -284.99500 (action) -285.00500 (recognition\056) -415.99700 (A) -285.00500 (limitation) ] TJ 11.95000 TL T* [ (of) -304.02000 (tw) 8.98833 (o\055stream) -304.01500 (Con) 38.98460 (vNets) -303.00300 (\133) 0.12743 (26) 0.28425 (\135) -304.00600 (and) -305.01000 (iDT) -304.01500 (\133) -1.12722 (40) 0.32346 (\135) -304.00600 (is) -304.01500 (that) -305 (both) -304 (al\055) ] TJ 11.96020 TL T* [ (gorithms) -208.00600 (require) -207.01800 (optical) -207.98100 <036f> 24.01950 (w) -206.98200 (as) -208.00100 (input\054) -215.99000 (which) -208.00100 (is) -207.99600 (e) 14.01180 (xpensi) 24.98500 (v) 15.00180 (e) -207.00100 (to) ] TJ 11.95000 TL T* [ (e) 14.01180 (xtract) -254.98600 (\050it) -256.99800 (tak) 8.99569 (es) -255.99100 (usually) -255.99600 (0\05606) -256.99600 (seconds) -257.01500 (to) -255.98100 (e) 14.01180 (xtract) -256.00600 (optical) -256.99100 <036f> 24.01950 (w) ] TJ 11.95980 TL T* [ (between) -315.00800 (a) -315.99800 (pair) -315.98600 (of) -316.01800 (frames) -316.01800 (\133) 0.08822 (26) 0.28425 (\135\051\054) -332.01500 (b) 19 (ut) -314.00900 (is) -316.01300 (only) -315.99800 (able) -315.98400 (to) -314.98900 (capture) ] TJ 11.95000 TL T* [ (temporal) -250.01700 (information) -249.99000 (in) -249.98300 (video) -249.99700 (clips) -250.00700 (of) -249.99200 (short) -250.00700 (duration\056) ] TJ 11.95980 -12.29020 Td [ (T) 80.01280 (o) -319.98800 (a) 18.99120 (v) 19.98120 (oid) -321.00200 (e) 14.01180 (xtracting) -320.00200 (optical) -321.01700 <036f> 24.01950 (w) 65.00120 (\054) -337.99400 (3D) -321.01200 (Con) 38.98210 (vNets) -320.99700 (are) -320.99200 (pro\055) ] TJ -11.95980 -11.95000 Td [ (posed) -315 (in) -316.00800 (\133) -0.06861 (35) -0.69593 (\135) -314.98400 (to) -316.00800 (generate) -315.01800 (a) -315.99800 (video) -315.98400 (representation\054) -331.98000 (with) -315.99800 (em\055) ] TJ 11.95980 TL T* [ (phasis) -233.00300 (on) -234.02000 (ef) 24.01950 <026369656e63> 15.01890 (y) -232.98600 (impro) 14.00200 (v) 15.00180 (ement\056) -304.00600 (This) -232.99600 (approach\054) -237.00500 (ho) 24.00480 (we) 24.98010 (v) 15.00180 (er) 40.01620 (\054) ] TJ 11.95000 TL T* [ (can) -343.01500 (only) -343.99200 (cope) -342.98800 (with) -343.01200 (16) -343.99700 (frames) -342.99300 (or) -344.01200 (so) -343.01700 (each) -343.98300 (time) -343.01200 (\133) 0.32346 (35) 0.28425 (\135\056) -590.01000 (V) 110.01200 (ery) ] TJ 11.96020 TL T* [ (recently) 64.00640 (\054) -280.98800 (Long) -275.99700 (Short\055T) 69.00290 (erm) -273.99200 (Memory) -276.01600 (\050LSTM\051) -276.01600 (\133) 0.16663 (11) 0.28425 (\135) -274.99200 (has) -276.00200 (been) ] TJ 11.95000 TL T* [ (applied) -345.00500 (to) -346.00200 (video) -346.01700 (analysis) -344.98700 (\133) -0.22544 (20) 0.28425 (\135\054) -368.99700 (inspired) -346.00700 (by) -345.99700 (the) -345.00700 (general) -346.00700 (re\055) ] TJ 11.95980 TL T* [ (current) -314.99600 (encoder) 18.98380 (\055decoder) -315.01300 (frame) 24.00970 (w) 9.99303 (ork) -313.99400 (\133) 0.12743 (31) 0.32346 (\135\056) -505.98800 (A) -315.97900 (plausible) -315.00800 (fea\055) ] TJ 11.95000 TL T* [ (ture) -382.01200 (of) -382.98400 (LSTM) -382.99700 (is) -382 (that) -382.98200 (LSTM) -382.99400 (is) -382 (capable) -383.01400 (of) -382.98400 (modeling) -382.98000 (data) ] TJ 11.96020 TL T* [ (sequences\056) -408.00400 (Ho) 24.01220 (we) 24.98260 (v) 15.00180 (er) 40.01620 (\054) -290.00600 (as) -281.98600 (this) -282.99500 (paper) -282.99000 (tries) -282.01500 (to) -282.99500 (cope) -283 (with\054) -291.00800 (there) ] TJ 11.95000 TL T* [ (are) -249.99000 (still) -250.01000 (a) -250.01200 (fe) 24.01950 (w) -249.01500 (challenges) -249.99500 (remain) -249.98300 (unaddressed\056) ] TJ 11.95980 -12.28980 Td [ (First\054) -494.99600 (a) -446.99100 (lar) 17.01850 (ge) -445.98600 (number) -446.98100 (of) -445.99100 (long\055range) -447.01000 (dependencies) -447.00500 (are) ] TJ -11.95980 -11.95000 Td [ (usually) -360.99400 (dif) 24.00970 <0263756c74> -360.00900 (to) -361.99900 (capture\056) -642.98900 (Ev) 14.01180 (en) -360.98400 (though) -360.99900 (LSTM) -360.99900 (can) -361.00900 (d) -0.99489 (eal) ] TJ 11.96020 TL T* [ (with) -284.00500 (long) -284.98500 (video) -285.01000 (clips) -284 (in) -284.99500 (principal\054) -293.01300 (it) -285.00500 (has) -284 (been) -285 (reported) -285.01400 (that) ] TJ 11.95000 TL T* [ (the) -365.00500 (f) 8.98833 (a) 20.01060 (v) 19.98120 (orable) -364.99500 (length) -365.01500 (of) -366.00800 (video) -366.01200 (clips) -365.00300 (to) -365.99800 (be) -364.98300 (feed) -366.01700 (into) -365.99300 (LSTM) ] TJ 11.96020 TL T* [ (f) 8.98833 (alls) -412 (in) -411.98800 (the) -412.99500 (range) -413.00300 (of) -413.01700 (30) -413.00300 (to) -413.00800 (80) -411.98300 (frames) -413.01700 (\133) -0.34307 (20) 0.32346 (\054) -412.44900 (37) -0.69593 (\135\056) -798.00600 (In) -411.99800 (order) ] TJ 11.95000 TL T* [ (to) -407.98900 (model) -409.00600 (longer) -409.01300 (video) -408.98400 (clips) -408.01400 (while) -408.98400 (attaining) -408.99400 (similar) -408.98900 (good) ] TJ 11.95980 TL T* [ (performance) -403.98700 (as) -403.01900 (in) -403.99000 (\133) -0.34307 (20) 0.28425 (\054) -404.41400 (37) 0.28425 (\135\054) -442.00200 (we) -403.01000 (propose) -404.00500 (to) -403.99000 (di) 24 (vide) -403.01000 (a) -403.98000 (long) ] TJ 11.95000 TL T* [ (video) -329.98100 (clip) -330.99300 (into) -330.98300 (a) -329.99500 (fe) 24.01950 (w) -330.01700 (short) -331.01000 (frame) -329.99500 (chunks\054) -350.99600 (feed) -331.00500 (the) -329.99000 (chunks) ] TJ 11.96020 TL T* [ (into) -274.01400 (LSTM\054) -273.00700 (and) -273.99700 (composite) -274.01200 (the) -274.00200 (LSTM) -273.99700 (outputs) -273.98200 (of) -273.98700 (the) -274.00200 (frame) ] TJ 11.95000 TL T* [ (chunks) -284.00200 (into) -284.99200 (one) -285.01400 (v) 13.98240 (ector) 39.99420 (\054) -292.98600 (which) -283.98500 (can) -284.98500 (then) -284.99500 (be) -285 (fed) -284.99000 (into) -284.01000 (another) ] TJ 11.96020 TL T* [ (LSTM) -216.99200 (at) -216.98200 (a) -217 (higher) -217.01400 (le) 23.98770 (v) 15.00180 (el) -216.00200 (to) -217.00900 (unco) 14.00690 (v) 15.00180 (er) -216.00900 (the) -216.99400 (temporal) -217.98400 (information) ] TJ 11.95000 TL T* [ (among) -236.01500 (the) -236.01200 (composited) -235.01500 (v) 13.98240 (ectors) -234.98600 (o) 13.98240 (v) 15.00180 (er) -234.98600 (a) -236.01500 (longer) -235.99000 (duration\056) -306 (Such) ] TJ 11.95980 TL T* [ (a) -355.99000 (hierarchical) -356.98700 (structure) -356.98500 <7369676e690263616e746c79> -355.98000 (reduces) -357.01900 (the) -357.00400 (length) -357.01400 (of) ] TJ 11.95000 TL T* [ (input) -221.01800 (information) -221.99600 <036f> 24.01710 (w) -220.98100 (b) 19 (ut) -221.00800 (is) -221.99300 (still) -220.99800 (capable) -221.98800 (of) -221.99800 (e) 14.01180 (xploiting) -221.00800 (tem\055) ] TJ 11.96020 TL T* [ (poral) -365.00800 (information) -366.00500 (o) 13.98240 (v) 15.00180 (er) -364.99800 (longer) -366.00300 (time) -365.00800 (afterw) 8.99814 (ards) -364.99300 (at) -366.00800 (a) -365.98800 (higher) ] TJ 11.95000 TL T* [ (le) 23.98770 (v) 15.00180 (el\056) ] TJ 11.95980 -12.28010 Td [ (Second\054) -318.99800 (additional) -306.99000 (non\055linearity) -306.00500 (has) -305.99500 (been) -306.01500 (demonstrated) ] TJ -11.95980 -11.95980 Td [ (helpful) -319.99500 (for) -320.98500 (impro) 14.00200 (ving) -319.01700 (model) -320.98300 (training) -320.01700 (for) -320.98700 (visual) -320.00700 (tasks) -321.01200 (such) ] 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 (1029) Tj T* ET Q endstream endobj 15 0 obj << /Filter /FlateDecode /Resources << /ExtGState << /a0 << /CA 1 /ca 1 >> >> /XObject << /x18 16 0 R >> >> /Length 28 /Group << /Type /Group /S /Transparency /CS /DeviceRGB /I true >> /BBox [ 78 746 96 765 ] /Type /XObject /Subtype /Form >> stream x+O4PH/VЯ0Pp 0 endstream endobj 16 0 obj << /Filter /FlateDecode /Resources 17 0 R /Length 107 /Type /XObject /BBox [ 78 746 96 765 ] /Subtype /Form >> stream xe AC̬wʠ =p,?]%+H-
Jc "82w8VSnGW;"
endstream
endobj
17 0 obj
<<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
>>
endobj
18 0 obj
<<
/Filter /FlateDecode
/Resources <<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
/XObject <<
/x15 19 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
19 0 obj
<<
/Filter /FlateDecode
/Resources 20 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
20 0 obj
<<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
>>
endobj
21 0 obj
<<
/Filter /FlateDecode
/Resources <<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
/XObject <<
/x24 22 0 R
>>
>>
/Length 28
/Group <<
/Type /Group
/S /Transparency
/CS /DeviceRGB
/I true
>>
/BBox [ 132 751 480 772 ]
/Type /XObject
/Subtype /Form
>>
stream
x+O4PH/VЯ02Qp
0
endstream
endobj
22 0 obj
<<
/Filter /FlateDecode
/Resources 23 0 R
/Length 53223
/Type /XObject
/BBox [ 132 751 480 772 ]
/Subtype /Form
>>
stream
xtI:6%Q㨈?7rA= u%6 ?Y(WbWo{B>9
x`Znϳ|8{3?0x*z ǃ|,@:w>`c|*ϻⳅKO3`g
:_|}}><.6`Z{{3]#<_o"~:ͺgk7/Ұ@|K yp ]03ʷCmş8˽Y?>(3!Bwqs.Z8,~~=rMT̩y+/*w: uBZ_`ߵp`%M?ɝ1ɳw=vDۉy&xb4Q>d@ sg~lA