%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 (Learning to Localize Sound Source in Visual Scenes) /Producer (PyPDF2) /Author (Arda Senocak\054 Tae\055Hyun Oh\054 Junsik Kim\054 Ming\055Hsuan Yang\054 In So Kweon) /Subject (2018 IEEE Conference on Computer Vision and Pattern Recognition) >> endobj 3 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 13 0 R /Resources << /XObject << /x8 14 0 R /R21 17 0 R /x6 18 0 R /R22 21 0 R /x12 24 0 R /x10 27 0 R >> /ExtGState << /s9 30 0 R /s11 33 0 R /a0 << /CA 1 /ca 1 >> /R7 36 0 R /s5 37 0 R /s7 40 0 R >> /Font << /F2 43 0 R /R8 44 0 R /R10 47 0 R /R12 51 0 R /F1 54 0 R /R16 55 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R19 22 0 R /R15 59 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 4 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 61 0 R /Resources << /XObject << /R46 62 0 R /R45 63 0 R /R44 64 0 R /R43 65 0 R /R42 66 0 R /R38 67 0 R /R39 68 0 R /R37 69 0 R >> /ExtGState << /R7 36 0 R >> /Font << /F2 70 0 R /R49 71 0 R /R8 44 0 R /R47 76 0 R /R40 79 0 R /R10 47 0 R /F1 83 0 R /R16 55 0 R /R35 84 0 R /R32 88 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R19 22 0 R /R15 59 0 R >> >> /Group 95 0 R /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 5 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 96 0 R /Resources << /ColorSpace << /R19 22 0 R /R15 59 0 R >> /ExtGState << /R7 36 0 R >> /Font << /F2 97 0 R /R58 98 0 R /R49 71 0 R /R8 44 0 R /R72 101 0 R /R70 106 0 R /R74 111 0 R /F1 114 0 R /R10 47 0 R /R64 115 0 R /R66 118 0 R /R60 123 0 R /R16 55 0 R /R62 127 0 R /R68 130 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 6 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 133 0 R /Resources << /ColorSpace << /R19 22 0 R /R15 59 0 R >> /ExtGState << /R7 36 0 R >> /Font << /R49 71 0 R /R72 101 0 R /R70 106 0 R /R74 111 0 R /F2 134 0 R /R88 135 0 R /R90 139 0 R /R82 142 0 R /R92 145 0 R /R94 150 0 R /R86 153 0 R /R84 158 0 R /R8 44 0 R /F1 161 0 R /R10 47 0 R /R64 115 0 R /R66 118 0 R /R60 123 0 R /R16 55 0 R /R62 127 0 R /R68 130 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 7 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 162 0 R /Resources << /XObject << /R109 163 0 R /R108 164 0 R /R102 165 0 R /R112 166 0 R /R113 167 0 R /R110 168 0 R /R111 169 0 R /R105 170 0 R /R104 171 0 R /R107 172 0 R /R106 173 0 R /R103 174 0 R >> /ExtGState << /R7 36 0 R >> /Font << /F2 175 0 R /R8 44 0 R /R49 71 0 R /R72 101 0 R /F1 176 0 R /R74 111 0 R /R10 47 0 R /R64 115 0 R /R66 118 0 R /R60 123 0 R /R16 55 0 R /R62 127 0 R /R88 135 0 R /R82 142 0 R /R92 145 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R19 22 0 R /R15 59 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 8 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 177 0 R /Resources << /XObject << /R121 178 0 R /R127 179 0 R /R131 180 0 R /R125 181 0 R /R124 182 0 R /R123 183 0 R /R122 184 0 R /R136 185 0 R /R137 186 0 R /R138 187 0 R /R139 188 0 R /R129 189 0 R /R128 190 0 R /R145 191 0 R /R144 192 0 R /R147 193 0 R /R146 194 0 R /R141 195 0 R /R140 196 0 R /R143 197 0 R /R126 198 0 R /R150 199 0 R /R142 200 0 R /R149 201 0 R /R148 202 0 R /R132 203 0 R /R130 204 0 R /R133 205 0 R /R134 206 0 R /R135 207 0 R >> /ExtGState << /R7 36 0 R >> /Font << /F2 208 0 R /R49 71 0 R /R8 44 0 R /F1 209 0 R /R74 111 0 R /R10 47 0 R /R64 115 0 R /R66 118 0 R /R60 123 0 R /R16 55 0 R /R62 127 0 R /R92 145 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R19 22 0 R /R15 59 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 9 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 210 0 R /Resources << /XObject << /R202 211 0 R /R211 212 0 R /R200 214 0 R /R201 215 0 R /R206 216 0 R /R207 217 0 R /R204 218 0 R /R205 219 0 R /R189 220 0 R /R188 221 0 R /R208 222 0 R /R209 223 0 R /R190 224 0 R /R191 225 0 R /R185 226 0 R /R184 227 0 R /R163 228 0 R /R162 229 0 R /R161 230 0 R /R160 231 0 R /R167 232 0 R /R187 233 0 R /R165 234 0 R /R164 235 0 R /R169 236 0 R /R168 237 0 R /R203 238 0 R /R192 239 0 R /R210 213 0 R /R213 240 0 R /R212 241 0 R /R196 242 0 R /R197 243 0 R /R194 244 0 R /R195 245 0 R /R198 246 0 R /R199 247 0 R /R193 248 0 R /R166 249 0 R /R183 250 0 R /R158 251 0 R /R159 252 0 R /R186 253 0 R /R182 254 0 R >> /ExtGState << /R7 36 0 R >> /Font << /R180 255 0 R /R178 258 0 R /R8 44 0 R /R49 71 0 R /R174 262 0 R /R176 267 0 R /R170 271 0 R /F2 274 0 R /R172 275 0 R /F1 279 0 R /R10 47 0 R /R64 115 0 R /R66 118 0 R /R60 123 0 R /R16 55 0 R /R88 135 0 R >> /ProcSet [ /Text /ImageC /ImageB /PDF /ImageI ] /ColorSpace << /R19 22 0 R /R15 59 0 R >> >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 10 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 280 0 R /Resources << /ColorSpace << /R19 22 0 R /R15 59 0 R >> /ExtGState << /R221 281 0 R /R222 282 0 R /R7 36 0 R >> /Font << /F2 283 0 R /R8 44 0 R /R49 71 0 R /R223 284 0 R /R10 47 0 R /F1 288 0 R /R16 55 0 R >> /ProcSet [ /ImageC /Text /PDF /ImageI /ImageB ] >> /MediaBox [ 0 0 612 792 ] /Annots [ ] >> endobj 11 0 obj << /Parent 1 0 R /Rotate 0 /Type /Page /Contents 289 0 R /Resources << /ColorSpace << /R19 22 0 R /R15 59 0 R >> /ExtGState << /R7 36 0 R >> /Font << /R10 47 0 R /F2 290 0 R /F1 291 0 R /R8 44 0 R /R16 55 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 16361 >> stream q q q 0.1 0 0 0.1 0 0 cm /R7 gs q 166 186 5788 7548 re W n 0 g q 10 0 0 10 0 0 cm BT /R8 14.3462 Tf 1 0 0 1 139.296 675.067 Tm [ (Lear) 14.9893 (ning) -249.994 (to) -250.012 (Localize) -249.988 (Sound) -250.002 (Sour) 18.0048 (ce) -250.013 (in) -249.993 (V) 37.0137 (isual) -250.005 (Scenes) ] TJ /R10 11.9552 Tf -46.6789 -37.8582 Td [ (Arda) -249.988 (Senocak) ] TJ ET Q 0 1 0 0 k q 10 0 0 10 0 0 cm BT /R12 7.9701 Tf 1 0 0 1 160.007 641.548 Tm (1) Tj ET Q /R15 cs 0 scn q 10 0 0 10 0 0 cm BT /R10 11.9552 Tf 1 0 0 1 176.695 637.209 Tm [ (T) 80.0137 (ae\055Hyun) -249.987 (Oh) ] TJ ET Q 0 1 0 0 k q 10 0 0 10 0 0 cm BT /R12 7.9701 Tf 1 0 0 1 241.803 641.548 Tm (2) Tj ET Q /R15 cs 0 scn q 10 0 0 10 0 0 cm BT /R10 11.9552 Tf 1 0 0 1 258.49 637.209 Tm [ (Junsik) -250.016 (Kim) ] TJ ET Q 0 1 0 0 k q 10 0 0 10 0 0 cm BT /R12 7.9701 Tf 1 0 0 1 313.292 641.548 Tm (1) Tj ET Q /R15 cs 0 scn q 10 0 0 10 0 0 cm BT /R10 11.9552 Tf 1 0 0 1 329.979 637.209 Tm [ (Ming\055Hsuan) -249.993 (Y) 99.9847 (ang) ] TJ ET Q 0 1 0 0 k q 10 0 0 10 0 0 cm BT /R12 7.9701 Tf 1 0 0 1 418.1 641.548 Tm (3) Tj ET Q /R15 cs 0 scn q 10 0 0 10 0 0 cm BT /R10 11.9552 Tf 1 0 0 1 434.788 637.209 Tm [ (In) -250.014 (So) -250.014 (Kweon) ] TJ ET Q 0 1 0 0 k q 10 0 0 10 0 0 cm BT /R12 7.9701 Tf 1 0 0 1 497.875 641.548 Tm (1) Tj ET Q /R15 cs 0 scn q 10 0 0 10 0 0 cm BT /R10 11.9552 Tf 1 0 0 1 220 623.262 Tm [ (Dept\056) -310.009 (EE\054) -249.989 (KAIST) 74.0088 (\054) -250.012 (South) -249.995 (K) 34.996 (orea) ] TJ ET Q 0 1 0 0 k q 10 0 0 10 0 0 cm BT /R12 7.9701 Tf 1 0 0 1 370.491 627.6 Tm (1) Tj ET Q /R15 cs 0 scn q 10 0 0 10 0 0 cm BT /R10 11.9552 Tf 1 0 0 1 237.963 609.314 Tm [ (MIT) -250.004 (CSAIL\054) -249.987 (MA\054) -250.006 (USA) ] TJ ET Q 0 1 0 0 k q 10 0 0 10 0 0 cm BT /R12 7.9701 Tf 1 0 0 1 352.529 613.652 Tm (2) Tj ET Q /R15 cs 0 scn q 10 0 0 10 0 0 cm BT /R10 11.9552 Tf 1 0 0 1 157.99 595.366 Tm [ (Dept\056) -310.009 (EECS\054) -250.012 (Uni) 24.9957 (v) 14.9851 (ersity) -249.989 (of) -250.014 (California\054) -249.993 (Merced\054) -249.991 (CA\054) -249.985 (USA) ] TJ ET Q 0 1 0 0 k q 10 0 0 10 0 0 cm BT /R12 7.9701 Tf 1 0 0 1 432.502 599.705 Tm (3) Tj ET Q /R15 cs 0 scn q 10 0 0 10 0 0 cm BT /R8 11.9552 Tf 1 0 0 1 145.995 554.32 Tm (Abstract) Tj /R16 9.9626 Tf -83.9277 -24.691 Td [ (V) 73.9913 (isual) -268.003 (e) 15.0128 (vents) -267.004 (ar) 36.9852 (e) -268 (usually) -268.003 (accompanied) -268.007 (by) -267.014 (sounds) -268.002 (in) -267.982 (our) ] TJ -11.9551 -11.9559 Td [ (daily) -326.984 (lives\056) -542.013 (W) 91.9859 (e) -327.009 (pose) -326.992 (the) -326.997 (question\072) -464.007 (Can) -326.982 (the) -328.016 (mac) 14.9803 (hine) -326.992 (learn) ] TJ 11.9551 TL T* [ (the) -412.982 (corr) 36.9871 (espondence) -413.996 (between) -413.012 (visual) -413.984 (s) 0.98635 (cene) -414.017 (and) -413.014 (the) -414.002 (sound\054) ] TJ T* [ (and) -263 (localize) -262.006 (the) -263.008 (sound) -262.989 (sour) 36.9963 (ce) -262.004 (only) -263.003 (by) -261.994 (observing) -262.983 (sound) -262.988 (and) ] TJ T* [ (visual) -274.987 (scene) -275.015 (pair) 9.9824 (s) -273.987 (lik) 10.0081 (e) -275.018 (human\077) -385.006 (In) -273.992 (this) -274.987 (paper) 111.018 (\054) -281.009 (we) -275.003 (pr) 44.9851 (opose) -274.996 (a) ] TJ T* [ (no) 10.0081 (vel) -261.016 (unsupervised) -259.997 (algorithm) -261.004 (to) -259.984 (addr) 36.9951 (ess) -261.008 (the) -261.008 (pr) 46.0046 (o) -1.01454 (bl) 0.99248 (em) -261.018 (of) -261.003 (lo\055) ] TJ 11.9563 TL T* [ (calizing) -364.999 (the) -364.01 (sound) -365.011 (sour) 36.9963 (ce) -363.987 (in) -364.986 (visual) -363.992 (scenes\056) -654.001 (A) -364.993 (two\055str) 37.0159 (eam) ] TJ 11.9547 TL T* [ (network) -427 (structur) 37.0122 (e) -426.012 (whic) 14.9987 (h) -426.983 (handles) -426.997 (eac) 15.0134 (h) -426.002 (modality) 55.0055 (\054) -471.017 (with) -426.995 (at\055) ] TJ T* [ (tention) -293.997 (mec) 15.011 (hanism) -295.018 (is) -294.016 (de) 15.0183 (veloped) -295.001 (for) -294.011 (sound) -294.002 (sour) 36.9963 (ce) -295.019 (localiza\055) ] TJ T* [ (tion\056) -419.981 (Mor) 36.9871 (eo) 9.99894 (ver) 110.999 (\054) -295.987 (although) -285.987 (our) -286.995 (network) -286.003 (is) -286.999 (formulated) -287.004 (within) ] TJ T* [ (the) -394.005 (unsupervised) -394.014 (learning) -394.016 (fr) 14.9914 (ame) 14.9816 (work\054) -429.985 (it) -393.998 (can) -394.988 (be) -394.013 (e) 19.9918 (xtended) ] TJ T* [ (to) -298.997 (a) -299.004 (uni\336ed) -298.986 (ar) 36.9852 (c) 15.0122 (hitectur) 37.0036 (e) -299.014 (with) -299.015 (a) -299.004 (simple) -300.004 (modi\336cation) -299 (for) -298.992 (the) ] TJ 11.9559 TL T* [ (supervised) -373.988 (and) -374.001 (semi\055supervised) -373 (learning) -374.018 (settings) -374.016 (as) -374.004 (well\056) ] TJ 11.9551 TL T* [ (Meanwhile) 9.99466 (\054) -424.985 (a) -390.008 (ne) 15.0183 (w) -390 (s) 0.98635 (ound) -389.994 (sour) 36.9963 (ce) -389.983 (dataset) -389.981 (is) -390.001 (de) 15.0171 (veloped) -390.004 (for) ] TJ T* [ (performance) -355.99 (e) 15.0122 (valuation\056) -626.987 (Our) -356.006 (empirical) -355.987 (e) 15.0122 (valuation) -356.017 (shows) ] TJ T* [ (that) -373.011 (the) -372.008 (unsupervised) -372.998 (method) -373.015 (e) 15.0122 (ventually) -371.992 (go) -372.987 (thr) 44.99 (ough) -373.016 (false) ] TJ T* [ (conclusion) -339.005 (in) -338.99 (some) -338.992 (cases\056) -576.991 (W) 91.9859 (e) -339.007 (also) -339.018 (show) -339.012 (that) -339.017 (e) 15.0122 (ven) -339.005 (with) -339.01 (a) ] TJ T* [ (fe) 14.981 (w) -372.983 (supervision\054) -404.004 (i\056e) 14.9828 (\056\054) -402.987 (semi\055supervised) -373 (setup\054) -403.989 (false) -373.014 (conclu\055) ] TJ 11.9559 TL T* [ (sion) -250.013 (is) -249.984 (able) -249.985 (to) -249.985 (be) -249.996 (corr) 36.9865 (ected) -249.999 (ef) 18 (fectively) 55.0178 (\056) ] TJ /R8 11.9552 Tf 29.7051 TL T* [ (1\056) -249.99 (Intr) 18.0146 (oduction) ] TJ /R10 9.9626 Tf 11.9551 -19.7102 Td [ (V) 59.9931 (isual) -197.005 (e) 25.0111 (v) 14.9828 (ents) -197.012 (are) -196.999 (typically) -196.99 (coherent) -196.981 (with) -197.984 (s) 0.98513 (ounds) -198.018 (and) -196.982 (the) 14.9852 (y) ] TJ -11.9551 -11.9551 Td [ (are) -296.001 (inte) 14.9932 (grated\056) -447.016 (When) -295.984 (we) -295.991 (see) -295.992 (that) -296.014 (a) -295.995 (car) -296.002 (is) -294.997 (mo) 14.995 (ving\054) -308.003 (we) -295.992 (hear) ] TJ 11.9547 TL T* [ (the) -229.993 (engine) -231.007 (sound) -230.014 (at) -229.998 (the) -231.013 (same) -229.991 (time\056) -304.003 (Sound) -230.006 (carries) -230.008 (rich) -231.015 (infor) 20.0065 (\055) ] TJ T* [ (mation) -206.012 (re) 15.0098 (g) 4.98446 (arding) -205.01 (the) -205.997 (spatial) -205 (and) -205.999 (temporal) -205 (cues) -206.002 (of) -204.983 (the) -205.997 (source) ] TJ 11.9563 TL T* [ (within) -295.995 (a) -297.014 (visual) -296.003 (scene\056) -448.984 (As) -297 (sho) 24.9922 (wn) -295.982 (in) -296.997 (the) -295.982 (bottom) -295.987 (e) 15.0122 (xample) -296.99 (of) ] TJ 11.9547 TL (Figure) ' ET Q /R19 cs 0 0 1 scn q 10 0 0 10 0 0 cm BT /R10 9.9626 Tf 1 0 0 1 79.0934 205.244 Tm (1) Tj ET Q /R15 cs 0 scn q 10 0 0 10 0 0 cm BT /R10 9.9626 Tf 1 0 0 1 84.0746 205.244 Tm [ (\054) -309.005 (the) -297.001 (engine) -298.015 (sound) -298.003 (suggests) -297.013 (where) -297.994 (the) -297.002 (source) -297.985 (may) ] TJ -33.9625 -11.9551 Td [ (be) -364.017 (in) -362.986 (the) -364.01 (ph) 4.98936 (ysical) -364 (w) 10 (orld) -362.985 (\133) ] TJ ET Q /R19 cs 0 0 1 scn q 10 0 0 10 0 0 cm BT /R10 9.9626 Tf 1 0 0 1 157.17 193.289 Tm (11) Tj ET Q /R15 cs 0 scn q 10 0 0 10 0 0 cm BT /R10 9.9626 Tf 1 0 0 1 167.133 193.289 Tm [ (\135\056) -650.991 (This) -363.996 (implies) -362.996 (that) -364.003 (sound) -363.991 (is) ] TJ -117.021 -11.9551 Td [ (not) -274.996 (only) -274.991 (complementary) -275.014 (to) -275 (the) -275.006 (visual) -274.987 (information\054) -281.985 (b) 20.0016 (ut) -275 (also) ] TJ 11.9551 TL T* [ (correlated) -249.997 (to) -249.985 (visual) -250.011 (e) 25.0105 (v) 14.9828 (ents\056) ] TJ 11.9551 -12.7367 Td [ (Humans) -718.01 (observ) 14.9914 (e) -716.983 (tremendous) -718.004 (number) -716.998 (of) -717.995 (combined) ] TJ /R10 7.9701 Tf 2.39063 -18.8551 Td [ (\052Ackno) 25.0126 (wledgment\072) -537.991 (W) 79.9803 (e) -364 (thank) -363.984 (to) -364.993 (W) 79.9803 (ei\055Sheng) -363.995 (Lai) -364.016 (for) -363.986 (helpful) -364.007 (discus\055) ] TJ -14.3457 -9.46523 Td [ (sions\054) -232.982 (and) -229.981 (to) -228.986 (the) -228.98 (annotators\056) -303.005 (A\056) -230.002 (Senocak\054) -233.01 (J\056) -228.99 (Kim) -229.993 (and) -229.002 (I\056S\056) -229.011 (Kweon) -230 (were) -228.984 (sup\055) ] TJ 9.46367 TL T* [ (ported) -221 (by) -222 (Instit) 1.00473 (ute) -222.02 (for) -221.02 (Information) -220.977 (\046) -221.978 (communicati) 1.02005 (ons) -222.003 (T) 70.0127 (echnology) -221.001 (Promo\055) ] TJ 9.46523 TL T* [ (tion) -193.017 (\050IITP\051) -194.024 (grant) -193.002 (funded) -194.016 (by) -192.984 (the) -193.985 (K) 35.0079 (orea) -192.99 (go) 15.0204 (v) 14.9837 (ernment) -193.981 (\050MSIT\051) -192.979 (\0502017\0550\05501780\051\056) ] TJ 9.46406 TL T* [ (The) -257.977 (part) -256.997 (of) -257.992 (this) -257.016 (w) 10.0121 (ork) -258.005 (w) 10.0121 (as) -257.992 (done) -257 (when) -258.002 (T) 73.9826 (\056\055H\056) -256.997 (Oh) -257.994 (w) 10.0136 (as) -257.992 (in) -257.021 (KAIST) 73.9948 (\054) -258.002 (and) -256.988 (he) -258.002 (w) 10.0136 (as) ] TJ 9.46484 TL T* [ (partly) -239.995 (funded) -241.018 (by) -239.985 (HBKU) -240.989 (QCRI) -240.016 (at) -239.995 (MIT) 73.9841 (\056) -240.995 (M\056\055H\056) -240.007 (Y) 99.9969 (ang) -239.981 (is) -240.982 (supported) -239.996 (in) -240.995 (part) -239.99 (by) ] TJ 9.46406 TL T* [ (NSF) -250.01 (CAREER) -249.976 (\050No\056) -309.984 (1149783\051\056) ] TJ ET Q Q q 166 186 5788 7548 re W n q 2362.47 0 0 1116.19 3088.62 4526.63 cm /R22 Do Q /R15 cs 0 scn q 10 0 0 10 0 0 cm BT /R10 8.9664 Tf 1 0 0 1 308.862 443.122 Tm [ (Figure) -306.009 (1\056) ] TJ /R8 8.9664 Tf 35.6211 0 Td [ (Wher) 18.017 (e) -305.979 (do) -305.984 (these) -305.99 (sounds) -304.98 (come) -306.014 (fr) 18.0034 (om\077) ] TJ /R10 8.9664 Tf 142.189 0 Td [ (W) 80.0079 (e) -305.976 (sho) 25.012 (w) -306.014 (an) -305.982 (e) 15.0137 (x\055) ] TJ -177.81 -10.959 Td [ (ample) -254.005 (of) -252.998 (interacti) 25.0147 (v) 14.9865 (e) -254.005 (sound) -252.982 (source) -253.992 (localization) -254.019 (by) -252.984 (the) -253.997 (proposed) -253.017 (al\055) ] TJ 10.959 TL T* [ (gorithm\056) -303.008 (In) -226.989 (this) -228 (paper) 39.9849 (\054) -232.008 (we) -228.005 (demonstrate) -228.016 (ho) 25.0066 (w) -226.987 (to) -228.011 (learn) -228.021 (to) -228.011 (localize) -226.981 (the) ] TJ T* [ (sources) -249.978 (\050objects\051) -249.978 (from) -249.997 (the) -249.989 (sound) -250.017 (signals\056) ] TJ /R10 9.9626 Tf 26.8449 TL T* [ (visual\055audio) -446.002 (e) 15.0122 (xamples) -446.996 (and) -446 (learn) -446.984 (the) -445.994 (correlation) -447.009 (between) ] TJ 11.9559 TL T* [ (them) -348.981 (throughout) -348.013 (their) -348.988 (whole) -348.986 (lif) 0.99003 (e) -349.005 (\133) ] TJ ET Q /R19 cs 0 0 1 scn q 10 0 0 10 0 0 cm BT /R10 9.9626 Tf 1 0 0 1 449.086 371.444 Tm (11) Tj ET Q /R15 cs 0 scn q 10 0 0 10 0 0 cm BT /R10 9.9626 Tf 1 0 0 1 459.048 371.444 Tm [ (\135) -349.005 (unconsciously) 64.9843 (\056) -606.012 (Be\055) ] TJ -150.186 -11.9551 Td [ (cause) -345.004 (of) -346.018 (the) -344.994 (correlation) -344.984 (between) -344.996 (the) -346.011 (sound) -345.016 (and) -344.996 (the) -346.011 (visual) ] TJ 11.9551 TL T* [ (e) 25.0105 (v) 14.9828 (ents\054) -312.989 (humans) -299.999 (can) -299.992 (understand) -300.984 (t) 0.98513 (he) -301.009 (object) -300.014 (or) -299.989 (the) -299.979 (e) 25.0105 (v) 14.9828 (ent) -300.999 (that) ] TJ T* [ (causes) -287.986 (sound) -288.003 (and) -287.986 (can) -287.994 (localize) -287.981 (the) -287.981 (sound) -288.986 (source) -287.986 (e) 25.0105 (v) 14.9828 (en) -287.991 (with\055) ] TJ T* [ (out) -354.982 (separate) -355.982 (education\056) -625.989 (Naturally) 64.9941 (\054) -382.01 (videos) -355.014 (and) -354.995 (their) -356.009 (corre\055) ] TJ T* [ (sponding) -227.016 (sounds) -228.009 (also) -227.006 (come) -227.991 (together) -227.001 (in) -227.991 (a) -226.987 (synchronized) -228.011 (w) 10.0032 (ay) 65.0088 (\056) ] TJ 11.9563 TL T* [ (Gi) 24.986 (v) 14.9828 (en) -233.02 (a) -232.986 (ple) 0.98758 (n) -1.01454 (t) 0.98758 (y) -233.015 (of) -233.017 (video) -233.003 (and) -233.015 (sound) -232.015 (clip) -233.005 (pairs\054) -235.985 (can) -232.986 (a) -232.986 (machine) ] TJ 11.9547 TL T* [ (model) -294.005 (learn) -293.99 (to) -294.017 (associate) -294.015 (the) -293.98 (sound) -294.005 (with) -294.005 (visual) -294.005 (scene) -293.99 (to) -294.02 (re\055) ] TJ T* [ (v) 14.9828 (eal) -255.999 (the) -255.009 (sound) -256.009 (source) -255.992 (location) -256.006 (without) -255.016 (an) 15.0171 (y) -255.992 (supervision) -255.992 (in) -255.987 (a) ] TJ T* [ (w) 10 (ay) -285.011 (similar) -285.999 (to) -285 (human) -284.989 (perception) -286.011 (to) -285.001 (localize) -285.001 (sound) -286.006 (sources) ] TJ T* [ (in) -249.985 (visual) -250.01 (scenes\077) ] TJ 11.9551 -13.716 Td [ (In) -255.994 (this) -254.992 (w) 10 (ork\054) -257.006 (we) -255.999 (analyze) -255.996 (whether) -255.006 (we) -255.996 (can) -254.982 (model) -256.011 (the) -255.987 (spa\055) ] TJ -11.9551 -11.9551 Td [ (tial) -342.004 (correspondence) -342.014 (between) -342.016 (visual) -341.997 (and) -342.016 (audio) -342.002 (information) ] TJ T* [ (by) -283.002 (le) 25.0179 (v) 14.9828 (eraging) -283.002 (the) -283.004 (correlation) -282.997 (between) -283.007 (the) -283.002 (modalities) -283.002 (based) ] TJ T* [ (on) -249.988 (simply) -249.995 (w) 10 (atching) -250.997 (and) -249.993 (listening) -249.993 (to) -249.988 (videos) -250.012 (in) -251.007 (unsupervised) ] TJ T* [ (w) 10 (ay) 65.0088 (\054) ] TJ /R16 9.9626 Tf 21.1719 0 Td (i\056e) Tj /R10 9.9626 Tf 9.68398 0 Td [ (\056\054) -283.987 (learning) -276.995 (based) -277.985 (sound) -276.988 (source) -277.008 (localization\056) -392.018 (W) 79.9866 (e) -277.998 (de\055) ] TJ -30.8559 -11.9559 Td [ (sign) -367.015 (our) -368.01 (m) 0.99248 (odel) -368.005 (by) -366.988 (using) -367.007 (tw) 10.0081 (o\055stream) -366.983 (netw) 10.0081 (ork) -368.012 (architecture) ] TJ T* [ (\050sound) -250.015 (and) -249.993 (visual) -250.01 (netw) 10.0081 (orks\051) -249.988 (where) -250.002 (each) -250.007 (netw) 10.0081 (ork) -249.993 (f) 9.99343 (acilitates) ] TJ T* [ (each) -305.015 (modality) -305.003 (and) -303.984 (localization) -305.013 (module) -305.018 (which) -304.993 (incorporates) ] TJ T* [ (the) -249.99 (attention) -250.002 (mechanism) -250.002 (as) -249.997 (illustrated) -249.997 (in) -249.988 (Figure) ] TJ ET Q /R19 cs 0 0 1 scn q 10 0 0 10 0 0 cm BT /R10 9.9626 Tf 1 0 0 1 500.632 142.536 Tm (2) Tj ET Q /R15 cs 0 scn q 10 0 0 10 0 0 cm BT /R10 9.9626 Tf 1 0 0 1 505.613 142.536 Tm (\056) Tj -184.796 -13.7148 Td [ (The) -251.992 (learning) -252.019 (task) -253.002 (for) -252 (sound) -252.012 (source) -251.992 (localization) -252.982 (from) -251.987 (lis\055) ] TJ -11.9551 -11.9563 Td [ (tening) -306.003 (is) -306.015 (challenging\054) -319.988 (especially) -306.008 (from) -306.018 (unlabeled) -306.008 (data\056) -479.013 (Un\055) ] TJ 11.9547 TL T* [ (constrained) -320 (videos) -320 (are) -319.998 (lik) 10.0179 (ely) -320.015 (to) -320.015 (contain) -320.986 (unrelat) 0.99493 (ed) -321.005 (audio) -320.005 (to) ] TJ T* [ (the) -376.989 (visual) -377.008 (contents) -377.008 (or) -376.993 (audio) -377.016 (source) -376.991 (that) -376.981 (is) -378 (of) 25.0154 (f) -377 (the) -376.986 (screen\054) ] TJ /R16 9.9626 Tf T* [ (e) 15.0122 (\056g) ] TJ /R10 9.9626 Tf 11.7461 0 Td [ (\056\054) -364.993 (narration\054) -366 (commenting\054) ] TJ /R16 9.9626 Tf 106.659 0 Td (etc) Tj /R10 9.9626 Tf 11.616 0 Td [ (\056) -342.016 (Another) -342.982 (challenge) -342.012 (arises) ] TJ ET Q 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 32 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 35 Tm (4358) 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
/BitsPerComponent 8
/Height 429
/Length 17870
/ColorSpace /DeviceGray
/DecodeParms <<
/Columns 908
/Predictor 15
>>
/Width 908
/Subtype /Image
>>
stream
xy@gnE4-T,O4MT@ʹ,γyji$Zi?+)2EXDA{,>?wYg>>μ @"{g (d J @F (d J @F (d J 1=^WZٲ\
3lTo֏,>XQ(/}Sa9 002EveQ?yM]n2 O8"1R2z~QAF8>ne{{