%PDF-1.3 1 0 obj << /Kids [ 3 0 R 4 0 R 5 0 R 6 0 R ] /Type /Pages /Count 4 >> endobj 2 0 obj << /Title (Visual to Sound\072 Generating Natural Sound for Videos in the Wild) /Producer (PyPDF2) /Author (Yipin Zhou\054 Zhaowen Wang\054 Chen Fang\054 Trung Bui\054 Tamara L\056 Berg) /Subject (2020 IEEE Conference on Computer Vision and Pattern Recognition Workshops) >> endobj 3 0 obj << /Parent 1 0 R /Rotate 0 /Contents 8 0 R /Resources << /XObject << /R29 9 0 R /x8 10 0 R /x10 13 0 R /x6 16 0 R /x12 19 0 R /R30 22 0 R >> /ExtGState << /s9 23 0 R /s11 26 0 R /a0 << /CA 1 /ca 1 >> /R21 29 0 R /s5 30 0 R /s7 33 0 R >> /Font << /F2 36 0 R /F1 37 0 R /R22 38 0 R /R24 41 0 R /R26 45 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /Group 49 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 50 0 R 51 0 R 52 0 R 53 0 R 54 0 R 55 0 R 56 0 R 57 0 R 58 0 R 59 0 R 60 0 R ] /Type /Page >> endobj 4 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 61 0 R /Resources << /ExtGState << /R21 29 0 R >> /Font << /F1 62 0 R /R72 63 0 R /R70 67 0 R /R24 41 0 R /F2 71 0 R /R64 72 0 R /R66 75 0 R /R60 78 0 R /R62 82 0 R /R22 38 0 R /R68 85 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 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 ] >> endobj 5 0 obj << /Parent 1 0 R /Rotate 0 /Contents 112 0 R /Resources << /XObject << /R83 113 0 R /R91 114 0 R /R92 115 0 R /R84 116 0 R >> /ExtGState << /R21 29 0 R >> /Font << /F2 117 0 R /R89 118 0 R /F1 122 0 R /R60 78 0 R /R22 38 0 R /R87 123 0 R /R24 41 0 R /R85 126 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] >> /Group 49 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 130 0 R 131 0 R 132 0 R 133 0 R 134 0 R ] /Type /Page >> endobj 6 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 135 0 R /Resources << /ExtGState << /R21 29 0 R >> /Font << /F2 136 0 R /F1 137 0 R /R60 78 0 R /R22 38 0 R /R24 41 0 R /R26 45 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ 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 7 0 obj << /Type /Catalog /Pages 1 0 R >> endobj 8 0 obj << /Length 20545 >> stream q q q 0.1 0 0 0.1 0 0 cm /R21 gs 0 g q 10 0 0 10 0 0 cm BT /R22 14.3462 Tf 1 0 0 1 93.9629 757.272 Tm [ (V) 37.0137 (isual) -250.004 (to) -250.012 (Sound\072) -309.99 (Generating) -250.01 (Natural) -249.993 (Sound) -250.002 (f) 24.9923 (or) -249.995 (V) 37.0137 (ideos) -249.988 (in) -249.991 (the) -250.008 (W) 18.0048 (ild) ] TJ /R24 11.9552 Tf 17.6391 -22.309 Td [ (Y) 54.9915 (ipin) -249.993 (Zhou) ] TJ /R24 7.9701 Tf 54.8031 4.33828 Td (1) Tj /R24 11.9552 Tf 4.48281 -4.33828 Td [ (\054) -250.012 (Zhao) 24.9926 (wen) -250.008 (W) 80 (ang) ] TJ /R24 7.9701 Tf 80.7453 4.33828 Td (2) Tj /R24 11.9552 Tf 4.48281 -4.33828 Td [ (\054) -250.012 (Chen) -249.995 (F) 15.0158 (ang) ] TJ /R24 7.9701 Tf 57.934 4.33828 Td (2) Tj /R24 11.9552 Tf 4.48398 -4.33828 Td [ (\054) -250.012 (T) 34.9898 (rung) -250.004 (Bui) ] TJ /R24 7.9701 Tf 55.041 4.33828 Td (2) Tj /R24 11.9552 Tf 4.4832 -4.33828 Td [ (\054) -250.012 (and) -249.987 (T) 80.0147 (amara) -249.997 (L\056) -250.018 (Ber) 18.0075 (g) ] TJ /R24 7.9701 Tf 101.08 4.33828 Td (1) Tj -332.363 -23.9082 Td (1) Tj /R24 11.9552 Tf 4.48281 -4.33789 Td [ (Uni) 24.9946 (v) 14.9862 (ersity) -249.989 (of) -250.014 (North) -250.016 (Carolina) -250 (at) -249.987 (Chapel) -249.993 (Hill\054) ] TJ /R24 7.9701 Tf 214.031 4.33789 Td (2) Tj /R24 11.9552 Tf 4.48398 -4.33789 Td [ (Adobe) -249.997 (Research) ] TJ /R22 11.9552 Tf -223.778 -26.7352 Td (Abstract) Tj /R26 9.9626 Tf -83.9277 -18.2688 Td [ (As) -273.003 (two) -274.007 (of) -273.001 (the) -273.006 <027665> -273.997 (tr) 14.9914 (aditional) -272.999 (human) -272.991 (senses) -274.018 (\050sight\054) -278.983 (hear) 20.0089 (\055) ] TJ -11.9551 -11.9551 Td [ (ing) 10.0155 (\054) -434.984 (taste) 10.0192 (\054) -434.984 (smell\054) -435.003 (and) -397.997 (touc) 14.9901 (h\051\054) -434.983 (vision) -397.986 (and) -399.017 (sound) -397.985 (ar) 36.9865 (e) -398.017 (basic) ] TJ 11.9551 TL T* [ (sour) 36.9963 (ces) -321.988 (thr) 44.9913 (ough) -322.005 (whic) 14.9987 (h) -323 (humans) -321.995 (under) 10.0155 (stand) -321.995 (the) -322.017 (world\056) -527.01 (Of\055) ] TJ T* [ (ten) -334.995 (corr) 36.9871 (elated) -334.987 (during) -333.99 (natur) 15.0061 (al) -334.991 (e) 15.0122 (vents\054) -356.007 (these) -334.993 (two) -335.018 (modalities) ] TJ T* [ (combine) -326.988 (to) -326.012 (jointly) -327.009 (af) 18.0092 (fect) -326.018 (human) -326.981 (per) 36.9816 (ception\056) -540.005 (In) -325.982 (this) -327.019 (paper) 111.018 (\054) ] TJ T* [ (we) -400.003 (pose) -400 (the) -400.004 (task) -399.998 (of) -400 (g) 10.0032 (ener) 15.0196 (ating) -399.982 (sound) -401.004 (given) -400.004 (visual) -399.985 (input\056) ] TJ 11.9559 TL T* [ <53706563690263616c6c79> 54.9957 (\054) -337 (we) -320.016 (apply) -320.007 (learning\055based) -318.995 (methods) -319.998 (to) -320.013 (g) 10.0032 (ener) 15.0196 (ate) ] TJ 11.9551 TL T* [ (r) 14.984 (aw) -292.992 (waveform) -293.999 (samples) -292.982 (given) -293.001 (input) -292.981 (video) -293.983 (fr) 14.9901 (ames\056) -439.013 (W) 91.9871 (e) -293.995 (e) 15.0122 (val\055) ] TJ T* [ (uate) -399.019 (our) -399.996 (models) -399.009 (on) -398.982 (a) -398.987 (dataset) -399.981 (of) -399.019 (videos) -399.017 (containing) -398.985 (a) -400.007 (va\055) ] TJ T* [ (riety) -350.989 (of) -352.007 (sounds) -351.008 (\050suc) 14.9816 (h) -352.015 (as) -350.989 (ambient) -351.011 (sounds) -351.988 (and) -350.986 (sounds) -351.988 (fr) 44.9851 (om) ] TJ T* [ (people\057animals\051\056) -395.993 (Our) -278.019 (e) 19.9918 (xperiments) -278.985 (show) -278.983 (that) -278.008 (the) -279.005 (g) 10.0032 (ener) 15.0196 (ated) ] TJ 11.9563 TL T* [ (sounds) -226.008 (ar) 36.9852 (e) -226.006 (fairly) -226.013 (r) 37.0196 (ealistic) -226.01 (and) -225.987 (have) -226 (go) -1.00964 (od) -225.992 (tempor) 15 (al) -225.989 (sync) 15.0196 (hr) 44.9851 (o\055) ] TJ 11.9551 TL T* [ (nization) -249.989 (with) -250.004 (the) -249.99 (visual) -250.011 (inputs\056) ] TJ /R22 11.9552 Tf 18.9379 TL T* [ (1\056) -249.99 (Intr) 18.0146 (oduction) ] TJ /R24 9.9626 Tf 11.9551 -16.1211 Td [ (The) -420.984 (visual) -421.002 (and) -420.982 (auditory) -421 (senses) -421.013 (are) -421.981 (ar) 17.9896 (guably) -421.006 (the) -420.981 (most) ] TJ -11.9551 -11.9559 Td [ (important) -412.988 (channels) -413.995 (through) -413 (which) -413.013 (humans) -413 (percei) 25.0105 (v) 14.9828 (e) -414.014 (their) ] TJ 11.9551 TL T* [ (surrounding) -303.006 (en) 39.9946 (vironments\054) -314.998 (and) -303.002 (the) 14.9852 (y) -303.003 (are) -302.001 (often) -302.998 (entertwined\056) ] TJ T* [ (From) -423 (life\055long) -421.982 (observ) 24.9897 (ations) -423.002 (of) -422.987 (the) -422 (natural) -423.015 (w) 10 (orld\054) -465.984 (people) ] TJ T* [ (are) -280.004 (able) -281.008 (to) -280.019 (learn) -281.012 (the) -279.985 (association) -279.996 (between) -281.005 (vision) -279.997 (and) -281.007 (sound\056) ] TJ T* [ (F) 14.992 (or) -372.015 (instance\054) -403.011 (when) -372.988 (seeing) -372.002 (a) -373.001 <03617368> -371.996 (of) -372.015 (lightning) -372.982 (in) -372.004 (the) -372.989 (sk) 14.9901 (y) 65.0137 (\054) ] TJ T* [ (one) -426.982 (might) -426.01 (co) 15.0183 (v) 14.9828 (er) -426.995 (their) -427.015 (ears) -426.013 (subconsciously) 64.9892 (\054) -471.017 (kno) 24.9909 (wing) -427.007 (that) ] TJ 11.9559 TL T* [ (the) -376.988 (crack) -375.987 (of) -376.994 (thunder) -377.016 (is) -376.982 (comi) 1 (ng) -376.986 (ne) 15.0183 (xt\056) -689.988 (Alternati) 25 (v) 14.9828 (ely) 65.0161 (\054) -408.988 (hear) 19.9893 (\055) ] TJ 11.9551 TL T* [ (ing) -229.984 (lea) 19.9954 (v) 14.9828 (es) -229.999 (rustling) -230 (in) -229.989 (the) -229.013 (wind) -229.981 (might) -230.003 (conjure) -229.991 (up) -229.991 (a) -230.006 (picture) -229.989 (of) ] TJ T* [ (a) -250.002 (peaceful) -250.001 (forest) -249.99 (scene\056) ] TJ 11.9551 -11.982 Td [ (In) -366.996 (this) -367.991 (paper) 39.9909 (\054) -396.01 (we) -368.018 (e) 15.0122 (xplore) -366.983 (whether) -366.989 (computational) -368.002 (mod\055) ] TJ -11.9551 -11.9547 Td [ (els) -384.985 (can) -384.998 (learn) -384.995 (the) -384.007 (relationship) -384.991 (between) -384.987 (visuals) -385.002 (and) -384.99 (sound\056) ] TJ T* [ (Models) -272.003 (of) -271.992 (this) -272.007 (relationship) -271.99 (could) -272.015 (be) -271.993 (fundamental) -272.004 (for) -271.996 (man) 14.9901 (y) ] TJ 11.9559 TL T* [ (applications) -350.998 (such) -350.008 (as) -351 (combining) -350.011 (videos) -351.015 (with) -351.017 (automatically) ] TJ 11.9551 TL T* [ (generated) -299.987 (ambient) -300 (sound) -300.983 (to) -300.016 (enhance) -299.992 (the) -299.982 (e) 15.0122 (xperience) -299.987 (of) -301.006 (im\055) ] TJ T* [ (mersion) -204.017 (in) -203.993 (virtual) -203.981 (reality\073) -219.98 (adding) -203.982 (sound) -204.017 (ef) 25.0081 (fects) -204.003 (to) -205.012 (videos) -203.981 (au\055) ] TJ T* [ (tomatically) -350.014 (to) -348.988 (reduce) -349.987 (tedious) -349.989 (manual) -349.019 (sound) -349.993 (editing) -349.988 (w) 10 (ork\073) ] TJ T* [ (Or) -356.02 (enabling) -357.013 (equal) -356.01 (accessibility) -356.018 (by) -356.009 (associating) -355.982 (sound) -357.012 (with) ] TJ T* [ (visual) -317.019 (informat) 1.00596 (ion) -316.989 (for) -317.007 (people) -316.012 (with) -316.984 (visual) -315.999 (impairments) -316.996 (\050al\055) ] TJ 11.9563 TL T* [ (lo) 24.9885 (wing) -308.007 (them) -308.007 (to) -306.995 (\223see\224) -307.999 (the) -308.02 (w) 10 (orld) -308.014 (through) -307.998 (sound\051\056) -482.993 (While) -308.012 (all) ] TJ 11.9547 TL T* [ (of) -208.983 (these) -209.994 (tasks) -209.004 (require) -210.002 (po) 24.986 (werful) -208.981 (high\055le) 24.9983 (v) 14.9828 (el) -210.002 (inference) -208.992 (and) -210 (rea\055) ] TJ T* [ (soning) -309.013 (ability) 65.0039 (\054) -325.002 (in) -308.995 (this) -308.982 (w) 10 (ork) -309.98 (we) -309.008 (tak) 10.0057 (e) -309.012 (a) -309.992 <02727374> -308.983 (step) -308.993 (to) 24.9885 (w) 10 (ard) -309.99 (this) ] TJ T* [ (goal\054) -229.005 (narro) 25.0148 (wing) -223.002 (do) 24.986 (wn) -223.994 (the) -223.994 (task) -223.008 (to) -223.99 (generating) -223.004 (audio) -223.985 (for) -224.004 (video) ] TJ T* [ (based) -285.984 (on) -285.982 (the) -285.984 (vie) 24.9836 (w) 10 (able) -285.988 (content\056) -417.985 <53706563690263616c6c79> 65.0137 (\054) -295.007 (we) -285.994 (train) -286.018 (mod\055) ] TJ 11.9559 TL T* [ (els) -201.016 (to) -201.993 (directly) -201.006 (predict) -201.993 (ra) 15.0098 (w) -200.984 (audio) -201.987 (signals) -200.993 (\050w) 9.99833 (a) 19.9918 (v) 14.9828 (eform) -201.998 (samples\051) ] TJ 11.9551 TL T* [ (from) -296.019 (input) -296 (videos\056) -448.016 (The) -295.985 (models) -295.998 (are) -296.982 (e) 15.0122 (xpected) -295.98 (to) -296.017 (learn) -295.99 (asso\055) ] TJ T* [ (ciations) -248.008 (between) -247.99 (generated) -249.016 (sound) -248.011 (and) -247.992 (visual) -248.01 (inputs) -248.003 (for) -249.02 (v) 24.9811 (ar) 19.9893 (\055) ] TJ T* [ (ious) -339.018 (scenes) -338.998 (and) -338.996 (object) -338.006 (interactions\056) -577.005 (Existing) -338.995 (w) 10 (orks) -338.99 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 267.217 164.713 Tm (9) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 272.198 164.713 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 278.066 164.713 Tm (2) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 283.048 164.713 Tm (\135) Tj -232.936 -11.9547 Td [ (handle) -330.01 (sound) -329.998 (generation) -330.985 (gi) 24.9885 (v) 14.9828 (en) -329.983 (input) -329.994 (of) -329.982 (videos\057images) -330.989 (un\055) ] TJ 11.9551 TL T* [ (der) -315.009 (e) 15.0128 (xperimental) -315.014 (settings) -315.006 (\050e\056g\056\054) -332.018 (to) -314.994 (generate) -315.009 (a) -315.991 (hitti) 1.01454 (ng) -316.016 (sound) ] TJ 11.9563 TL T* [ (or) -384.013 (where) -384.019 (the) -384.007 (input) -383.985 (videos) -383.99 (are) -383.987 (recorded) -384.019 (indoor) -383.99 (with) -383.992 <0278> 14.9975 (ed) ] TJ 11.9547 TL T* [ (background\051\056) -391.991 (In) -277.992 (our) -277.005 (w) 10 (ork\054) -283.982 (we) -277.994 (deal) -277.009 (with) -276.991 (generating) -278.015 (natural) ] TJ T* [ (sound) -250.01 (from) -249.987 (videos) -250.012 (collected) -249.983 (in) -249.985 (the) -249.99 (wild\056) ] TJ 11.9551 -11.982 Td [ (T) 79.9903 (o) -410.985 (enable) -410.987 (learning\054) -451.003 (we) -410.991 (introduce) -409.988 (a) -410.995 (dataset) -411.017 (that) -411.015 (is) -411.017 (de\055) ] TJ -11.9551 -11.9551 Td [ (ri) 25.0191 (v) 14.9828 (ed) -372.016 (from) -372.986 (AudioSet) -372.014 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 141.639 81 Tm (5) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 146.62 81 Tm [ (\135\056) -678.006 (The) -372.011 (dataset) -372.984 (includes) -371.989 (sounds) -373.004 (of) ] TJ ET Q q 1724.36 0 0 761.757 3407.56 6181.07 cm /R30 Do Q q 10 0 0 10 0 0 cm BT /R24 8.9664 Tf 1 0 0 1 308.862 597.186 Tm [ (Figure) -281.002 (1\056) -281.019 (Dataset) -281.999 (statistics\072) -372.019 (the) -281.008 (table) -282.021 (sho) 25.012 (ws) -281.002 (the) -281.008 (number) -280.981 (of) -281.013 (videos) ] TJ 10.959 TL T* [ (with) -192.987 (a) 20.021 (v) 14.9865 (eraged) -192.978 (length) -194 (for) -192.978 (each) -193.006 (cate) 14.9811 (gory) 64.9833 (\054) -204.997 (while) -193.006 (the) -193.006 (pie) -193.006 (chart) -192.995 (presents) ] TJ T* [ (the) -249.989 (distrib) 20.0129 (ution) -250.006 (of) -249.992 (video) -249.989 (lengths\056) ] TJ /R24 9.9626 Tf 22.0461 TL T* [ (humans\057animals) -292.983 (and) -293.005 (other) -291.981 (natural) -292.995 (sounds) -293.015 (spanning) -293.015 (10) -293 (cat\055) ] TJ 11.9551 TL T* [ (e) 15.0122 (gories) -253.982 (\050Baby) -253.982 (crying\054) -255 (Human) -254.016 (snoring\054) -254.982 (Dog\054) -255.982 (W) 79.9866 (ater) -254.002 <036f> 24.9958 (wing\054) ] TJ T* [ (Fire) 24.9909 (w) 10 (orks\054) -403.996 (Rail) -373.014 (transport\054) -403.982 (Printers\054) -404.021 (Drum\054) -403.991 (Helicopter) -373.996 (and) ] TJ 11.9563 TL T* [ (Chainsa) 15.0048 (w\051\056) ] TJ 11.9551 -12.732 Td [ (Our) -354.014 (model) -353.995 (learns) -354.012 (a) -353.005 (mapping) -353.985 (from) -354.01 (video) -354 (frames) -354.014 (to) -354.01 (au\055) ] TJ -11.9551 -11.9559 Td [ (dio) -356.982 (using) -357.009 (a) -357.004 (video) -356.982 (encoder) -357 (plus) -357.994 (sound) -357.014 (generator) -356.994 (structure\056) ] TJ 11.9551 TL T* [ (F) 14.9926 (or) -301.986 (sound) -302.001 (generation\054) -316.006 (we) -301.989 (use) -301.984 (a) -301.994 (hierarchical) -301.984 (recurrent) -303.013 (neu\055) ] TJ T* [ (ral) -321.005 (netw) 10.0081 (ork) -322.017 (proposed) -320.981 (by) -322.015 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 414.536 468.758 Tm (7) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 419.517 468.758 Tm [ (\135\056) -523.991 (W) 79.9866 (e) -321.99 (present) -320.995 (3) -321.981 (v) 24.9811 (ariants) -320.991 (to) -322.015 (en\055) ] TJ -110.655 -11.9551 Td [ (code) -339.98 (the) -340.995 (visual) -339.995 (information\054) -362.993 (which) -340.007 (can) -339.987 (be) -339.982 (combined) -341.012 (with) ] TJ 11.9551 TL T* [ (the) -390.006 (sound) -389.006 (generation) -389.996 (netw) 10.0081 (ork) -390.008 (to) -390.004 (form) -388.984 (a) -390.018 (complete) -389.989 (frame\055) ] TJ T* [ (w) 10 (ork) -386.007 (\050Sec\056) ] TJ ET Q 1 0 0 rg q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 357.111 432.893 Tm (3) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 362.092 432.893 Tm [ (\051\056) -717.998 (T) 79.9916 (o) -384.99 (e) 25.0105 (v) 24.9811 (aluate) -386.009 (the) -386.004 (proposed) -385.009 (models) -385.985 (and) -386.009 (the) ] TJ -53.2301 -11.9559 Td [ (generated) -386.012 (results\054) -421 (we) -386.016 (conduct) -385.994 (both) -385.99 (numerical) -387.014 (e) 25.0105 (v) 24.9811 (aluations) ] TJ 11.9551 TL T* [ (and) -390.008 (human) -389.011 (e) 15.0122 (xperiments) -390.016 (\050Sec\056) ] TJ ET Q 1 0 0 rg q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 434.63 408.982 Tm (4) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 439.611 408.982 Tm [ (\051\056) -729.016 (Please) -389.999 (see) -388.999 (our) -390.008 (supple\055) ] TJ -130.749 -11.9551 Td [ (mentary) -284.994 (video) -284.014 (to) -285 (see) -283.994 (and) -285.006 (hear) -285.016 (sound) -284.007 (generation) -284.992 (results\056) -413.985 (In) ] TJ 11.9547 TL T* [ (addition\054) -285.011 (recent) -277.017 (w) 10 (orks) -278.02 (\133) ] TJ ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 404.284 385.072 Tm (10) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 414.247 385.072 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 419.507 385.072 Tm (8) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 424.488 385.072 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 429.739 385.072 Tm (1) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 434.72 385.072 Tm (\054) Tj ET Q 0 1 0 rg q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 439.98 385.072 Tm (16) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 449.943 385.072 Tm [ (\135) -277.998 (learn) -277.013 (sound) -278.008 (source) -277.988 (lo\055) ] TJ -141.08 -11.9551 Td [ (calization) -236.017 (in) -235.007 (visual) -236.012 (scenes) -235.995 (based) -235.99 (on) -235.01 (the) -235.99 (concurrent) -235.99 (property) ] TJ 11.9551 TL T* [ (of) -245.015 (visual) -244.991 (and) -245.013 (sound\056) -307.998 (W) 79.9866 (e) -246.003 (also) -245.003 (apply) -244.998 (an) -245.018 (attention) -244.984 (mechanism) ] TJ 11.9559 TL T* [ (on) -227.011 (our) -227.014 (proposed) -226.996 (model) -226.996 (to) -227.011 (see) -226.987 (whether) -227.011 (it) -227.006 (can) -226.987 (learn) -226.982 (the) -227.011 (local\055) ] TJ 11.9551 TL T* [ (ization) -306.986 (as) -307.005 (a) -307.012 (by\055product) -306.981 (through) -307.017 (generation) -306.988 (task\056) -481.983 (Ho) 24.986 (we) 25.0154 (v) 14.9828 (er) 39.9835 (\054) ] TJ T* [ (we) -267.997 (observ) 14.9926 (e) -268 (that) -268.019 (the) -267.987 (at) 0.98268 (tention) -268 (maps) -268.014 (are) -268.004 (not) -268.014 (as) -267.995 (reasonable) -267.985 (as) ] TJ T* [ (the) -249.99 (carefully) -249.993 (designed) -250.012 (tasks) -250.017 (mentioned) -249.997 (abo) 14.9828 (v) 14.9828 (e\056) ] TJ 11.9551 -12.7328 Td [ (The) -235.995 (inno) 15 (v) 24.9811 (ations) -236.992 (introduced) -236.01 (by) -237.009 (our) -235.995 (paper) -236.98 (are\072) -303.013 (1\051) -236 (W) 79.9866 (e) -236.985 (pro\055) ] TJ -11.9551 -11.9551 Td [ (pose) -366.005 (a) -367.002 (ne) 25.0154 (w) -366.015 (problem) -367.01 (of) -366.015 (generating) -365.998 (sounds) -367.002 (from) -366.008 (videos) -367.012 (in) ] TJ T* [ (the) -279.005 (wild\073) -292 (2\051) -279.01 (W) 79.9866 (e) -279.017 (release) -278 (a) -279.017 (dataset) -278.998 (cont) 0.99493 (aining) -278.988 (28109) -278.988 (cleaned) ] TJ T* [ (videos) -270.989 (\05055) -271.986 (hours) -270.996 (in) -271.982 (total\051) -270.994 (spanning) -271.999 (10) -271.004 (object) -272.018 (cate) 15.0122 (gories\073) -282.007 (3\051) ] TJ 11.9563 TL T* [ (W) 79.9866 (e) -254.982 (e) 15.0122 (xplore) -255.002 (model) -254.992 (v) 24.9811 (ariants) -255.002 (for) -255.016 (the) -255.006 (generation) -254.997 (architectures\073) ] TJ 11.9551 TL T* [ (4\051) -333.981 (Numerical) -334.006 (and) -334.018 (human) -333.998 (e) 25.0105 (v) 24.9811 (aluations) -333.988 (are) -335.013 (pro) 14.9828 (vided) -334.003 (as) -333.983 (well) ] TJ T* [ (as) -249.995 (an) -249.997 (analysis) -250.01 (of) -249.995 (generated) -249.997 (results\056) ] TJ /R22 11.9552 Tf 20.5742 TL T* [ (2\056) -249.989 (V) 36.9891 (isually) -249.983 (Engaged) -250.002 (and) -249.997 (Gr) 17.9952 (ounded) -250.002 (A) 49.9913 (udioSet) ] TJ /R24 9.9626 Tf 11.9551 -19.707 Td [ (As) -239.989 (mentioned) -240.982 (by) -239.989 (Sec\056) ] TJ ET Q 1 0 0 rg q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 410.371 188.596 Tm (1) Tj ET Q 0 g q 10 0 0 10 0 0 cm BT /R24 9.9626 Tf 1 0 0 1 415.352 188.596 Tm [ (\054) -241.994 (we) -240 (collect) -241.004 (a) -240.004 (video) -239.98 (\050with) -240.999 (sound\051) ] TJ -106.49 -11.9551 Td [ (dataset) -338.99 (deri) 24.9811 (v) 14.9828 (ed) -339.002 (from) -338.992 (Au) -1.01209 (di) 0.99003 (oSet\056) -577.997 (Audioset) -339.007 (is) -338.988 (a) -339.987 (dataset) -338.988 (col\055) ] TJ 11.9547 TL T* [ (lected) -267.007 (for) -265.997 (audio) -266.997 (e) 25.0105 (v) 14.9828 (ent) -267.007 (recognition) -266 (b) 20.0016 (ut) -267.004 (not) -266.995 (ideal) -265.985 (for) -267.014 (our) -267.009 (task) ] TJ T* [ (because) -277.995 (man) 14.9901 (y) -277.008 (of) -277.99 (videos) -276.988 (and) -277.988 (audios) -276.988 (are) -278.003 (loosely) -276.983 (related\073) -292.02 (the) ] TJ T* [ (tar) 17.997 (get) -221.995 (sound) -222.014 (might) -222.005 (be) -220.983 (co) 15.0171 (v) 14.9828 (ered) -222.007 (by) -221.992 (other) -221.992 (sounds) -222.007 (\050lik) 10.013 (e) -222.007 (music\051\073) ] TJ T* [ (and) -279.007 (the) -279.985 (dataset) -279 (contains) -280.005 (some) -278.993 <6d69732d636c61737369026564> -280.017 (videos\056) -397.988 (All) -280.012 (of) ] TJ 11.9559 TL T* [ (these) -283.002 (sources) -283.002 (of) -283.009 (noise) -282.992 (tend) -282.997 (to) -283.002 (deter) -283.012 (the) -283.002 (models) -282.982 (from) -283.002 (learn\055) ] TJ 11.9551 TL T* [ (ing) -324.007 (the) -324.997 (correct) -323.992 (mapping) -323.99 (from) -324.995 (video) -324.005 (to) -324.015 (audio\056) -532.99 (T) 79.9916 (o) -325 (alle) 24.9811 (viate) ] TJ T* [ (these) -229.011 (issues\054) -234.01 (we) -228.984 (clean) -228.981 (a) -228.986 (subset) -230.016 (of) -229.021 (the) -229.011 (data\054) -233.015 (including) -229.996 (28\054109) ] TJ T* [ (videos) -266.01 (in) -265.002 (total) -266.012 (with) -266.012 (an) -265.015 (a) 19.9918 (v) 14.0026 (era) 1.00964 (g) -1.01454 (e) -265.02 (length) -266.01 (of) -265.995 (7) -265.01 (seconds\054) -270.014 (by) -265.985 (v) 14.9828 (er) 19.9869 (\055) ] TJ -13.741 -29.8879 Td (1) Tj ET Q Q Q q q 1 1 1 rg /a0 gs 48.406 786.422 515.188 -52.699 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 14.4 TL ET 1 1 1 rg n 270 47 72 14 re f* 0.5 0.5 0.5 rg BT /F2 9 Tf 10.8 TL ET BT 1 0 0 1 297 50 Tm (2500) Tj T* ET Q endstream endobj 9 0 obj << /Filter /FlateDecode /BitsPerComponent 8 /Height 842 /Length 3183 /ColorSpace /DeviceGray /DecodeParms << /Columns 1906 /Predictor 15 >> /Width 1906 /Subtype /Image >> stream xA 0 !ڥؽ{ r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( r \ H( pJ. endstream endobj 10 0 obj << /Filter /FlateDecode /Resources << /ExtGState << /a0 << /CA 1 /ca 1 >> >> /XObject << /x18 11 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 11 0 obj << /Filter /FlateDecode /Resources 12 0 R /Length 107 /Type /XObject /BBox [ 78 746 96 765 ] /Subtype /Form >> stream xe AC̬wʠ =p,?]%+H-
Jc "82w8VSnGW;"
endstream
endobj
12 0 obj
<<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
>>
endobj
13 0 obj
<<
/Filter /FlateDecode
/Resources <<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
/XObject <<
/x21 14 0 R
>>
>>
/Length 28
/Group <<
/Type /Group
/S /Transparency
/CS /DeviceRGB
/I true
>>
/BBox [ 96 752 110 775 ]
/Type /XObject
/Subtype /Form
>>
stream
x+O4PH/VЯ02Tp
0w
endstream
endobj
14 0 obj
<<
/Filter /FlateDecode
/Resources 15 0 R
/Length 118
/Type /XObject
/BBox [ 96 752 110 775 ]
/Subtype /Form
>>
stream
xe=QSp#lZq|ddLB:D[c¸>*FI&(\T=s&GuJѩIY\I}#
.!.
endstream
endobj
15 0 obj
<<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
>>
endobj
16 0 obj
<<
/Filter /FlateDecode
/Resources <<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
/XObject <<
/x15 17 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
17 0 obj
<<
/Filter /FlateDecode
/Resources 18 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
18 0 obj
<<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
>>
endobj
19 0 obj
<<
/Filter /FlateDecode
/Resources <<
/ExtGState <<
/a0 <<
/CA 1
/ca 1
>>
>>
/XObject <<
/x24 20 0 R
>>
>>
/Length 28
/Group <<
/Type /Group
/S /Transparency
/CS /DeviceRGB
/I true
>>
/BBox [ 112 751 500 772 ]
/Type /XObject
/Subtype /Form
>>
stream
x+O4PH/VЯ02Qp
0
endstream
endobj
20 0 obj
<<
/Filter /FlateDecode
/Resources 21 0 R
/Length 56490
/Type /XObject
/BBox [ 112 751 500 772 ]
/Subtype /Form
>>
stream
xtIJ5*|^~x?P9)<<ҥ+kh4}/?>]漿վ]Z Ǭ57W'b~?Kwk6:>rAニxeG1k-/LfߗA̺(egk5>
=KO1{Vux}Ol|_s؟|*>5=k|*a5x}On :97~1s7rX {|>IQQo6|'\#x_ڽC}M7(췀(N
سW#^Ň}|zblnB'\o=r;$s:z-
V3+έ昵 WD/nl ;q"3Pg휅\3C`k7760c9-Cȸ