id sid tid token lemma pos bracis-28377 1 1 dog dog NOUN bracis-28377 1 2 face face NOUN bracis-28377 1 3 recognition recognition NOUN bracis-28377 1 4 using use VERB bracis-28377 1 5 vision vision NOUN bracis-28377 1 6 transformer transformer NOUN bracis-28377 1 7 | | NOUN bracis-28377 1 8 springer springer NOUN bracis-28377 1 9 nature nature NOUN bracis-28377 1 10 link link PROPN bracis-28377 1 11 ( ( PUNCT bracis-28377 1 12 formerly formerly ADV bracis-28377 1 13 springerlink springerlink NOUN bracis-28377 1 14 ) ) PUNCT bracis-28377 1 15 skip skip VERB bracis-28377 1 16 to to ADP bracis-28377 1 17 main main ADJ bracis-28377 1 18 content content NOUN bracis-28377 1 19 advertisement advertisement NOUN bracis-28377 1 20 account account NOUN bracis-28377 1 21 menu menu NOUN bracis-28377 1 22 find find VERB bracis-28377 1 23 a a DET bracis-28377 1 24 journal journal NOUN bracis-28377 1 25 publish publish VERB bracis-28377 1 26 with with ADP bracis-28377 1 27 us we PRON bracis-28377 1 28 track track VERB bracis-28377 1 29 your your PRON bracis-28377 1 30 research research NOUN bracis-28377 1 31 search search NOUN bracis-28377 1 32 cart cart NOUN bracis-28377 1 33 home home NOUN bracis-28377 1 34 intelligent intelligent ADJ bracis-28377 1 35 systems system NOUN bracis-28377 1 36 conference conference NOUN bracis-28377 1 37 paper paper NOUN bracis-28377 1 38 dog dog NOUN bracis-28377 1 39 face face NOUN bracis-28377 1 40 recognition recognition NOUN bracis-28377 1 41 using use VERB bracis-28377 1 42 vision vision NOUN bracis-28377 1 43 transformer transformer NOUN bracis-28377 1 44 conference conference NOUN bracis-28377 1 45 paper paper NOUN bracis-28377 1 46 first first ADV bracis-28377 1 47 online online ADV bracis-28377 1 48 : : PUNCT bracis-28377 1 49 12 12 NUM bracis-28377 1 50 october october NOUN bracis-28377 1 51 2023 2023 NUM bracis-28377 1 52 pp pp ADP bracis-28377 1 53 33–47 33–47 NUM bracis-28377 1 54 cite cite VERB bracis-28377 1 55 this this DET bracis-28377 1 56 conference conference NOUN bracis-28377 1 57 paper paper NOUN bracis-28377 1 58 download download NOUN bracis-28377 1 59 book book NOUN bracis-28377 1 60 pdf pdf PROPN bracis-28377 1 61 download download NOUN bracis-28377 1 62 book book NOUN bracis-28377 1 63 epub epub PROPN bracis-28377 1 64 intelligent intelligent ADJ bracis-28377 1 65 systems system NOUN bracis-28377 1 66 ( ( PUNCT bracis-28377 1 67 bracis bracis NOUN bracis-28377 1 68 2023 2023 NUM bracis-28377 1 69 ) ) PUNCT bracis-28377 1 70 dog dog NOUN bracis-28377 1 71 face face NOUN bracis-28377 1 72 recognition recognition NOUN bracis-28377 1 73 using use VERB bracis-28377 1 74 vision vision NOUN bracis-28377 1 75 transformer transformer NOUN bracis-28377 1 76 download download NOUN bracis-28377 1 77 book book NOUN bracis-28377 1 78 pdf pdf PROPN bracis-28377 1 79 download download NOUN bracis-28377 1 80 book book PROPN bracis-28377 1 81 epub epub PROPN bracis-28377 1 82 victor victor PROPN bracis-28377 1 83 hugo hugo PROPN bracis-28377 1 84 braguim braguim VERB bracis-28377 1 85 canto canto PROPN bracis-28377 1 86     SPACE bracis-28377 1 87 orcid orcid NOUN bracis-28377 1 88 : : PUNCT bracis-28377 1 89 orcid.org/0000-0002-0511-44759 orcid.org/0000-0002-0511-44759 NOUN bracis-28377 1 90 , , PUNCT bracis-28377 1 91 joão joão PROPN bracis-28377 1 92 renato renato PROPN bracis-28377 1 93 ribeiro ribeiro PROPN bracis-28377 1 94 manesco manesco PROPN bracis-28377 1 95     SPACE bracis-28377 1 96 orcid orcid PROPN bracis-28377 1 97 : : PUNCT bracis-28377 1 98 orcid.org/0000-0002-1617-51429 orcid.org/0000-0002-1617-51429 ADJ bracis-28377 1 99 , , PUNCT bracis-28377 1 100 gustavo gustavo NOUN bracis-28377 1 101 botelho botelho PROPN bracis-28377 1 102 de de ADP bracis-28377 1 103 souza souza PROPN bracis-28377 1 104     SPACE bracis-28377 1 105 orcid orcid PROPN bracis-28377 1 106 : : PUNCT bracis-28377 1 107 orcid.org/0000-0002-4441-910810 orcid.org/0000-0002-4441-910810 PROPN bracis-28377 1 108 & & CCONJ bracis-28377 1 109 … … PUNCT bracis-28377 1 110 aparecido aparecido PROPN bracis-28377 1 111 nilceu nilceu PROPN bracis-28377 1 112 marana marana PROPN bracis-28377 1 113     SPACE bracis-28377 1 114 orcid orcid PROPN bracis-28377 1 115 : : PUNCT bracis-28377 1 116 orcid.org/0000-0003-4861-70619 orcid.org/0000-0003-4861-70619 ADV bracis-28377 1 117     SPACE bracis-28377 1 118 show show VERB bracis-28377 1 119 authors author NOUN bracis-28377 1 120 part part NOUN bracis-28377 1 121 of of ADP bracis-28377 1 122 the the DET bracis-28377 1 123 book book NOUN bracis-28377 1 124 series series NOUN bracis-28377 1 125 : : PUNCT bracis-28377 1 126 lecture lecture NOUN bracis-28377 1 127 notes note NOUN bracis-28377 1 128 in in ADP bracis-28377 1 129 computer computer NOUN bracis-28377 1 130 science science NOUN bracis-28377 1 131 ( ( PUNCT bracis-28377 1 132 ( ( PUNCT bracis-28377 1 133 lnai lnai ADJ bracis-28377 1 134 , , PUNCT bracis-28377 1 135 volume volume NOUN bracis-28377 1 136 14196 14196 NUM bracis-28377 1 137 ) ) PUNCT bracis-28377 1 138 ) ) PUNCT bracis-28377 1 139 included include VERB bracis-28377 1 140 in in ADP bracis-28377 1 141 the the DET bracis-28377 1 142 following follow VERB bracis-28377 1 143 conference conference NOUN bracis-28377 1 144 series series NOUN bracis-28377 1 145 : : PUNCT bracis-28377 1 146 brazilian brazilian ADJ bracis-28377 1 147 conference conference NOUN bracis-28377 1 148 on on ADP bracis-28377 1 149 intelligent intelligent ADJ bracis-28377 1 150 systems system NOUN bracis-28377 1 151 597 597 NUM bracis-28377 1 152 accesses access NOUN bracis-28377 1 153 abstract abstract VERB bracis-28377 1 154 the the DET bracis-28377 1 155 demand demand NOUN bracis-28377 1 156 for for ADP bracis-28377 1 157 effective effective ADJ bracis-28377 1 158 , , PUNCT bracis-28377 1 159 efficient efficient ADJ bracis-28377 1 160 and and CCONJ bracis-28377 1 161 safe safe ADJ bracis-28377 1 162 methods method NOUN bracis-28377 1 163 for for ADP bracis-28377 1 164 animal animal NOUN bracis-28377 1 165 identification identification NOUN bracis-28377 1 166 has have AUX bracis-28377 1 167 been be AUX bracis-28377 1 168 increasing increase VERB bracis-28377 1 169 significantly significantly ADV bracis-28377 1 170 , , PUNCT bracis-28377 1 171 due due ADP bracis-28377 1 172 to to ADP bracis-28377 1 173 the the DET bracis-28377 1 174 need need NOUN bracis-28377 1 175 for for ADP bracis-28377 1 176 traceability traceability NOUN bracis-28377 1 177 , , PUNCT bracis-28377 1 178 management management NOUN bracis-28377 1 179 , , PUNCT bracis-28377 1 180 and and CCONJ bracis-28377 1 181 control control NOUN bracis-28377 1 182 of of ADP bracis-28377 1 183 this this DET bracis-28377 1 184 population population NOUN bracis-28377 1 185 , , PUNCT bracis-28377 1 186 which which PRON bracis-28377 1 187 grows grow VERB bracis-28377 1 188 at at ADP bracis-28377 1 189 higher high ADJ bracis-28377 1 190 rates rate NOUN bracis-28377 1 191 than than ADP bracis-28377 1 192 the the DET bracis-28377 1 193 human human ADJ bracis-28377 1 194 population population NOUN bracis-28377 1 195 , , PUNCT bracis-28377 1 196 particularly particularly ADV bracis-28377 1 197 pets pet NOUN bracis-28377 1 198 . . PUNCT bracis-28377 2 1 motivated motivate VERB bracis-28377 2 2 by by ADP bracis-28377 2 3 the the DET bracis-28377 2 4 efficacy efficacy NOUN bracis-28377 2 5 of of ADP bracis-28377 2 6 modern modern ADJ bracis-28377 2 7 human human ADJ bracis-28377 2 8 identification identification NOUN bracis-28377 2 9 methods method NOUN bracis-28377 2 10 based base VERB bracis-28377 2 11 on on ADP bracis-28377 2 12 face face NOUN bracis-28377 2 13 biometrics biometric NOUN bracis-28377 2 14 features feature NOUN bracis-28377 2 15 , , PUNCT bracis-28377 2 16 in in ADP bracis-28377 2 17 this this DET bracis-28377 2 18 paper paper NOUN bracis-28377 2 19 , , PUNCT bracis-28377 2 20 we we PRON bracis-28377 2 21 propose propose VERB bracis-28377 2 22 a a DET bracis-28377 2 23 dog dog NOUN bracis-28377 2 24 face face NOUN bracis-28377 2 25 recognition recognition NOUN bracis-28377 2 26 method method NOUN bracis-28377 2 27 based base VERB bracis-28377 2 28 on on ADP bracis-28377 2 29 vision vision NOUN bracis-28377 2 30 transformers transformer NOUN bracis-28377 2 31 , , PUNCT bracis-28377 2 32 a a DET bracis-28377 2 33 deep deep ADJ bracis-28377 2 34 learning learning NOUN bracis-28377 2 35 approach approach NOUN bracis-28377 2 36 that that PRON bracis-28377 2 37 decomposes decompose VERB bracis-28377 2 38 the the DET bracis-28377 2 39 input input NOUN bracis-28377 2 40 image image NOUN bracis-28377 2 41 into into ADP bracis-28377 2 42 a a DET bracis-28377 2 43 sequence sequence NOUN bracis-28377 2 44 of of ADP bracis-28377 2 45 patches patch NOUN bracis-28377 2 46 and and CCONJ bracis-28377 2 47 applies apply VERB bracis-28377 2 48 self self NOUN bracis-28377 2 49 - - PUNCT bracis-28377 2 50 attention attention NOUN bracis-28377 2 51 to to ADP bracis-28377 2 52 these these DET bracis-28377 2 53 patches patch NOUN bracis-28377 2 54 to to PART bracis-28377 2 55 capture capture VERB bracis-28377 2 56 spatial spatial ADJ bracis-28377 2 57 relationships relationship NOUN bracis-28377 2 58 between between ADP bracis-28377 2 59 them they PRON bracis-28377 2 60 . . PUNCT bracis-28377 3 1 results result NOUN bracis-28377 3 2 obtained obtain VERB bracis-28377 3 3 on on ADP bracis-28377 3 4 dogfacenet dogfacenet NOUN bracis-28377 3 5 , , PUNCT bracis-28377 3 6 a a DET bracis-28377 3 7 public public ADJ bracis-28377 3 8 database database NOUN bracis-28377 3 9 of of ADP bracis-28377 3 10 dog dog NOUN bracis-28377 3 11 face face NOUN bracis-28377 3 12 images image NOUN bracis-28377 3 13 , , PUNCT bracis-28377 3 14 show show VERB bracis-28377 3 15 that that SCONJ bracis-28377 3 16 the the DET bracis-28377 3 17 proposed propose VERB bracis-28377 3 18 method method NOUN bracis-28377 3 19 , , PUNCT bracis-28377 3 20 which which PRON bracis-28377 3 21 uses use VERB bracis-28377 3 22 the the DET bracis-28377 3 23 efficientformer efficientformer NOUN bracis-28377 3 24 - - PUNCT bracis-28377 3 25 l1 l1 PROPN bracis-28377 3 26 architecture architecture NOUN bracis-28377 3 27 , , PUNCT bracis-28377 3 28 outperforms outperform VERB bracis-28377 3 29 the the DET bracis-28377 3 30 state state NOUN bracis-28377 3 31 - - PUNCT bracis-28377 3 32 of of ADP bracis-28377 3 33 - - PUNCT bracis-28377 3 34 the the DET bracis-28377 3 35 - - PUNCT bracis-28377 3 36 art art NOUN bracis-28377 3 37 method method NOUN bracis-28377 3 38 proposed propose VERB bracis-28377 3 39 previously previously ADV bracis-28377 3 40 in in ADP bracis-28377 3 41 literature literature NOUN bracis-28377 3 42 based base VERB bracis-28377 3 43 on on ADP bracis-28377 3 44 resnet resnet NOUN bracis-28377 3 45 , , PUNCT bracis-28377 3 46 a a DET bracis-28377 3 47 deep deep ADJ bracis-28377 3 48 convolutional convolutional ADJ bracis-28377 3 49 neural neural ADJ bracis-28377 3 50 network network NOUN bracis-28377 3 51 . . PUNCT bracis-28377 4 1 download download PROPN bracis-28377 4 2 conference conference PROPN bracis-28377 4 3 paper paper NOUN bracis-28377 4 4 pdf pdf NOUN bracis-28377 4 5 similar similar ADJ bracis-28377 4 6 content content NOUN bracis-28377 4 7 being be AUX bracis-28377 4 8 viewed view VERB bracis-28377 4 9 by by ADP bracis-28377 4 10 others other NOUN bracis-28377 4 11 dogface dogface NOUN bracis-28377 4 12 detection detection NOUN bracis-28377 4 13 and and CCONJ bracis-28377 4 14 localization localization NOUN bracis-28377 4 15 of of ADP bracis-28377 4 16 dogface dogface NOUN bracis-28377 4 17 ’s ’s PART bracis-28377 4 18 landmarks landmark NOUN bracis-28377 4 19 chapter chapter NOUN bracis-28377 4 20 © © PROPN bracis-28377 4 21 2019 2019 NUM bracis-28377 4 22 a a DET bracis-28377 4 23 deep deep ADJ bracis-28377 4 24 learning learning NOUN bracis-28377 4 25 approach approach NOUN bracis-28377 4 26 for for ADP bracis-28377 4 27 dog dog NOUN bracis-28377 4 28 face face NOUN bracis-28377 4 29 verification verification NOUN bracis-28377 4 30 and and CCONJ bracis-28377 4 31 recognition recognition NOUN bracis-28377 4 32 chapter chapter NOUN bracis-28377 4 33 © © PROPN bracis-28377 4 34 2019 2019 NUM bracis-28377 4 35 deep deep ADJ bracis-28377 4 36 learning learning NOUN bracis-28377 4 37 techniques technique NOUN bracis-28377 4 38 for for ADP bracis-28377 4 39 dogs dog NOUN bracis-28377 4 40 image image NOUN bracis-28377 4 41 dataset dataset VERB bracis-28377 4 42 analysis analysis NOUN bracis-28377 4 43 chapter chapter NOUN bracis-28377 4 44 © © PROPN bracis-28377 4 45 2025 2025 NUM bracis-28377 4 46 explore explore VERB bracis-28377 4 47 related relate VERB bracis-28377 4 48 subjects subject NOUN bracis-28377 4 49 discover discover VERB bracis-28377 4 50 the the DET bracis-28377 4 51 latest late ADJ bracis-28377 4 52 articles article NOUN bracis-28377 4 53 , , PUNCT bracis-28377 4 54 books book NOUN bracis-28377 4 55 and and CCONJ bracis-28377 4 56 news news NOUN bracis-28377 4 57 in in ADP bracis-28377 4 58 related related ADJ bracis-28377 4 59 subjects subject NOUN bracis-28377 4 60 , , PUNCT bracis-28377 4 61 suggested suggest VERB bracis-28377 4 62 using use VERB bracis-28377 4 63 machine machine NOUN bracis-28377 4 64 learning learning NOUN bracis-28377 4 65 . . PUNCT bracis-28377 5 1 automated automate VERB bracis-28377 5 2 pattern pattern NOUN bracis-28377 5 3 recognition recognition NOUN bracis-28377 5 4 biometrics biometric NOUN bracis-28377 5 5 computer computer NOUN bracis-28377 5 6 vision vision PROPN bracis-28377 5 7 object object VERB bracis-28377 5 8 recognition recognition NOUN bracis-28377 5 9 pattern pattern NOUN bracis-28377 5 10 vision vision NOUN bracis-28377 5 11 computer computer NOUN bracis-28377 5 12 imaging imaging NOUN bracis-28377 5 13 , , PUNCT bracis-28377 5 14 vision vision NOUN bracis-28377 5 15 , , PUNCT bracis-28377 5 16 pattern pattern NOUN bracis-28377 5 17 recognition recognition NOUN bracis-28377 5 18 and and CCONJ bracis-28377 5 19 graphics graphic NOUN bracis-28377 5 20 1 1 NUM bracis-28377 5 21 introduction introduction NOUN bracis-28377 5 22 according accord VERB bracis-28377 5 23 to to ADP bracis-28377 5 24 pet pet PROPN bracis-28377 5 25 brazil brazil PROPN bracis-28377 5 26 institute institute NOUN bracis-28377 5 27 [ [ X bracis-28377 5 28 9 9 NUM bracis-28377 5 29 ] ] PUNCT bracis-28377 5 30 , , PUNCT bracis-28377 5 31 in in ADP bracis-28377 5 32 2018 2018 NUM bracis-28377 5 33 , , PUNCT bracis-28377 5 34 there there PRON bracis-28377 5 35 were be VERB bracis-28377 5 36 54.2 54.2 NUM bracis-28377 5 37 million million NUM bracis-28377 5 38 dogs dog NOUN bracis-28377 5 39 , , PUNCT bracis-28377 5 40 39.8 39.8 NUM bracis-28377 5 41 million million NUM bracis-28377 5 42 birds bird NOUN bracis-28377 5 43 , , PUNCT bracis-28377 5 44 23.9 23.9 NUM bracis-28377 5 45 million million NUM bracis-28377 5 46 cats cat NOUN bracis-28377 5 47 , , PUNCT bracis-28377 5 48 19.1 19.1 NUM bracis-28377 5 49 million million NUM bracis-28377 5 50 fishes fish NOUN bracis-28377 5 51 , , PUNCT bracis-28377 5 52 and and CCONJ bracis-28377 5 53 2.3 2.3 NUM bracis-28377 5 54 million million NUM bracis-28377 5 55 reptiles reptile NOUN bracis-28377 5 56 and and CCONJ bracis-28377 5 57 small small ADJ bracis-28377 5 58 mammals mammal NOUN bracis-28377 5 59 , , PUNCT bracis-28377 5 60 considered consider VERB bracis-28377 5 61 pets pet NOUN bracis-28377 5 62 , , PUNCT bracis-28377 5 63 in in ADP bracis-28377 5 64 the the DET bracis-28377 5 65 homes home NOUN bracis-28377 5 66 of of ADP bracis-28377 5 67 brazilian brazilian ADJ bracis-28377 5 68 families family NOUN bracis-28377 5 69 . . PUNCT bracis-28377 6 1 compared compare VERB bracis-28377 6 2 to to ADP bracis-28377 6 3 the the DET bracis-28377 6 4 2013 2013 NUM bracis-28377 6 5 census census NOUN bracis-28377 6 6 data datum NOUN bracis-28377 6 7 , , PUNCT bracis-28377 6 8 the the DET bracis-28377 6 9 cat cat NOUN bracis-28377 6 10 and and CCONJ bracis-28377 6 11 dog dog NOUN bracis-28377 6 12 populations population NOUN bracis-28377 6 13 have have AUX bracis-28377 6 14 grown grow VERB bracis-28377 6 15 in in ADP bracis-28377 6 16 households household NOUN bracis-28377 6 17 at at ADP bracis-28377 6 18 8.1 8.1 NUM bracis-28377 6 19 % % NOUN bracis-28377 6 20 and and CCONJ bracis-28377 6 21 5 5 NUM bracis-28377 6 22 % % NOUN bracis-28377 6 23 , , PUNCT bracis-28377 6 24 respectively respectively ADV bracis-28377 6 25 , , PUNCT bracis-28377 6 26 with with ADP bracis-28377 6 27 an an DET bracis-28377 6 28 overall overall ADJ bracis-28377 6 29 average average ADJ bracis-28377 6 30 growth growth NOUN bracis-28377 6 31 of of ADP bracis-28377 6 32 5.2 5.2 NUM bracis-28377 6 33 % % NOUN bracis-28377 6 34 . . PUNCT bracis-28377 7 1 in in ADP bracis-28377 7 2 addition addition NOUN bracis-28377 7 3 , , PUNCT bracis-28377 7 4 data datum NOUN bracis-28377 7 5 from from ADP bracis-28377 7 6 the the DET bracis-28377 7 7 world world PROPN bracis-28377 7 8 health health NOUN bracis-28377 7 9 organization organization NOUN bracis-28377 7 10 ( ( PUNCT bracis-28377 7 11 who who PRON bracis-28377 7 12 ) ) PUNCT bracis-28377 7 13 , , PUNCT bracis-28377 7 14 published publish VERB bracis-28377 7 15 by by ADP bracis-28377 7 16 são são PROPN bracis-28377 7 17 paulo paulo PROPN bracis-28377 7 18 university university PROPN bracis-28377 7 19 ( ( PUNCT bracis-28377 7 20 usp usp NOUN bracis-28377 7 21 ) ) PUNCT bracis-28377 8 1 [ [ X bracis-28377 8 2 14 14 NUM bracis-28377 8 3 ] ] PUNCT bracis-28377 8 4 , , PUNCT bracis-28377 8 5 indicate indicate VERB bracis-28377 8 6 that that SCONJ bracis-28377 8 7 in in ADP bracis-28377 8 8 brazil brazil PROPN bracis-28377 8 9 there there PRON bracis-28377 8 10 are be VERB bracis-28377 8 11 approximately approximately ADV bracis-28377 8 12 30 30 NUM bracis-28377 8 13 million million NUM bracis-28377 8 14 lost lose VERB bracis-28377 8 15 or or CCONJ bracis-28377 8 16 abandoned abandon VERB bracis-28377 8 17 animals animal NOUN bracis-28377 8 18 , , PUNCT bracis-28377 8 19 with with ADP bracis-28377 8 20 10 10 NUM bracis-28377 8 21 million million NUM bracis-28377 8 22 cats cat NOUN bracis-28377 8 23 and and CCONJ bracis-28377 8 24 20 20 NUM bracis-28377 8 25 million million NUM bracis-28377 8 26 dogs dog NOUN bracis-28377 8 27 . . PUNCT bracis-28377 9 1 the the DET bracis-28377 9 2 uncontrolled uncontrolled ADJ bracis-28377 9 3 proliferation proliferation NOUN bracis-28377 9 4 and and CCONJ bracis-28377 9 5 the the DET bracis-28377 9 6 inappropriate inappropriate ADJ bracis-28377 9 7 management management NOUN bracis-28377 9 8 of of ADP bracis-28377 9 9 animals animal NOUN bracis-28377 9 10 in in ADP bracis-28377 9 11 society society NOUN bracis-28377 9 12 can can AUX bracis-28377 9 13 be be AUX bracis-28377 9 14 factors factor NOUN bracis-28377 9 15 with with ADP bracis-28377 9 16 a a DET bracis-28377 9 17 very very ADV bracis-28377 9 18 negative negative ADJ bracis-28377 9 19 impact impact NOUN bracis-28377 9 20 on on ADP bracis-28377 9 21 public public ADJ bracis-28377 9 22 health health NOUN bracis-28377 9 23 , , PUNCT bracis-28377 9 24 as as SCONJ bracis-28377 9 25 animals animal NOUN bracis-28377 9 26 can can AUX bracis-28377 9 27 transmit transmit VERB bracis-28377 9 28 many many ADJ bracis-28377 9 29 diseases disease NOUN bracis-28377 9 30 and and CCONJ bracis-28377 9 31 parasites parasite NOUN bracis-28377 9 32 . . PUNCT bracis-28377 10 1 therefore therefore ADV bracis-28377 10 2 , , PUNCT bracis-28377 10 3 it it PRON bracis-28377 10 4 is be AUX bracis-28377 10 5 of of ADP bracis-28377 10 6 paramount paramount ADJ bracis-28377 10 7 importance importance NOUN bracis-28377 10 8 to to PART bracis-28377 10 9 carry carry VERB bracis-28377 10 10 out out ADP bracis-28377 10 11 monitoring monitoring NOUN bracis-28377 10 12 , , PUNCT bracis-28377 10 13 control control NOUN bracis-28377 10 14 and and CCONJ bracis-28377 10 15 also also ADV bracis-28377 10 16 traceability traceability NOUN bracis-28377 10 17 of of ADP bracis-28377 10 18 animals animal NOUN bracis-28377 10 19 through through ADP bracis-28377 10 20 complete complete ADJ bracis-28377 10 21 medical medical ADJ bracis-28377 10 22 records record NOUN bracis-28377 10 23 and and CCONJ bracis-28377 10 24 effective effective ADJ bracis-28377 10 25 identification identification NOUN bracis-28377 10 26 mechanisms mechanism NOUN bracis-28377 10 27 to to PART bracis-28377 10 28 enable enable VERB bracis-28377 10 29 monitoring monitoring NOUN bracis-28377 10 30 and and CCONJ bracis-28377 10 31 evaluation evaluation NOUN bracis-28377 10 32 of of ADP bracis-28377 10 33 the the DET bracis-28377 10 34 animal animal NOUN bracis-28377 10 35 population population NOUN bracis-28377 10 36 in in ADP bracis-28377 10 37 the the DET bracis-28377 10 38 country country NOUN bracis-28377 10 39 . . PUNCT bracis-28377 11 1 some some DET bracis-28377 11 2 countries country NOUN bracis-28377 11 3 have have AUX bracis-28377 11 4 already already ADV bracis-28377 11 5 approved approve VERB bracis-28377 11 6 mandatory mandatory ADJ bracis-28377 11 7 pet pet ADJ bracis-28377 11 8 identification identification NOUN bracis-28377 11 9 registration registration NOUN bracis-28377 11 10 , , PUNCT bracis-28377 11 11 such such ADJ bracis-28377 11 12 as as ADP bracis-28377 11 13 south south PROPN bracis-28377 11 14 korea korea PROPN bracis-28377 11 15 , , PUNCT bracis-28377 11 16 where where SCONJ bracis-28377 11 17 all all DET bracis-28377 11 18 dogs dog NOUN bracis-28377 11 19 must must AUX bracis-28377 11 20 be be AUX bracis-28377 11 21 monitored monitor VERB bracis-28377 11 22 through through ADP bracis-28377 11 23 rfid rfid NOUN bracis-28377 11 24 ( ( PUNCT bracis-28377 11 25 radio radio NOUN bracis-28377 11 26 frequency frequency NOUN bracis-28377 11 27 identification identification NOUN bracis-28377 11 28 ) ) PUNCT bracis-28377 11 29 , , PUNCT bracis-28377 11 30 tags tag NOUN bracis-28377 11 31 on on ADP bracis-28377 11 32 collars collar NOUN bracis-28377 11 33 or or CCONJ bracis-28377 11 34 other other ADJ bracis-28377 11 35 devices device NOUN bracis-28377 11 36 [ [ X bracis-28377 11 37 10 10 NUM bracis-28377 11 38 ] ] PUNCT bracis-28377 11 39 . . PUNCT bracis-28377 12 1 as as SCONJ bracis-28377 12 2 occurs occur VERB bracis-28377 12 3 in in ADP bracis-28377 12 4 human human ADJ bracis-28377 12 5 identification identification NOUN bracis-28377 12 6 , , PUNCT bracis-28377 12 7 animal animal NOUN bracis-28377 12 8 identification identification NOUN bracis-28377 12 9 is be AUX bracis-28377 12 10 also also ADV bracis-28377 12 11 susceptible susceptible ADJ bracis-28377 12 12 to to ADP bracis-28377 12 13 fraud fraud NOUN bracis-28377 12 14 . . PUNCT bracis-28377 13 1 in in ADP bracis-28377 13 2 the the DET bracis-28377 13 3 uk uk PROPN bracis-28377 13 4 , , PUNCT bracis-28377 13 5 for for ADP bracis-28377 13 6 example example NOUN bracis-28377 13 7 , , PUNCT bracis-28377 13 8 pet pet ADJ bracis-28377 13 9 insurance insurance NOUN bracis-28377 13 10 fraud fraud NOUN bracis-28377 13 11 cases case NOUN bracis-28377 13 12 have have AUX bracis-28377 13 13 been be AUX bracis-28377 13 14 increasing increase VERB bracis-28377 13 15 by by ADP bracis-28377 13 16 400 400 NUM bracis-28377 13 17 % % NOUN bracis-28377 13 18 per per ADP bracis-28377 13 19 year year NOUN bracis-28377 13 20 , , PUNCT bracis-28377 13 21 reaching reach VERB bracis-28377 13 22 2 2 NUM bracis-28377 13 23 million million NUM bracis-28377 13 24 , , PUNCT bracis-28377 13 25 annually annually ADV bracis-28377 13 26 , , PUNCT bracis-28377 13 27 in in ADP bracis-28377 13 28 claims claim NOUN bracis-28377 13 29 , , PUNCT bracis-28377 13 30 according accord VERB bracis-28377 13 31 to to ADP bracis-28377 13 32 youtalk youtalk NOUN bracis-28377 13 33 - - PUNCT bracis-28377 13 34 insurance insurance NOUN bracis-28377 13 35 [ [ X bracis-28377 13 36 18 18 NUM bracis-28377 13 37 ] ] PUNCT bracis-28377 13 38 . . PUNCT bracis-28377 14 1 therefore therefore ADV bracis-28377 14 2 , , PUNCT bracis-28377 14 3 in in ADP bracis-28377 14 4 addition addition NOUN bracis-28377 14 5 to to ADP bracis-28377 14 6 issues issue NOUN bracis-28377 14 7 of of ADP bracis-28377 14 8 effectiveness effectiveness NOUN bracis-28377 14 9 and and CCONJ bracis-28377 14 10 efficiency efficiency NOUN bracis-28377 14 11 , , PUNCT bracis-28377 14 12 animal animal NOUN bracis-28377 14 13 identification identification NOUN bracis-28377 14 14 methods method NOUN bracis-28377 14 15 also also ADV bracis-28377 14 16 need need VERB bracis-28377 14 17 to to PART bracis-28377 14 18 be be AUX bracis-28377 14 19 robust robust ADJ bracis-28377 14 20 enough enough ADV bracis-28377 14 21 to to PART bracis-28377 14 22 discourage discourage VERB bracis-28377 14 23 , , PUNCT bracis-28377 14 24 stop stop VERB bracis-28377 14 25 or or CCONJ bracis-28377 14 26 even even ADV bracis-28377 14 27 prevent prevent VERB bracis-28377 14 28 fraud fraud NOUN bracis-28377 14 29 . . PUNCT bracis-28377 15 1 according accord VERB bracis-28377 15 2 to to ADP bracis-28377 15 3 kumar kumar PROPN bracis-28377 15 4 et et PROPN bracis-28377 15 5 al al PROPN bracis-28377 15 6 . . PUNCT bracis-28377 16 1 [ [ X bracis-28377 16 2 11 11 NUM bracis-28377 16 3 ] ] PUNCT bracis-28377 16 4 , , PUNCT bracis-28377 16 5 animal animal NOUN bracis-28377 16 6 identification identification NOUN bracis-28377 16 7 approaches approach NOUN bracis-28377 16 8 can can AUX bracis-28377 16 9 be be AUX bracis-28377 16 10 classified classify VERB bracis-28377 16 11 into into ADP bracis-28377 16 12 invasive invasive ADJ bracis-28377 16 13 or or CCONJ bracis-28377 16 14 non non ADJ bracis-28377 16 15 - - ADJ bracis-28377 16 16 invasive invasive ADJ bracis-28377 16 17 . . PUNCT bracis-28377 17 1 examples example NOUN bracis-28377 17 2 of of ADP bracis-28377 17 3 invasive invasive ADJ bracis-28377 17 4 approaches approach NOUN bracis-28377 17 5 are be AUX bracis-28377 17 6 ear ear NOUN bracis-28377 17 7 tags tag NOUN bracis-28377 17 8 , , PUNCT bracis-28377 17 9 ear ear X bracis-28377 17 10 tattoos tattoo NOUN bracis-28377 17 11 and and CCONJ bracis-28377 17 12 microchip microchip NOUN bracis-28377 17 13 implantation implantation NOUN bracis-28377 17 14 in in ADP bracis-28377 17 15 the the DET bracis-28377 17 16 animal animal NOUN bracis-28377 17 17 ’s ’s PART bracis-28377 17 18 body body NOUN bracis-28377 17 19 . . PUNCT bracis-28377 18 1 examples example NOUN bracis-28377 18 2 of of ADP bracis-28377 18 3 non non ADJ bracis-28377 18 4 - - ADJ bracis-28377 18 5 invasive invasive ADJ bracis-28377 18 6 approaches approach NOUN bracis-28377 18 7 are be AUX bracis-28377 18 8 rfid rfid NOUN bracis-28377 18 9 ( ( PUNCT bracis-28377 18 10 radio radio NOUN bracis-28377 18 11 frequency frequency NOUN bracis-28377 18 12 identification identification NOUN bracis-28377 18 13 ) ) PUNCT bracis-28377 18 14 devices device NOUN bracis-28377 18 15 , , PUNCT bracis-28377 18 16 collars collar VERB bracis-28377 18 17 with with ADP bracis-28377 18 18 gps gps PROPN bracis-28377 18 19 ( ( PUNCT bracis-28377 18 20 global global ADJ bracis-28377 18 21 positioning positioning NOUN bracis-28377 18 22 system system NOUN bracis-28377 18 23 ) ) PUNCT bracis-28377 18 24 , , PUNCT bracis-28377 18 25 internet internet NOUN bracis-28377 18 26 of of ADP bracis-28377 18 27 things thing NOUN bracis-28377 18 28 ( ( PUNCT bracis-28377 18 29 iot iot NOUN bracis-28377 18 30 ) ) PUNCT bracis-28377 18 31 devices device NOUN bracis-28377 18 32 , , PUNCT bracis-28377 18 33 bluetooth bluetooth PROPN bracis-28377 18 34 devices device NOUN bracis-28377 18 35 , , PUNCT bracis-28377 18 36 and and CCONJ bracis-28377 18 37 biometric biometric ADJ bracis-28377 18 38 features feature NOUN bracis-28377 18 39 ( ( PUNCT bracis-28377 18 40 e.g. e.g. ADV bracis-28377 18 41 face face NOUN bracis-28377 18 42 , , PUNCT bracis-28377 18 43 iris iris NOUN bracis-28377 18 44 , , PUNCT bracis-28377 18 45 retina retina NOUN bracis-28377 18 46 and and CCONJ bracis-28377 18 47 muzzles muzzle NOUN bracis-28377 18 48 ) ) PUNCT bracis-28377 18 49 . . PUNCT bracis-28377 19 1 among among ADP bracis-28377 19 2 these these DET bracis-28377 19 3 approaches approach NOUN bracis-28377 19 4 , , PUNCT bracis-28377 19 5 animal animal NOUN bracis-28377 19 6 identification identification NOUN bracis-28377 19 7 based base VERB bracis-28377 19 8 on on ADP bracis-28377 19 9 biometric biometric ADJ bracis-28377 19 10 features feature NOUN bracis-28377 19 11 has have AUX bracis-28377 19 12 proved prove VERB bracis-28377 19 13 to to PART bracis-28377 19 14 be be AUX bracis-28377 19 15 the the DET bracis-28377 19 16 best good ADJ bracis-28377 19 17 choice choice NOUN bracis-28377 19 18 , , PUNCT bracis-28377 19 19 since since SCONJ bracis-28377 19 20 it it PRON bracis-28377 19 21 is be AUX bracis-28377 19 22 effective effective ADJ bracis-28377 19 23 , , PUNCT bracis-28377 19 24 efficient efficient ADJ bracis-28377 19 25 and and CCONJ bracis-28377 19 26 does do AUX bracis-28377 19 27 not not PART bracis-28377 19 28 require require VERB bracis-28377 19 29 special special ADJ bracis-28377 19 30 devices device NOUN bracis-28377 19 31 , , PUNCT bracis-28377 19 32 which which PRON bracis-28377 19 33 would would AUX bracis-28377 19 34 introduce introduce VERB bracis-28377 19 35 more more ADJ bracis-28377 19 36 costs cost NOUN bracis-28377 19 37 to to ADP bracis-28377 19 38 the the DET bracis-28377 19 39 identification identification NOUN bracis-28377 19 40 process process NOUN bracis-28377 19 41 , , PUNCT bracis-28377 19 42 being be AUX bracis-28377 19 43 also also ADV bracis-28377 19 44 important important ADJ bracis-28377 19 45 to to PART bracis-28377 19 46 discourage discourage VERB bracis-28377 19 47 or or CCONJ bracis-28377 19 48 prevent prevent VERB bracis-28377 19 49 fraud fraud NOUN bracis-28377 19 50 . . PUNCT bracis-28377 20 1 among among ADP bracis-28377 20 2 several several ADJ bracis-28377 20 3 human human ADJ bracis-28377 20 4 biometric biometric ADJ bracis-28377 20 5 traits trait NOUN bracis-28377 20 6 , , PUNCT bracis-28377 20 7 the the DET bracis-28377 20 8 face face NOUN bracis-28377 20 9 is be AUX bracis-28377 20 10 one one NUM bracis-28377 20 11 of of ADP bracis-28377 20 12 the the DET bracis-28377 20 13 most most ADV bracis-28377 20 14 popular popular ADJ bracis-28377 20 15 due due ADP bracis-28377 20 16 to to ADP bracis-28377 20 17 its its PRON bracis-28377 20 18 high high ADJ bracis-28377 20 19 coverage coverage NOUN bracis-28377 20 20 , , PUNCT bracis-28377 20 21 high high ADJ bracis-28377 20 22 acceptability acceptability NOUN bracis-28377 20 23 , , PUNCT bracis-28377 20 24 easiness easiness NOUN bracis-28377 20 25 of of ADP bracis-28377 20 26 capture capture NOUN bracis-28377 20 27 at at ADP bracis-28377 20 28 a a DET bracis-28377 20 29 distance distance NOUN bracis-28377 20 30 , , PUNCT bracis-28377 20 31 convenience convenience NOUN bracis-28377 20 32 of of ADP bracis-28377 20 33 use use NOUN bracis-28377 20 34 and and CCONJ bracis-28377 20 35 lower low ADJ bracis-28377 20 36 costs cost NOUN bracis-28377 20 37 since since SCONJ bracis-28377 20 38 digital digital ADJ bracis-28377 20 39 cameras camera NOUN bracis-28377 20 40 are be AUX bracis-28377 20 41 cheap cheap ADJ bracis-28377 20 42 and and CCONJ bracis-28377 20 43 popular popular ADJ bracis-28377 20 44 sensors sensor NOUN bracis-28377 20 45 . . PUNCT bracis-28377 21 1 motivated motivate VERB bracis-28377 21 2 by by ADP bracis-28377 21 3 the the DET bracis-28377 21 4 success success NOUN bracis-28377 21 5 of of ADP bracis-28377 21 6 modern modern ADJ bracis-28377 21 7 human human ADJ bracis-28377 21 8 face face NOUN bracis-28377 21 9 recognition recognition NOUN bracis-28377 21 10 methods method NOUN bracis-28377 21 11 , , PUNCT bracis-28377 21 12 in in ADP bracis-28377 21 13 this this DET bracis-28377 21 14 paper paper NOUN bracis-28377 21 15 , , PUNCT bracis-28377 21 16 we we PRON bracis-28377 21 17 propose propose VERB bracis-28377 21 18 a a DET bracis-28377 21 19 dog dog NOUN bracis-28377 21 20 face face NOUN bracis-28377 21 21 recognition recognition NOUN bracis-28377 21 22 method method NOUN bracis-28377 21 23 based base VERB bracis-28377 21 24 on on ADP bracis-28377 21 25 vision vision NOUN bracis-28377 21 26 transformers transformer NOUN bracis-28377 21 27 , , PUNCT bracis-28377 21 28 a a DET bracis-28377 21 29 deep deep ADJ bracis-28377 21 30 learning learning NOUN bracis-28377 21 31 approach approach NOUN bracis-28377 21 32 that that PRON bracis-28377 21 33 decomposes decompose VERB bracis-28377 21 34 the the DET bracis-28377 21 35 input input NOUN bracis-28377 21 36 image image NOUN bracis-28377 21 37 into into ADP bracis-28377 21 38 a a DET bracis-28377 21 39 sequence sequence NOUN bracis-28377 21 40 of of ADP bracis-28377 21 41 patches patch NOUN bracis-28377 21 42 and and CCONJ bracis-28377 21 43 applies apply VERB bracis-28377 21 44 self self NOUN bracis-28377 21 45 - - PUNCT bracis-28377 21 46 attention attention NOUN bracis-28377 21 47 to to ADP bracis-28377 21 48 these these DET bracis-28377 21 49 patches patch NOUN bracis-28377 21 50 to to PART bracis-28377 21 51 capture capture VERB bracis-28377 21 52 spatial spatial ADJ bracis-28377 21 53 relationships relationship NOUN bracis-28377 21 54 between between ADP bracis-28377 21 55 them they PRON bracis-28377 21 56 . . PUNCT bracis-28377 22 1 2 2 NUM bracis-28377 22 2 biometric biometric ADJ bracis-28377 22 3 dog dog NOUN bracis-28377 22 4 identification identification NOUN bracis-28377 22 5 biometric biometric ADJ bracis-28377 22 6 dog dog NOUN bracis-28377 22 7 identification identification NOUN bracis-28377 22 8 approaches approach NOUN bracis-28377 22 9 have have AUX bracis-28377 22 10 been be AUX bracis-28377 22 11 reported report VERB bracis-28377 22 12 in in ADP bracis-28377 22 13 the the DET bracis-28377 22 14 literature literature NOUN bracis-28377 22 15 by by ADP bracis-28377 22 16 using use VERB bracis-28377 22 17 snout snout NOUN bracis-28377 22 18 and and CCONJ bracis-28377 22 19 facial facial ADJ bracis-28377 22 20 features feature NOUN bracis-28377 22 21 , , PUNCT bracis-28377 22 22 among among ADP bracis-28377 22 23 others other NOUN bracis-28377 22 24 [ [ X bracis-28377 22 25 10 10 NUM bracis-28377 22 26 , , PUNCT bracis-28377 22 27 12 12 NUM bracis-28377 22 28 , , PUNCT bracis-28377 22 29 17 17 NUM bracis-28377 22 30 , , PUNCT bracis-28377 22 31 21 21 NUM bracis-28377 22 32 ] ] PUNCT bracis-28377 22 33 . . PUNCT bracis-28377 23 1 this this DET bracis-28377 23 2 section section NOUN bracis-28377 23 3 presents present VERB bracis-28377 23 4 some some DET bracis-28377 23 5 approaches approach NOUN bracis-28377 23 6 proposed propose VERB bracis-28377 23 7 so so ADV bracis-28377 23 8 far far ADV bracis-28377 23 9 , , PUNCT bracis-28377 23 10 their their PRON bracis-28377 23 11 identification identification NOUN bracis-28377 23 12 rates rate NOUN bracis-28377 23 13 , , PUNCT bracis-28377 23 14 as as ADV bracis-28377 23 15 well well ADV bracis-28377 23 16 as as ADP bracis-28377 23 17 their their PRON bracis-28377 23 18 advantages advantage NOUN bracis-28377 23 19 and and CCONJ bracis-28377 23 20 disadvantages disadvantage NOUN bracis-28377 23 21 . . PUNCT bracis-28377 24 1 2.1 2.1 NUM bracis-28377 24 2 identification identification NOUN bracis-28377 24 3 using use VERB bracis-28377 24 4 snout snout NOUN bracis-28377 24 5 features feature VERB bracis-28377 24 6 snout snout NOUN bracis-28377 24 7 is be AUX bracis-28377 24 8 the the DET bracis-28377 24 9 region region NOUN bracis-28377 24 10 of of ADP bracis-28377 24 11 an an DET bracis-28377 24 12 animal animal NOUN bracis-28377 24 13 ’s ’s PART bracis-28377 24 14 face face NOUN bracis-28377 24 15 consisting consist VERB bracis-28377 24 16 of of ADP bracis-28377 24 17 its its PRON bracis-28377 24 18 nose nose NOUN bracis-28377 24 19 , , PUNCT bracis-28377 24 20 mouth mouth NOUN bracis-28377 24 21 , , PUNCT bracis-28377 24 22 and and CCONJ bracis-28377 24 23 jaw jaw NOUN bracis-28377 24 24 . . PUNCT bracis-28377 25 1 in in ADP bracis-28377 25 2 many many ADJ bracis-28377 25 3 animals animal NOUN bracis-28377 25 4 , , PUNCT bracis-28377 25 5 this this DET bracis-28377 25 6 structure structure NOUN bracis-28377 25 7 is be AUX bracis-28377 25 8 called call VERB bracis-28377 25 9 muzzle muzzle NOUN bracis-28377 25 10 . . PUNCT bracis-28377 26 1 the the DET bracis-28377 26 2 muzzle muzzle NOUN bracis-28377 26 3 information information NOUN bracis-28377 26 4 can can AUX bracis-28377 26 5 be be AUX bracis-28377 26 6 used use VERB bracis-28377 26 7 as as ADP bracis-28377 26 8 a a DET bracis-28377 26 9 biometric biometric ADJ bracis-28377 26 10 feature feature NOUN bracis-28377 26 11 for for ADP bracis-28377 26 12 dog dog NOUN bracis-28377 26 13 identification identification NOUN bracis-28377 26 14 , , PUNCT bracis-28377 26 15 as as SCONJ bracis-28377 26 16 reported report VERB bracis-28377 26 17 by by ADP bracis-28377 26 18 jang jang PROPN bracis-28377 26 19 et et PROPN bracis-28377 26 20 al al PROPN bracis-28377 26 21 . . PUNCT bracis-28377 26 22     SPACE bracis-28377 27 1 [ [ X bracis-28377 27 2 10 10 NUM bracis-28377 27 3 ] ] PUNCT bracis-28377 27 4 . . PUNCT bracis-28377 28 1 figure figure NOUN bracis-28377 28 2     SPACE bracis-28377 28 3 1 1 NUM bracis-28377 28 4 shows show NOUN bracis-28377 28 5 , , PUNCT bracis-28377 28 6 on on ADP bracis-28377 28 7 the the DET bracis-28377 28 8 left left NOUN bracis-28377 28 9 , , PUNCT bracis-28377 28 10 a a DET bracis-28377 28 11 device device NOUN bracis-28377 28 12 used use VERB bracis-28377 28 13 to to PART bracis-28377 28 14 capture capture VERB bracis-28377 28 15 images image NOUN bracis-28377 28 16 of of ADP bracis-28377 28 17 dog dog NOUN bracis-28377 28 18 snouts snout NOUN bracis-28377 28 19 and and CCONJ bracis-28377 28 20 , , PUNCT bracis-28377 28 21 on on ADP bracis-28377 28 22 the the DET bracis-28377 28 23 right right NOUN bracis-28377 28 24 , , PUNCT bracis-28377 28 25 a a DET bracis-28377 28 26 captured capture VERB bracis-28377 28 27 image image NOUN bracis-28377 28 28 , , PUNCT bracis-28377 28 29 with with ADP bracis-28377 28 30 the the DET bracis-28377 28 31 region region NOUN bracis-28377 28 32 of of ADP bracis-28377 28 33 interest interest NOUN bracis-28377 28 34 for for ADP bracis-28377 28 35 biometric biometric ADJ bracis-28377 28 36 identification identification NOUN bracis-28377 28 37 highlighted highlight VERB bracis-28377 28 38 . . PUNCT bracis-28377 29 1 fig fig NOUN bracis-28377 29 2 . . PUNCT bracis-28377 30 1 1 1 X bracis-28377 30 2 . . X bracis-28377 30 3 device device NOUN bracis-28377 30 4 used use VERB bracis-28377 30 5 to to PART bracis-28377 30 6 collect collect VERB bracis-28377 30 7 images image NOUN bracis-28377 30 8 of of ADP bracis-28377 30 9 dog dog NOUN bracis-28377 30 10 snouts snout NOUN bracis-28377 30 11 ( ( PUNCT bracis-28377 30 12 on on ADP bracis-28377 30 13 the the DET bracis-28377 30 14 left left NOUN bracis-28377 30 15 ) ) PUNCT bracis-28377 30 16 and and CCONJ bracis-28377 30 17 the the DET bracis-28377 30 18 region region NOUN bracis-28377 30 19 of of ADP bracis-28377 30 20 interest interest NOUN bracis-28377 30 21 of of ADP bracis-28377 30 22 a a DET bracis-28377 30 23 collected collect VERB bracis-28377 30 24 image image NOUN bracis-28377 30 25 , , PUNCT bracis-28377 30 26 used use VERB bracis-28377 30 27 for for ADP bracis-28377 30 28 feature feature NOUN bracis-28377 30 29 extraction extraction NOUN bracis-28377 30 30 ( ( PUNCT bracis-28377 30 31 on on ADP bracis-28377 30 32 the the DET bracis-28377 30 33 right right NOUN bracis-28377 30 34 ) ) PUNCT bracis-28377 31 1 [ [ X bracis-28377 31 2 10 10 NUM bracis-28377 31 3 ] ] PUNCT bracis-28377 31 4 . . PUNCT bracis-28377 32 1 full full ADJ bracis-28377 32 2 size size NOUN bracis-28377 32 3 image image NOUN bracis-28377 32 4 jang jang PROPN bracis-28377 32 5 et et PROPN bracis-28377 32 6 al al PROPN bracis-28377 32 7 . . PUNCT bracis-28377 32 8     SPACE bracis-28377 33 1 [ [ X bracis-28377 33 2 10 10 NUM bracis-28377 33 3 ] ] PUNCT bracis-28377 33 4 have have AUX bracis-28377 33 5 proposed propose VERB bracis-28377 33 6 the the DET bracis-28377 33 7 use use NOUN bracis-28377 33 8 of of ADP bracis-28377 33 9 some some DET bracis-28377 33 10 feature feature NOUN bracis-28377 33 11 extraction extraction NOUN bracis-28377 33 12 methods method NOUN bracis-28377 33 13 , , PUNCT bracis-28377 33 14 such such ADJ bracis-28377 33 15 as as ADP bracis-28377 33 16 sift sift ADJ bracis-28377 33 17 ( ( PUNCT bracis-28377 33 18 scale scale NOUN bracis-28377 33 19 - - PUNCT bracis-28377 33 20 invariant invariant ADJ bracis-28377 33 21 feature feature NOUN bracis-28377 33 22 transform transform NOUN bracis-28377 33 23 ) ) PUNCT bracis-28377 33 24 , , PUNCT bracis-28377 33 25 surf surf NOUN bracis-28377 33 26 ( ( PUNCT bracis-28377 33 27 speeded speed VERB bracis-28377 33 28 up up ADP bracis-28377 33 29 robust robust ADJ bracis-28377 33 30 features feature NOUN bracis-28377 33 31 ) ) PUNCT bracis-28377 33 32 , , PUNCT bracis-28377 33 33 brisk brisk ADJ bracis-28377 33 34 ( ( PUNCT bracis-28377 33 35 binary binary ADJ bracis-28377 33 36 robust robust ADJ bracis-28377 33 37 invariant invariant ADJ bracis-28377 33 38 scalable scalable ADJ bracis-28377 33 39 keypoints keypoint NOUN bracis-28377 33 40 ) ) PUNCT bracis-28377 33 41 and and CCONJ bracis-28377 33 42 orb orb NOUN bracis-28377 33 43 ( ( PUNCT bracis-28377 33 44 oriented orient VERB bracis-28377 33 45 fast fast ADJ bracis-28377 33 46 and and CCONJ bracis-28377 33 47 rotated rotate VERB bracis-28377 33 48 brief brief NOUN bracis-28377 33 49 ) ) PUNCT bracis-28377 33 50 , , PUNCT bracis-28377 33 51 and and CCONJ bracis-28377 33 52 the the DET bracis-28377 33 53 method method NOUN bracis-28377 33 54 that that PRON bracis-28377 33 55 presented present VERB bracis-28377 33 56 the the DET bracis-28377 33 57 best good ADJ bracis-28377 33 58 performance performance NOUN bracis-28377 33 59 was be AUX bracis-28377 33 60 the the DET bracis-28377 33 61 orb orb NOUN bracis-28377 33 62 with with ADP bracis-28377 33 63 eer eer PROPN bracis-28377 33 64 ( ( PUNCT bracis-28377 33 65 equal equal ADJ bracis-28377 33 66 error error NOUN bracis-28377 33 67 rate rate NOUN bracis-28377 33 68 ) ) PUNCT bracis-28377 33 69 of of ADP bracis-28377 33 70 0.35 0.35 NUM bracis-28377 33 71 % % NOUN bracis-28377 33 72 , , PUNCT bracis-28377 33 73 when when SCONJ bracis-28377 33 74 evaluated evaluate VERB bracis-28377 33 75 with with ADP bracis-28377 33 76 55 55 NUM bracis-28377 33 77 dog dog NOUN bracis-28377 33 78 muzzle muzzle NOUN bracis-28377 33 79 pattern pattern NOUN bracis-28377 33 80 images image NOUN bracis-28377 33 81 acquired acquire VERB bracis-28377 33 82 from from ADP bracis-28377 33 83 11 11 NUM bracis-28377 33 84 dogs dog NOUN bracis-28377 33 85 and and CCONJ bracis-28377 33 86 990 990 NUM bracis-28377 33 87 images image NOUN bracis-28377 33 88 augmented augment VERB bracis-28377 33 89 by by ADP bracis-28377 33 90 the the DET bracis-28377 33 91 image image NOUN bracis-28377 33 92 deformation deformation NOUN bracis-28377 33 93 ( ( PUNCT bracis-28377 33 94 i.e. i.e. X bracis-28377 33 95 , , PUNCT bracis-28377 33 96 angle angle NOUN bracis-28377 33 97 , , PUNCT bracis-28377 33 98 illumination illumination NOUN bracis-28377 33 99 , , PUNCT bracis-28377 33 100 noise noise NOUN bracis-28377 33 101 , , PUNCT bracis-28377 33 102 affine affine NOUN bracis-28377 33 103 transform transform NOUN bracis-28377 33 104 ) ) PUNCT bracis-28377 33 105 . . PUNCT bracis-28377 34 1 however however ADV bracis-28377 34 2 , , PUNCT bracis-28377 34 3 despite despite SCONJ bracis-28377 34 4 this this DET bracis-28377 34 5 low low ADJ bracis-28377 34 6 error error NOUN bracis-28377 34 7 rate rate NOUN bracis-28377 34 8 , , PUNCT bracis-28377 34 9 the the DET bracis-28377 34 10 technique technique NOUN bracis-28377 34 11 proposed propose VERB bracis-28377 34 12 by by ADP bracis-28377 34 13 these these DET bracis-28377 34 14 authors author NOUN bracis-28377 34 15 has have VERB bracis-28377 34 16 some some DET bracis-28377 34 17 disadvantages disadvantage NOUN bracis-28377 34 18 : : PUNCT bracis-28377 34 19 even even ADV bracis-28377 34 20 though though SCONJ bracis-28377 34 21 it it PRON bracis-28377 34 22 is be AUX bracis-28377 34 23 not not PART bracis-28377 34 24 an an DET bracis-28377 34 25 invasive invasive ADJ bracis-28377 34 26 technique technique NOUN bracis-28377 34 27 , , PUNCT bracis-28377 34 28 it it PRON bracis-28377 34 29 is be AUX bracis-28377 34 30 necessary necessary ADJ bracis-28377 34 31 to to PART bracis-28377 34 32 immobilize immobilize VERB bracis-28377 34 33 the the DET bracis-28377 34 34 animal animal NOUN bracis-28377 34 35 to to PART bracis-28377 34 36 capture capture VERB bracis-28377 34 37 the the DET bracis-28377 34 38 image image NOUN bracis-28377 34 39 and and CCONJ bracis-28377 34 40 extract extract VERB bracis-28377 34 41 features feature NOUN bracis-28377 34 42 from from ADP bracis-28377 34 43 the the DET bracis-28377 34 44 snout snout NOUN bracis-28377 34 45 ; ; PUNCT bracis-28377 34 46 it it PRON bracis-28377 34 47 is be AUX bracis-28377 34 48 important important ADJ bracis-28377 34 49 to to PART bracis-28377 34 50 avoid avoid VERB bracis-28377 34 51 light light ADJ bracis-28377 34 52 reflection reflection NOUN bracis-28377 34 53 and and CCONJ bracis-28377 34 54 blur blur NOUN bracis-28377 34 55 in in ADP bracis-28377 34 56 muzzle muzzle NOUN bracis-28377 34 57 images image NOUN bracis-28377 34 58 ; ; PUNCT bracis-28377 34 59 it it PRON bracis-28377 34 60 is be AUX bracis-28377 34 61 necessary necessary ADJ bracis-28377 34 62 to to PART bracis-28377 34 63 have have VERB bracis-28377 34 64 good good ADJ bracis-28377 34 65 hygiene hygiene NOUN bracis-28377 34 66 of of ADP bracis-28377 34 67 the the DET bracis-28377 34 68 muzzle muzzle NOUN bracis-28377 34 69 . . PUNCT bracis-28377 35 1 figure figure NOUN bracis-28377 35 2     SPACE bracis-28377 35 3 2 2 NUM bracis-28377 35 4 shows show VERB bracis-28377 35 5 examples example NOUN bracis-28377 35 6 of of ADP bracis-28377 35 7 dog dog NOUN bracis-28377 35 8 muzzle muzzle NOUN bracis-28377 35 9 images image NOUN bracis-28377 35 10 that that PRON bracis-28377 35 11 were be AUX bracis-28377 35 12 discarded discard VERB bracis-28377 35 13 in in ADP bracis-28377 35 14 the the DET bracis-28377 35 15 work work NOUN bracis-28377 35 16 by by ADP bracis-28377 35 17 jang jang PROPN bracis-28377 35 18 et et PROPN bracis-28377 35 19 al al PROPN bracis-28377 35 20 . . PUNCT bracis-28377 35 21     SPACE bracis-28377 36 1 [ [ X bracis-28377 36 2 10 10 NUM bracis-28377 36 3 ] ] PUNCT bracis-28377 36 4 due due ADP bracis-28377 36 5 to to ADP bracis-28377 36 6 light light ADJ bracis-28377 36 7 reflection reflection NOUN bracis-28377 36 8 and and CCONJ bracis-28377 36 9 blurring blurring NOUN bracis-28377 36 10 caused cause VERB bracis-28377 36 11 by by ADP bracis-28377 36 12 dog dog NOUN bracis-28377 36 13 movements movement NOUN bracis-28377 36 14 . . PUNCT bracis-28377 37 1 fig fig NOUN bracis-28377 37 2 . . PUNCT bracis-28377 38 1 2 2 X bracis-28377 38 2 . . X bracis-28377 38 3 dog dog NOUN bracis-28377 38 4 muzzle muzzle NOUN bracis-28377 38 5 images image NOUN bracis-28377 38 6 with with ADP bracis-28377 38 7 light light ADJ bracis-28377 38 8 reflection reflection NOUN bracis-28377 38 9 ( ( PUNCT bracis-28377 38 10 1–2 1–2 NOUN bracis-28377 38 11 ) ) PUNCT bracis-28377 38 12 and and CCONJ bracis-28377 38 13 blurring blur VERB bracis-28377 38 14 ( ( PUNCT bracis-28377 38 15 3–4 3–4 NUM bracis-28377 38 16 ) ) PUNCT bracis-28377 38 17 [ [ X bracis-28377 38 18 10 10 NUM bracis-28377 38 19 ] ] PUNCT bracis-28377 38 20 . . PUNCT bracis-28377 39 1 full full ADJ bracis-28377 39 2 size size NOUN bracis-28377 39 3 image image NOUN bracis-28377 39 4 2.2 2.2 NUM bracis-28377 39 5 identification identification NOUN bracis-28377 39 6 using use VERB bracis-28377 39 7 facial facial ADJ bracis-28377 39 8 features feature NOUN bracis-28377 39 9 recent recent ADJ bracis-28377 39 10 works work NOUN bracis-28377 39 11 on on ADP bracis-28377 39 12 dog dog NOUN bracis-28377 39 13 biometric biometric ADJ bracis-28377 39 14 identification identification NOUN bracis-28377 39 15 based base VERB bracis-28377 39 16 on on ADP bracis-28377 39 17 face face NOUN bracis-28377 39 18 features feature NOUN bracis-28377 39 19 have have AUX bracis-28377 39 20 used use VERB bracis-28377 39 21 the the DET bracis-28377 39 22 dogfacenet dogfacenet NOUN bracis-28377 39 23 database database NOUN bracis-28377 39 24 , , PUNCT bracis-28377 39 25 proposed propose VERB bracis-28377 39 26 by by ADP bracis-28377 39 27 mougeot mougeot PROPN bracis-28377 39 28 , , PUNCT bracis-28377 39 29 li li PROPN bracis-28377 39 30 and and CCONJ bracis-28377 39 31 jia jia PROPN bracis-28377 40 1 [ [ X bracis-28377 40 2 17 17 NUM bracis-28377 40 3 ] ] PUNCT bracis-28377 40 4 . . PUNCT bracis-28377 41 1 this this DET bracis-28377 41 2 database database NOUN bracis-28377 41 3 contains contain VERB bracis-28377 41 4 3,148 3,148 NUM bracis-28377 41 5 images image NOUN bracis-28377 41 6 of of ADP bracis-28377 41 7 485 485 NUM bracis-28377 41 8 dogs dog NOUN bracis-28377 41 9 , , PUNCT bracis-28377 41 10 with with ADP bracis-28377 41 11 at at ADV bracis-28377 41 12 least least ADV bracis-28377 41 13 two two NUM bracis-28377 41 14 images image NOUN bracis-28377 41 15 per per ADP bracis-28377 41 16 individual individual NOUN bracis-28377 41 17 . . PUNCT bracis-28377 42 1 however however ADV bracis-28377 42 2 , , PUNCT bracis-28377 42 3 the the DET bracis-28377 42 4 publicly publicly ADV bracis-28377 42 5 available available ADJ bracis-28377 42 6 database database NOUN bracis-28377 42 7 that that PRON bracis-28377 42 8 was be AUX bracis-28377 42 9 developed develop VERB bracis-28377 42 10 after after SCONJ bracis-28377 42 11 the the DET bracis-28377 42 12 published publish VERB bracis-28377 42 13 work work NOUN bracis-28377 42 14 contains contain VERB bracis-28377 42 15 8,363 8,363 NUM bracis-28377 42 16 images image NOUN bracis-28377 42 17 of of ADP bracis-28377 42 18 1,393 1,393 NUM bracis-28377 42 19 dogs dog NOUN bracis-28377 42 20 . . PUNCT bracis-28377 43 1 figure figure NOUN bracis-28377 43 2     SPACE bracis-28377 43 3 3 3 NUM bracis-28377 43 4 shows show VERB bracis-28377 43 5 four four NUM bracis-28377 43 6 images image NOUN bracis-28377 43 7 of of ADP bracis-28377 43 8 four four NUM bracis-28377 43 9 dogs dog NOUN bracis-28377 43 10 captured capture VERB bracis-28377 43 11 from from ADP bracis-28377 43 12 different different ADJ bracis-28377 43 13 viewpoints viewpoint NOUN bracis-28377 43 14 that that PRON bracis-28377 43 15 compose compose VERB bracis-28377 43 16 the the DET bracis-28377 43 17 dogfacenet dogfacenet NOUN bracis-28377 43 18 dataset dataset NOUN bracis-28377 43 19 . . PUNCT bracis-28377 44 1 fig fig NOUN bracis-28377 44 2 . . PUNCT bracis-28377 45 1 3 3 X bracis-28377 45 2 . . X bracis-28377 45 3 samples sample NOUN bracis-28377 45 4 of of ADP bracis-28377 45 5 four four NUM bracis-28377 45 6 dog dog NOUN bracis-28377 45 7 images image NOUN bracis-28377 45 8 from from ADP bracis-28377 45 9 the the DET bracis-28377 45 10 dogfacenet dogfacenet NOUN bracis-28377 45 11 dataset dataset NOUN bracis-28377 45 12 [ [ X bracis-28377 45 13 17 17 NUM bracis-28377 45 14 ] ] PUNCT bracis-28377 45 15 . . PUNCT bracis-28377 46 1 full full ADJ bracis-28377 46 2 size size NOUN bracis-28377 46 3 image image NOUN bracis-28377 46 4 the the DET bracis-28377 46 5 images image NOUN bracis-28377 46 6 of of ADP bracis-28377 46 7 the the DET bracis-28377 46 8 dog dog NOUN bracis-28377 46 9 ’s ’s PART bracis-28377 46 10 faces face NOUN bracis-28377 46 11 were be AUX bracis-28377 46 12 all all ADV bracis-28377 46 13 aligned align VERB bracis-28377 46 14 , , PUNCT bracis-28377 46 15 using use VERB bracis-28377 46 16 the the DET bracis-28377 46 17 animal animal NOUN bracis-28377 46 18 ’s ’s PART bracis-28377 46 19 eyes eye NOUN bracis-28377 46 20 as as ADP bracis-28377 46 21 a a DET bracis-28377 46 22 reference reference NOUN bracis-28377 46 23 . . PUNCT bracis-28377 47 1 figure figure NOUN bracis-28377 47 2     SPACE bracis-28377 47 3 4 4 NUM bracis-28377 47 4 shows show VERB bracis-28377 47 5 the the DET bracis-28377 47 6 process process NOUN bracis-28377 47 7 of of ADP bracis-28377 47 8 detecting detect VERB bracis-28377 47 9 the the DET bracis-28377 47 10 fiducial fiducial ADJ bracis-28377 47 11 points point NOUN bracis-28377 47 12 ( ( PUNCT bracis-28377 47 13 eyes eye NOUN bracis-28377 47 14 and and CCONJ bracis-28377 47 15 snout snout NOUN bracis-28377 47 16 ) ) PUNCT bracis-28377 47 17 on on ADP bracis-28377 47 18 the the DET bracis-28377 47 19 face face NOUN bracis-28377 47 20 of of ADP bracis-28377 47 21 a a DET bracis-28377 47 22 dog dog NOUN bracis-28377 47 23 ( ( PUNCT bracis-28377 47 24 on on ADP bracis-28377 47 25 the the DET bracis-28377 47 26 left left NOUN bracis-28377 47 27 ) ) PUNCT bracis-28377 47 28 , , PUNCT bracis-28377 47 29 used use VERB bracis-28377 47 30 for for ADP bracis-28377 47 31 the the DET bracis-28377 47 32 alignment alignment NOUN bracis-28377 47 33 of of ADP bracis-28377 47 34 the the DET bracis-28377 47 35 images image NOUN bracis-28377 47 36 , , PUNCT bracis-28377 47 37 and and CCONJ bracis-28377 47 38 the the DET bracis-28377 47 39 result result NOUN bracis-28377 47 40 of of ADP bracis-28377 47 41 the the DET bracis-28377 47 42 alignment alignment NOUN bracis-28377 47 43 and and CCONJ bracis-28377 47 44 cropping crop VERB bracis-28377 47 45 processes process NOUN bracis-28377 47 46 applied apply VERB bracis-28377 47 47 on on ADP bracis-28377 47 48 the the DET bracis-28377 47 49 six six NUM bracis-28377 47 50 images image NOUN bracis-28377 47 51 of of ADP bracis-28377 47 52 one one NUM bracis-28377 47 53 dog dog NOUN bracis-28377 47 54 from from ADP bracis-28377 47 55 the the DET bracis-28377 47 56 dogfacenet dogfacenet NOUN bracis-28377 47 57 dataset dataset NOUN bracis-28377 47 58 ( ( PUNCT bracis-28377 47 59 on on ADP bracis-28377 47 60 the the DET bracis-28377 47 61 right right NOUN bracis-28377 47 62 ) ) PUNCT bracis-28377 47 63 . . PUNCT bracis-28377 48 1 the the DET bracis-28377 48 2 dog dog NOUN bracis-28377 48 3 face face VERB bracis-28377 48 4 images image NOUN bracis-28377 48 5 of of ADP bracis-28377 48 6 the the DET bracis-28377 48 7 dogfacenet dogfacenet NOUN bracis-28377 48 8 database database NOUN bracis-28377 48 9 were be AUX bracis-28377 48 10 divided divide VERB bracis-28377 48 11 into into ADP bracis-28377 48 12 training training NOUN bracis-28377 48 13 and and CCONJ bracis-28377 48 14 test test NOUN bracis-28377 48 15 sets set NOUN bracis-28377 48 16 , , PUNCT bracis-28377 48 17 with with ADP bracis-28377 48 18 90 90 NUM bracis-28377 48 19 % % NOUN bracis-28377 48 20 and and CCONJ bracis-28377 48 21 10 10 NUM bracis-28377 48 22 % % NOUN bracis-28377 48 23 , , PUNCT bracis-28377 48 24 respectively respectively ADV bracis-28377 48 25 . . PUNCT bracis-28377 49 1 fig fig NOUN bracis-28377 49 2 . . PUNCT bracis-28377 50 1 4 4 X bracis-28377 50 2 . . X bracis-28377 50 3 detection detection NOUN bracis-28377 50 4 of of ADP bracis-28377 50 5 the the DET bracis-28377 50 6 animal animal NOUN bracis-28377 50 7 ’s ’s PART bracis-28377 50 8 eyes eye NOUN bracis-28377 50 9 and and CCONJ bracis-28377 50 10 snout snout NOUN bracis-28377 50 11 ( ( PUNCT bracis-28377 50 12 left leave VERB bracis-28377 50 13 ) ) PUNCT bracis-28377 50 14 and and CCONJ bracis-28377 50 15 the the DET bracis-28377 50 16 results result NOUN bracis-28377 50 17 obtained obtain VERB bracis-28377 50 18 after after ADP bracis-28377 50 19 the the DET bracis-28377 50 20 aligning aligning NOUN bracis-28377 50 21 and and CCONJ bracis-28377 50 22 cropping crop VERB bracis-28377 50 23 processes process NOUN bracis-28377 50 24 , , PUNCT bracis-28377 50 25 applied apply VERB bracis-28377 50 26 to to ADP bracis-28377 50 27 six six NUM bracis-28377 50 28 images image NOUN bracis-28377 50 29 of of ADP bracis-28377 50 30 one one NUM bracis-28377 50 31 dog dog NOUN bracis-28377 50 32 from from ADP bracis-28377 50 33 the the DET bracis-28377 50 34 dogfacenet dogfacenet NOUN bracis-28377 50 35 dataset dataset NOUN bracis-28377 50 36 [ [ X bracis-28377 50 37 17 17 NUM bracis-28377 50 38 ] ] PUNCT bracis-28377 50 39 . . PUNCT bracis-28377 51 1 full full ADJ bracis-28377 51 2 size size NOUN bracis-28377 51 3 image image NOUN bracis-28377 51 4 dog dog NOUN bracis-28377 51 5 identification identification NOUN bracis-28377 51 6 through through ADP bracis-28377 51 7 facial facial ADJ bracis-28377 51 8 features feature NOUN bracis-28377 51 9 is be AUX bracis-28377 51 10 a a DET bracis-28377 51 11 challenging challenging ADJ bracis-28377 51 12 task task NOUN bracis-28377 51 13 , , PUNCT bracis-28377 51 14 given give VERB bracis-28377 51 15 the the DET bracis-28377 51 16 high high ADJ bracis-28377 51 17 interclass interclass NOUN bracis-28377 51 18 similarity similarity NOUN bracis-28377 51 19 and and CCONJ bracis-28377 51 20 the the DET bracis-28377 51 21 high high ADJ bracis-28377 51 22 intraclass intraclass NOUN bracis-28377 51 23 variability variability NOUN bracis-28377 51 24 that that PRON bracis-28377 51 25 can can AUX bracis-28377 51 26 occur occur VERB bracis-28377 51 27 . . PUNCT bracis-28377 52 1 figure figure VERB bracis-28377 52 2     SPACE bracis-28377 52 3 5 5 NUM bracis-28377 52 4 shows show NOUN bracis-28377 52 5 , , PUNCT bracis-28377 52 6 on on ADP bracis-28377 52 7 the the DET bracis-28377 52 8 left left ADJ bracis-28377 52 9 , , PUNCT bracis-28377 52 10 six six NUM bracis-28377 52 11 pairs pair NOUN bracis-28377 52 12 of of ADP bracis-28377 52 13 images image NOUN bracis-28377 52 14 captured capture VERB bracis-28377 52 15 from from ADP bracis-28377 52 16 distinct distinct ADJ bracis-28377 52 17 pairs pair NOUN bracis-28377 52 18 of of ADP bracis-28377 52 19 individuals individual NOUN bracis-28377 52 20 that that PRON bracis-28377 52 21 present present VERB bracis-28377 52 22 faces face VERB bracis-28377 52 23 with with ADP bracis-28377 52 24 high high ADJ bracis-28377 52 25 similarity similarity NOUN bracis-28377 52 26 ( ( PUNCT bracis-28377 52 27 false false ADJ bracis-28377 52 28 positive positive ADJ bracis-28377 52 29 pairs pair NOUN bracis-28377 52 30 of of ADP bracis-28377 52 31 canine canine NOUN bracis-28377 52 32 image image NOUN bracis-28377 52 33 faces face NOUN bracis-28377 52 34 according accord VERB bracis-28377 52 35 to to ADP bracis-28377 52 36 the the DET bracis-28377 52 37 method method NOUN bracis-28377 52 38 proposed propose VERB bracis-28377 52 39 by by ADP bracis-28377 52 40 mougeot mougeot PROPN bracis-28377 52 41 , , PUNCT bracis-28377 52 42 li li PROPN bracis-28377 52 43 and and CCONJ bracis-28377 52 44 jia jia PROPN bracis-28377 53 1 [ [ X bracis-28377 53 2 17 17 NUM bracis-28377 53 3 ] ] NUM bracis-28377 53 4 ) ) PUNCT bracis-28377 53 5 , , PUNCT bracis-28377 53 6 and and CCONJ bracis-28377 53 7 , , PUNCT bracis-28377 53 8 on on ADP bracis-28377 53 9 the the DET bracis-28377 53 10 right right NOUN bracis-28377 53 11 , , PUNCT bracis-28377 53 12 six six NUM bracis-28377 53 13 pairs pair NOUN bracis-28377 53 14 of of ADP bracis-28377 53 15 images image NOUN bracis-28377 53 16 , , PUNCT bracis-28377 53 17 each each DET bracis-28377 53 18 pair pair NOUN bracis-28377 53 19 from from ADP bracis-28377 53 20 the the DET bracis-28377 53 21 same same ADJ bracis-28377 53 22 individual individual NOUN bracis-28377 53 23 , , PUNCT bracis-28377 53 24 that that PRON bracis-28377 53 25 present present VERB bracis-28377 53 26 high high ADJ bracis-28377 53 27 variability variability NOUN bracis-28377 53 28 ( ( PUNCT bracis-28377 53 29 false false ADJ bracis-28377 53 30 negative negative ADJ bracis-28377 53 31 pairs pair NOUN bracis-28377 53 32 of of ADP bracis-28377 53 33 canine canine NOUN bracis-28377 53 34 image image NOUN bracis-28377 53 35 faces face NOUN bracis-28377 53 36 according accord VERB bracis-28377 53 37 to to ADP bracis-28377 53 38 the the DET bracis-28377 53 39 method method NOUN bracis-28377 53 40 proposed propose VERB bracis-28377 53 41 by by ADP bracis-28377 53 42 mougeot mougeot PROPN bracis-28377 53 43 , , PUNCT bracis-28377 53 44 li li PROPN bracis-28377 53 45 and and CCONJ bracis-28377 53 46 jia jia PROPN bracis-28377 54 1 [ [ X bracis-28377 54 2 17 17 NUM bracis-28377 54 3 ] ] PUNCT bracis-28377 54 4 ) ) PUNCT bracis-28377 54 5 . . PUNCT bracis-28377 55 1 fig fig NOUN bracis-28377 55 2 . . PUNCT bracis-28377 56 1 5 5 X bracis-28377 56 2 . . X bracis-28377 56 3 examples example NOUN bracis-28377 56 4 of of ADP bracis-28377 56 5 six six NUM bracis-28377 56 6 false false ADJ bracis-28377 56 7 positive positive ADJ bracis-28377 56 8 pairs pair NOUN bracis-28377 56 9 of of ADP bracis-28377 56 10 canine canine NOUN bracis-28377 56 11 faces face NOUN bracis-28377 56 12 ( ( PUNCT bracis-28377 56 13 left leave VERB bracis-28377 56 14 ) ) PUNCT bracis-28377 56 15 and and CCONJ bracis-28377 56 16 six six NUM bracis-28377 56 17 false false ADJ bracis-28377 56 18 negative negative ADJ bracis-28377 56 19 pairs pair NOUN bracis-28377 56 20 of of ADP bracis-28377 56 21 canine canine NOUN bracis-28377 56 22 faces face NOUN bracis-28377 56 23 ( ( PUNCT bracis-28377 56 24 right right ADJ bracis-28377 56 25 ) ) PUNCT bracis-28377 56 26 , , PUNCT bracis-28377 56 27 according accord VERB bracis-28377 56 28 to to ADP bracis-28377 56 29 the the DET bracis-28377 56 30 method method NOUN bracis-28377 56 31 proposed propose VERB bracis-28377 56 32 by by ADP bracis-28377 56 33 mougeot mougeot PROPN bracis-28377 56 34 , , PUNCT bracis-28377 56 35 li li PROPN bracis-28377 56 36 and and CCONJ bracis-28377 56 37 jia jia PROPN bracis-28377 57 1 [ [ X bracis-28377 57 2 17 17 NUM bracis-28377 57 3 ] ] PUNCT bracis-28377 57 4 . . PUNCT bracis-28377 58 1 full full ADJ bracis-28377 58 2 size size NOUN bracis-28377 58 3 image image NOUN bracis-28377 58 4 mougeot mougeot NOUN bracis-28377 58 5 , , PUNCT bracis-28377 58 6 li li PROPN bracis-28377 58 7 and and CCONJ bracis-28377 58 8 jia jia PROPN bracis-28377 59 1 [ [ X bracis-28377 59 2 17 17 NUM bracis-28377 59 3 ] ] PUNCT bracis-28377 59 4 proposed propose VERB bracis-28377 59 5 a a DET bracis-28377 59 6 deep deep ADJ bracis-28377 59 7 vgg vgg NOUN bracis-28377 59 8 - - PUNCT bracis-28377 59 9 like like NOUN bracis-28377 59 10 and and CCONJ bracis-28377 59 11 a a DET bracis-28377 59 12 resnet resnet NOUN bracis-28377 59 13 - - PUNCT bracis-28377 59 14 like like ADJ bracis-28377 59 15 convolutional convolutional ADJ bracis-28377 59 16 neural neural ADJ bracis-28377 59 17 network network NOUN bracis-28377 59 18 ( ( PUNCT bracis-28377 59 19 cnn cnn PROPN bracis-28377 59 20 ) ) PUNCT bracis-28377 59 21 for for ADP bracis-28377 59 22 the the DET bracis-28377 59 23 dog dog NOUN bracis-28377 59 24 face face NOUN bracis-28377 59 25 verification verification NOUN bracis-28377 59 26 task task NOUN bracis-28377 59 27 containing contain VERB bracis-28377 59 28 48 48 NUM bracis-28377 59 29 dogs dog NOUN bracis-28377 59 30 from from ADP bracis-28377 59 31 an an DET bracis-28377 59 32 open open NOUN bracis-28377 59 33 - - PUNCT bracis-28377 59 34 set set NOUN bracis-28377 59 35 of of ADP bracis-28377 59 36 test test NOUN bracis-28377 59 37 data datum NOUN bracis-28377 59 38 of of ADP bracis-28377 59 39 dogfacenet dogfacenet NOUN bracis-28377 59 40 , , PUNCT bracis-28377 59 41 and and CCONJ bracis-28377 59 42 concluded conclude VERB bracis-28377 59 43 that that SCONJ bracis-28377 59 44 the the DET bracis-28377 59 45 resnet resnet NOUN bracis-28377 59 46 - - PUNCT bracis-28377 59 47 like like ADJ bracis-28377 59 48 cnn cnn PROPN bracis-28377 59 49 performed perform VERB bracis-28377 59 50 better well ADV bracis-28377 59 51 with with ADP bracis-28377 59 52 92 92 NUM bracis-28377 59 53 % % NOUN bracis-28377 59 54 of of ADP bracis-28377 59 55 accuracy accuracy NOUN bracis-28377 59 56 . . PUNCT bracis-28377 60 1 for for ADP bracis-28377 60 2 the the DET bracis-28377 60 3 identification identification NOUN bracis-28377 60 4 task task NOUN bracis-28377 60 5 , , PUNCT bracis-28377 60 6 the the DET bracis-28377 60 7 best good ADJ bracis-28377 60 8 performance performance NOUN bracis-28377 60 9 was be AUX bracis-28377 60 10 also also ADV bracis-28377 60 11 obtained obtain VERB bracis-28377 60 12 by by ADP bracis-28377 60 13 the the DET bracis-28377 60 14 resnet resnet NOUN bracis-28377 60 15 - - PUNCT bracis-28377 60 16 like like ADJ bracis-28377 60 17 cnn cnn PROPN bracis-28377 60 18 , , PUNCT bracis-28377 60 19 with with ADP bracis-28377 60 20 60.44 60.44 NUM bracis-28377 60 21 % % NOUN bracis-28377 60 22 of of ADP bracis-28377 60 23 accuracy accuracy NOUN bracis-28377 60 24 for for ADP bracis-28377 60 25 rank-1 rank-1 NUM bracis-28377 60 26 and and CCONJ bracis-28377 60 27 92 92 NUM bracis-28377 60 28 % % NOUN bracis-28377 60 29 for for ADP bracis-28377 60 30 rank-5 rank-5 PROPN bracis-28377 60 31 . . PUNCT bracis-28377 61 1 yoon yoon PROPN bracis-28377 61 2 , , PUNCT bracis-28377 61 3 so so ADV bracis-28377 61 4 and and CCONJ bracis-28377 61 5 rhee rhee NOUN bracis-28377 62 1 [ [ X bracis-28377 62 2 21 21 NUM bracis-28377 62 3 ] ] PUNCT bracis-28377 62 4 , , PUNCT bracis-28377 62 5 based base VERB bracis-28377 62 6 on on ADP bracis-28377 62 7 the the DET bracis-28377 62 8 studies study NOUN bracis-28377 62 9 of of ADP bracis-28377 62 10 mougeot mougeot NOUN bracis-28377 62 11 , , PUNCT bracis-28377 62 12 li li PROPN bracis-28377 62 13 and and CCONJ bracis-28377 62 14 jia jia PROPN bracis-28377 63 1 [ [ X bracis-28377 63 2 17 17 NUM bracis-28377 63 3 ] ] PUNCT bracis-28377 63 4 , , PUNCT bracis-28377 63 5 proposed propose VERB bracis-28377 63 6 a a DET bracis-28377 63 7 methodology methodology NOUN bracis-28377 63 8 for for ADP bracis-28377 63 9 using use VERB bracis-28377 63 10 vector vector NOUN bracis-28377 63 11 spaces space NOUN bracis-28377 63 12 to to PART bracis-28377 63 13 improve improve VERB bracis-28377 63 14 the the DET bracis-28377 63 15 performance performance NOUN bracis-28377 63 16 in in ADP bracis-28377 63 17 the the DET bracis-28377 63 18 dog dog NOUN bracis-28377 63 19 identification identification NOUN bracis-28377 63 20 process process NOUN bracis-28377 63 21 . . PUNCT bracis-28377 64 1 thus thus ADV bracis-28377 64 2 , , PUNCT bracis-28377 64 3 in in ADP bracis-28377 64 4 the the DET bracis-28377 64 5 scenario scenario NOUN bracis-28377 64 6 of of ADP bracis-28377 64 7 dog dog NOUN bracis-28377 64 8 facial facial ADJ bracis-28377 64 9 identification identification NOUN bracis-28377 64 10 , , PUNCT bracis-28377 64 11 the the DET bracis-28377 64 12 authors author NOUN bracis-28377 64 13 proposed propose VERB bracis-28377 64 14 , , PUNCT bracis-28377 64 15 using use VERB bracis-28377 64 16 the the DET bracis-28377 64 17 triplet triplet NOUN bracis-28377 64 18 loss loss NOUN bracis-28377 64 19 , , PUNCT bracis-28377 64 20 the the DET bracis-28377 64 21 removal removal NOUN bracis-28377 64 22 of of ADP bracis-28377 64 23 the the DET bracis-28377 64 24 l2 l2 NOUN bracis-28377 64 25 norm norm NOUN bracis-28377 64 26 , , PUNCT bracis-28377 64 27 for for ADP bracis-28377 64 28 comparing compare VERB bracis-28377 64 29 feature feature NOUN bracis-28377 64 30 vectors vector NOUN bracis-28377 64 31 from from ADP bracis-28377 64 32 the the DET bracis-28377 64 33 images image NOUN bracis-28377 64 34 of of ADP bracis-28377 64 35 the the DET bracis-28377 64 36 dogs dog NOUN bracis-28377 64 37 in in ADP bracis-28377 64 38 a a DET bracis-28377 64 39 new new ADJ bracis-28377 64 40 space space NOUN bracis-28377 64 41 within within ADP bracis-28377 64 42 a a DET bracis-28377 64 43 sphere sphere NOUN bracis-28377 64 44 of of ADP bracis-28377 64 45 radius radius NOUN bracis-28377 64 46 1 1 NUM bracis-28377 64 47 . . PUNCT bracis-28377 65 1 with with ADP bracis-28377 65 2 this this PRON bracis-28377 65 3 , , PUNCT bracis-28377 65 4 they they PRON bracis-28377 65 5 obtained obtain VERB bracis-28377 65 6 a a DET bracis-28377 65 7 mean mean ADJ bracis-28377 65 8 accuracy accuracy NOUN bracis-28377 65 9 of of ADP bracis-28377 65 10 88.8 88.8 NUM bracis-28377 65 11 % % NOUN bracis-28377 65 12 with with ADP bracis-28377 65 13 arcface+vl arcface+vl PRON bracis-28377 65 14 in in ADP bracis-28377 65 15 the the DET bracis-28377 65 16 verification verification NOUN bracis-28377 65 17 task task NOUN bracis-28377 65 18 . . PROPN bracis-28377 66 1 3 3 NUM bracis-28377 66 2 background background NOUN bracis-28377 66 3 in in ADP bracis-28377 66 4 this this DET bracis-28377 66 5 section section NOUN bracis-28377 66 6 , , PUNCT bracis-28377 66 7 we we PRON bracis-28377 66 8 describe describe VERB bracis-28377 66 9 the the DET bracis-28377 66 10 fundamental fundamental ADJ bracis-28377 66 11 concepts concept NOUN bracis-28377 66 12 used use VERB bracis-28377 66 13 in in ADP bracis-28377 66 14 this this DET bracis-28377 66 15 work work NOUN bracis-28377 66 16 , , PUNCT bracis-28377 66 17 such such ADJ bracis-28377 66 18 as as ADP bracis-28377 66 19 convolutional convolutional ADJ bracis-28377 66 20 neural neural ADJ bracis-28377 66 21 network network NOUN bracis-28377 66 22 ( ( PUNCT bracis-28377 66 23 cnn cnn PROPN bracis-28377 66 24 ) ) PUNCT bracis-28377 66 25 , , PUNCT bracis-28377 66 26 resnet resnet NOUN bracis-28377 66 27 architecture architecture NOUN bracis-28377 66 28 , , PUNCT bracis-28377 66 29 vision vision NOUN bracis-28377 66 30 transformers transformer NOUN bracis-28377 66 31 ( ( PUNCT bracis-28377 66 32 vit vit NOUN bracis-28377 66 33 ) ) PUNCT bracis-28377 66 34 , , PUNCT bracis-28377 66 35 efficientformer efficientformer NOUN bracis-28377 66 36 - - PUNCT bracis-28377 66 37 l1 l1 NOUN bracis-28377 66 38 architecture architecture NOUN bracis-28377 66 39 and and CCONJ bracis-28377 66 40 arcface arcface NOUN bracis-28377 66 41 function function NOUN bracis-28377 66 42 . . PUNCT bracis-28377 67 1 3.1 3.1 NUM bracis-28377 67 2 convolutional convolutional ADJ bracis-28377 67 3 neural neural ADJ bracis-28377 67 4 networks network NOUN bracis-28377 67 5 and and CCONJ bracis-28377 67 6     SPACE bracis-28377 67 7 the the DET bracis-28377 67 8     SPACE bracis-28377 67 9 resnet resnet ADJ bracis-28377 67 10 architecture architecture NOUN bracis-28377 67 11 convolutional convolutional ADJ bracis-28377 67 12 neural neural ADJ bracis-28377 67 13 networks network NOUN bracis-28377 67 14 ( ( PUNCT bracis-28377 67 15 cnn cnn PROPN bracis-28377 67 16 ) ) PUNCT bracis-28377 67 17 have have AUX bracis-28377 67 18 been be AUX bracis-28377 67 19 widely widely ADV bracis-28377 67 20 used use VERB bracis-28377 67 21 in in ADP bracis-28377 67 22 solutions solution NOUN bracis-28377 67 23 for for ADP bracis-28377 67 24 computer computer NOUN bracis-28377 67 25 vision vision NOUN bracis-28377 67 26 applications application NOUN bracis-28377 67 27 , , PUNCT bracis-28377 67 28 in in ADP bracis-28377 67 29 problems problem NOUN bracis-28377 67 30 such such ADJ bracis-28377 67 31 as as ADP bracis-28377 67 32 object object NOUN bracis-28377 67 33 detection detection NOUN bracis-28377 67 34 , , PUNCT bracis-28377 67 35 image image NOUN bracis-28377 67 36 recognition recognition NOUN bracis-28377 67 37 , , PUNCT bracis-28377 67 38 and and CCONJ bracis-28377 67 39 biometrics biometric NOUN bracis-28377 68 1 [ [ X bracis-28377 68 2 1 1 NUM bracis-28377 68 3 , , PUNCT bracis-28377 68 4 2 2 NUM bracis-28377 68 5 ] ] PUNCT bracis-28377 68 6 . . PUNCT bracis-28377 69 1 their their PRON bracis-28377 69 2 architectures architecture NOUN bracis-28377 69 3 are be AUX bracis-28377 69 4 similar similar ADJ bracis-28377 69 5 to to ADP bracis-28377 69 6 the the DET bracis-28377 69 7 ones one NOUN bracis-28377 69 8 of of ADP bracis-28377 69 9 the the DET bracis-28377 69 10 human human ADJ bracis-28377 69 11 visual visual ADJ bracis-28377 69 12 cortex cortex NOUN bracis-28377 69 13 . . PUNCT bracis-28377 70 1 according accord VERB bracis-28377 70 2 to to ADP bracis-28377 70 3 chollet chollet NOUN bracis-28377 70 4 [ [ X bracis-28377 70 5 1 1 NUM bracis-28377 70 6 ] ] PUNCT bracis-28377 70 7 , , PUNCT bracis-28377 70 8 there there PRON bracis-28377 70 9 are be VERB bracis-28377 70 10 some some DET bracis-28377 70 11 high high ADJ bracis-28377 70 12 - - PUNCT bracis-28377 70 13 performance performance NOUN bracis-28377 70 14 cnn cnn PROPN bracis-28377 70 15 architectures architecture NOUN bracis-28377 70 16 , , PUNCT bracis-28377 70 17 such such ADJ bracis-28377 70 18 as as ADP bracis-28377 70 19 alexnet alexnet ADJ bracis-28377 70 20 , , PUNCT bracis-28377 70 21 vgg vgg NOUN bracis-28377 70 22 , , PUNCT bracis-28377 70 23 googlenet googlenet NOUN bracis-28377 70 24 and and CCONJ bracis-28377 70 25 resnet resnet NOUN bracis-28377 70 26 . . PUNCT bracis-28377 71 1 convolutional convolutional ADJ bracis-28377 71 2 neural neural ADJ bracis-28377 71 3 networks network NOUN bracis-28377 71 4 are be AUX bracis-28377 71 5 organized organize VERB bracis-28377 71 6 into into ADP bracis-28377 71 7 layers layer NOUN bracis-28377 71 8 , , PUNCT bracis-28377 71 9 such such ADJ bracis-28377 71 10 as as ADP bracis-28377 71 11 the the DET bracis-28377 71 12 convolutional convolutional ADJ bracis-28377 71 13 layers layer NOUN bracis-28377 71 14 , , PUNCT bracis-28377 71 15 the the DET bracis-28377 71 16 pooling pool VERB bracis-28377 71 17 layers layer NOUN bracis-28377 71 18 and and CCONJ bracis-28377 71 19 the the DET bracis-28377 71 20 fully fully ADV bracis-28377 71 21 connected connected ADJ bracis-28377 71 22 layer layer NOUN bracis-28377 71 23 . . PUNCT bracis-28377 72 1 in in ADP bracis-28377 72 2 fig fig NOUN bracis-28377 72 3 . . PUNCT bracis-28377 72 4     SPACE bracis-28377 72 5 6 6 NUM bracis-28377 72 6 there there PRON bracis-28377 72 7 is be VERB bracis-28377 72 8 an an DET bracis-28377 72 9 example example NOUN bracis-28377 72 10 of of ADP bracis-28377 72 11 the the DET bracis-28377 72 12 architecture architecture NOUN bracis-28377 72 13 of of ADP bracis-28377 72 14 a a DET bracis-28377 72 15 cnn cnn PROPN bracis-28377 72 16 . . PUNCT bracis-28377 73 1 fig fig NOUN bracis-28377 73 2 . . PUNCT bracis-28377 74 1 6 6 NUM bracis-28377 74 2 . . NUM bracis-28377 74 3 simplified simplify VERB bracis-28377 74 4 architecture architecture NOUN bracis-28377 74 5 of of ADP bracis-28377 74 6 a a DET bracis-28377 74 7 convolutional convolutional ADJ bracis-28377 74 8 neural neural ADJ bracis-28377 74 9 network network NOUN bracis-28377 74 10 [ [ X bracis-28377 74 11 13 13 NUM bracis-28377 74 12 ] ] PUNCT bracis-28377 74 13 . . PUNCT bracis-28377 75 1 full full ADJ bracis-28377 75 2 size size NOUN bracis-28377 75 3 image image NOUN bracis-28377 75 4 in in ADP bracis-28377 75 5 cnns cnn NOUN bracis-28377 75 6 , , PUNCT bracis-28377 75 7 to to PART bracis-28377 75 8 limit limit VERB bracis-28377 75 9 the the DET bracis-28377 75 10 range range NOUN bracis-28377 75 11 of of ADP bracis-28377 75 12 a a DET bracis-28377 75 13 result result NOUN bracis-28377 75 14 of of ADP bracis-28377 75 15 a a DET bracis-28377 75 16 convolution convolution NOUN bracis-28377 75 17 layer layer NOUN bracis-28377 75 18 , , PUNCT bracis-28377 75 19 activation activation NOUN bracis-28377 75 20 functions function NOUN bracis-28377 75 21 are be AUX bracis-28377 75 22 used use VERB bracis-28377 75 23 , , PUNCT bracis-28377 75 24 such such ADJ bracis-28377 75 25 as as ADP bracis-28377 75 26 the the DET bracis-28377 75 27 linear linear ADJ bracis-28377 75 28 function function NOUN bracis-28377 75 29 relu relu NOUN bracis-28377 75 30 , , PUNCT bracis-28377 75 31 the the DET bracis-28377 75 32 sigmoid sigmoid NOUN bracis-28377 75 33 function function NOUN bracis-28377 75 34 , , PUNCT bracis-28377 75 35 the the DET bracis-28377 75 36 hyperbolic hyperbolic ADJ bracis-28377 75 37 tangent tangent NOUN bracis-28377 75 38 function function NOUN bracis-28377 75 39 , , PUNCT bracis-28377 75 40 etc etc X bracis-28377 75 41 . . X bracis-28377 75 42 thus thus ADV bracis-28377 75 43 , , PUNCT bracis-28377 75 44 according accord VERB bracis-28377 75 45 to to ADP bracis-28377 75 46 lecun lecun PROPN bracis-28377 75 47 et et PROPN bracis-28377 75 48 al al PROPN bracis-28377 75 49 . . PUNCT bracis-28377 76 1 [ [ X bracis-28377 76 2 6 6 NUM bracis-28377 76 3 ] ] PUNCT bracis-28377 76 4 , , PUNCT bracis-28377 76 5 the the DET bracis-28377 76 6 convolutional convolutional ADJ bracis-28377 76 7 layers layer NOUN bracis-28377 76 8 have have VERB bracis-28377 76 9 the the DET bracis-28377 76 10 objective objective NOUN bracis-28377 76 11 of of ADP bracis-28377 76 12 performing perform VERB bracis-28377 76 13 the the DET bracis-28377 76 14 extraction extraction NOUN bracis-28377 76 15 of of ADP bracis-28377 76 16 features feature NOUN bracis-28377 76 17 through through ADP bracis-28377 76 18 the the DET bracis-28377 76 19 convolution convolution NOUN bracis-28377 76 20 operation operation NOUN bracis-28377 76 21 . . PUNCT bracis-28377 77 1 furthermore furthermore ADV bracis-28377 77 2 , , PUNCT bracis-28377 77 3 the the DET bracis-28377 77 4 adjustments adjustment NOUN bracis-28377 77 5 of of ADP bracis-28377 77 6 the the DET bracis-28377 77 7 backpropagation backpropagation NOUN bracis-28377 77 8 algorithm algorithm NOUN bracis-28377 77 9 weights weight NOUN bracis-28377 77 10 are be AUX bracis-28377 77 11 performed perform VERB bracis-28377 77 12 in in ADP bracis-28377 77 13 this this DET bracis-28377 77 14 layer layer NOUN bracis-28377 77 15 . . PUNCT bracis-28377 78 1 also also ADV bracis-28377 78 2 according accord VERB bracis-28377 78 3 to to ADP bracis-28377 78 4 lecun lecun PROPN bracis-28377 78 5 et et PROPN bracis-28377 78 6 al al PROPN bracis-28377 78 7 . . PUNCT bracis-28377 79 1 [ [ X bracis-28377 79 2 6 6 NUM bracis-28377 79 3 ] ] PUNCT bracis-28377 79 4 , , PUNCT bracis-28377 79 5 the the DET bracis-28377 79 6 pooling pool VERB bracis-28377 79 7 layer layer NOUN bracis-28377 79 8 aims aim VERB bracis-28377 79 9 to to PART bracis-28377 79 10 reduce reduce VERB bracis-28377 79 11 dimensionality dimensionality NOUN bracis-28377 79 12 by by ADP bracis-28377 79 13 combining combine VERB bracis-28377 79 14 neighbouring neighbouring ADJ bracis-28377 79 15 pixels pixel NOUN bracis-28377 79 16 into into ADP bracis-28377 79 17 a a DET bracis-28377 79 18 single single ADJ bracis-28377 79 19 value value NOUN bracis-28377 79 20 . . PUNCT bracis-28377 80 1 consequently consequently ADV bracis-28377 80 2 , , PUNCT bracis-28377 80 3 there there PRON bracis-28377 80 4 is be VERB bracis-28377 80 5 a a DET bracis-28377 80 6 reduction reduction NOUN bracis-28377 80 7 in in ADP bracis-28377 80 8 time time NOUN bracis-28377 80 9 and and CCONJ bracis-28377 80 10 a a DET bracis-28377 80 11 minimization minimization NOUN bracis-28377 80 12 of of ADP bracis-28377 80 13 the the DET bracis-28377 80 14 use use NOUN bracis-28377 80 15 of of ADP bracis-28377 80 16 computational computational ADJ bracis-28377 80 17 resources resource NOUN bracis-28377 80 18 for for ADP bracis-28377 80 19 training training NOUN bracis-28377 80 20 . . PUNCT bracis-28377 81 1 the the DET bracis-28377 81 2 fully fully ADV bracis-28377 81 3 connected connect VERB bracis-28377 81 4 layer layer NOUN bracis-28377 81 5 aims aim VERB bracis-28377 81 6 to to PART bracis-28377 81 7 perform perform VERB bracis-28377 81 8 image image NOUN bracis-28377 81 9 classification classification NOUN bracis-28377 81 10 . . PUNCT bracis-28377 82 1 for for ADP bracis-28377 82 2 example example NOUN bracis-28377 82 3 , , PUNCT bracis-28377 82 4 in in ADP bracis-28377 82 5 a a DET bracis-28377 82 6 neural neural ADJ bracis-28377 82 7 network network NOUN bracis-28377 82 8 for for ADP bracis-28377 82 9 classifying classify VERB bracis-28377 82 10 dogs dog NOUN bracis-28377 82 11 , , PUNCT bracis-28377 82 12 the the DET bracis-28377 82 13 result result NOUN bracis-28377 82 14 of of ADP bracis-28377 82 15 the the DET bracis-28377 82 16 fully fully ADV bracis-28377 82 17 connected connect VERB bracis-28377 82 18 layer layer NOUN bracis-28377 82 19 expresses express VERB bracis-28377 82 20 which which DET bracis-28377 82 21 dog dog NOUN bracis-28377 82 22 is be AUX bracis-28377 82 23 associated associate VERB bracis-28377 82 24 with with ADP bracis-28377 82 25 the the DET bracis-28377 82 26 input input NOUN bracis-28377 82 27 image image NOUN bracis-28377 82 28 . . PUNCT bracis-28377 83 1 in in ADP bracis-28377 83 2 addition addition NOUN bracis-28377 83 3 to to ADP bracis-28377 83 4 conventional conventional ADJ bracis-28377 83 5 cnn cnn PROPN bracis-28377 83 6 architectures architecture NOUN bracis-28377 83 7 , , PUNCT bracis-28377 83 8 resnet resnet NOUN bracis-28377 83 9 is be AUX bracis-28377 83 10 a a DET bracis-28377 83 11 cnn cnn NOUN bracis-28377 83 12 architecture architecture NOUN bracis-28377 83 13 proposed propose VERB bracis-28377 83 14 by by ADP bracis-28377 83 15 he he PRON bracis-28377 83 16 et et PROPN bracis-28377 83 17 al al PROPN bracis-28377 83 18 . . PUNCT bracis-28377 84 1 [ [ X bracis-28377 84 2 8 8 NUM bracis-28377 84 3 ] ] PUNCT bracis-28377 84 4 and and CCONJ bracis-28377 84 5 its its PRON bracis-28377 84 6 main main ADJ bracis-28377 84 7 characteristic characteristic NOUN bracis-28377 84 8 , , PUNCT bracis-28377 84 9 as as SCONJ bracis-28377 84 10 its its PRON bracis-28377 84 11 own own ADJ bracis-28377 84 12 name name NOUN bracis-28377 84 13 indicates indicate VERB bracis-28377 84 14 residual residual ADJ bracis-28377 84 15 networks network NOUN bracis-28377 84 16 is be AUX bracis-28377 84 17 the the DET bracis-28377 84 18 use use NOUN bracis-28377 84 19 of of ADP bracis-28377 84 20 shortcut shortcut NOUN bracis-28377 84 21 connections connection NOUN bracis-28377 84 22 , , PUNCT bracis-28377 84 23 i.e. i.e. X bracis-28377 84 24 , , PUNCT bracis-28377 84 25 residual residual ADJ bracis-28377 84 26 information information NOUN bracis-28377 84 27 is be AUX bracis-28377 84 28 passed pass VERB bracis-28377 84 29 from from ADP bracis-28377 84 30 previous previous ADJ bracis-28377 84 31 layers layer NOUN bracis-28377 84 32 to to ADP bracis-28377 84 33 posterior posterior ADJ bracis-28377 84 34 ones one NOUN bracis-28377 84 35 , , PUNCT bracis-28377 84 36 which which PRON bracis-28377 84 37 improves improve VERB bracis-28377 84 38 and and CCONJ bracis-28377 84 39 , , PUNCT bracis-28377 84 40 consequently consequently ADV bracis-28377 84 41 , , PUNCT bracis-28377 84 42 change change VERB bracis-28377 84 43 the the DET bracis-28377 84 44 training training NOUN bracis-28377 84 45 process process NOUN bracis-28377 84 46 ( ( PUNCT bracis-28377 84 47 avoiding avoid VERB bracis-28377 84 48 data datum NOUN bracis-28377 84 49 vanishing vanishing NOUN bracis-28377 84 50 or or CCONJ bracis-28377 84 51 attenuation attenuation NOUN bracis-28377 84 52 as as ADV bracis-28377 84 53 well well ADV bracis-28377 84 54 as as ADP bracis-28377 84 55 saving save VERB bracis-28377 84 56 computational computational ADJ bracis-28377 84 57 resources resource NOUN bracis-28377 84 58 , , PUNCT bracis-28377 84 59 for for ADP bracis-28377 84 60 instance instance NOUN bracis-28377 84 61 ) ) PUNCT bracis-28377 84 62 . . PUNCT bracis-28377 85 1 in in ADP bracis-28377 85 2 fig fig NOUN bracis-28377 85 3 . . PUNCT bracis-28377 85 4     SPACE bracis-28377 85 5 7 7 NUM bracis-28377 85 6 , , PUNCT bracis-28377 85 7 there there PRON bracis-28377 85 8 is be VERB bracis-28377 85 9 an an DET bracis-28377 85 10 example example NOUN bracis-28377 85 11 of of ADP bracis-28377 85 12 using use VERB bracis-28377 85 13 shortcut shortcut NOUN bracis-28377 85 14 connections connection NOUN bracis-28377 85 15 and and CCONJ bracis-28377 85 16 routing route VERB bracis-28377 85 17 to to ADP bracis-28377 85 18 the the DET bracis-28377 85 19 output output NOUN bracis-28377 85 20 layer layer NOUN bracis-28377 85 21 of of ADP bracis-28377 85 22 the the DET bracis-28377 85 23 neural neural ADJ bracis-28377 85 24 network network NOUN bracis-28377 85 25 in in ADP bracis-28377 85 26 order order NOUN bracis-28377 85 27 to to PART bracis-28377 85 28 optimize optimize VERB bracis-28377 85 29 its its PRON bracis-28377 85 30 training training NOUN bracis-28377 85 31 process process NOUN bracis-28377 85 32 . . PUNCT bracis-28377 86 1 fig fig NOUN bracis-28377 86 2 . . PUNCT bracis-28377 87 1 7 7 X bracis-28377 87 2 . . X bracis-28377 87 3 resnet resnet NOUN bracis-28377 87 4 features feature NOUN bracis-28377 87 5 [ [ X bracis-28377 87 6 7 7 NUM bracis-28377 87 7 ] ] PUNCT bracis-28377 87 8 . . PUNCT bracis-28377 88 1 full full ADJ bracis-28377 88 2 size size NOUN bracis-28377 88 3 image image NOUN bracis-28377 88 4 according accord VERB bracis-28377 88 5 to to ADP bracis-28377 88 6 li li PROPN bracis-28377 88 7 et et PROPN bracis-28377 88 8 al al PROPN bracis-28377 88 9 . . PUNCT bracis-28377 89 1 [ [ X bracis-28377 89 2 15 15 NUM bracis-28377 89 3 ] ] PUNCT bracis-28377 89 4 , , PUNCT bracis-28377 89 5 several several ADJ bracis-28377 89 6 improvements improvement NOUN bracis-28377 89 7 were be AUX bracis-28377 89 8 made make VERB bracis-28377 89 9 to to ADP bracis-28377 89 10 the the DET bracis-28377 89 11 resnet resnet ADJ bracis-28377 89 12 architecture architecture NOUN bracis-28377 89 13 to to PART bracis-28377 89 14 increase increase VERB bracis-28377 89 15 its its PRON bracis-28377 89 16 performance performance NOUN bracis-28377 89 17 . . PUNCT bracis-28377 90 1 among among ADP bracis-28377 90 2 them they PRON bracis-28377 90 3 , , PUNCT bracis-28377 90 4 zhang zhang PROPN bracis-28377 90 5 et et PROPN bracis-28377 90 6 al al PROPN bracis-28377 90 7 . . PUNCT bracis-28377 91 1 [ [ X bracis-28377 91 2 22 22 NUM bracis-28377 91 3 ] ] PUNCT bracis-28377 91 4 proposed propose VERB bracis-28377 91 5 multi multi ADJ bracis-28377 91 6 - - ADJ bracis-28377 91 7 level level ADJ bracis-28377 91 8 shortcut shortcut NOUN bracis-28377 91 9 connections connection NOUN bracis-28377 91 10 and and CCONJ bracis-28377 91 11 targ targ NOUN bracis-28377 91 12 , , PUNCT bracis-28377 91 13 almeida almeida PROPN bracis-28377 91 14 and and CCONJ bracis-28377 91 15 lyman lyman PROPN bracis-28377 91 16 [ [ X bracis-28377 91 17 19 19 NUM bracis-28377 91 18 ] ] PUNCT bracis-28377 91 19 proposed propose VERB bracis-28377 91 20 using use VERB bracis-28377 91 21 more more ADJ bracis-28377 91 22 convolution convolution NOUN bracis-28377 91 23 layers layer NOUN bracis-28377 91 24 and and CCONJ bracis-28377 91 25 data datum NOUN bracis-28377 91 26 flow flow NOUN bracis-28377 91 27 between between ADP bracis-28377 91 28 layers layer NOUN bracis-28377 91 29 , , PUNCT bracis-28377 91 30 among among ADP bracis-28377 91 31 others other NOUN bracis-28377 91 32 . . PUNCT bracis-28377 92 1 also also ADV bracis-28377 92 2 , , PUNCT bracis-28377 92 3 different different ADJ bracis-28377 92 4 benchmark benchmark NOUN bracis-28377 92 5 architectures architecture NOUN bracis-28377 92 6 of of ADP bracis-28377 92 7 resnet resnet ADJ bracis-28377 92 8 neural neural ADJ bracis-28377 92 9 networks network NOUN bracis-28377 92 10 were be AUX bracis-28377 92 11 proposed propose VERB bracis-28377 92 12 , , PUNCT bracis-28377 92 13 however however ADV bracis-28377 92 14 , , PUNCT bracis-28377 92 15 one one NUM bracis-28377 92 16 of of ADP bracis-28377 92 17 the the DET bracis-28377 92 18 most most ADV bracis-28377 92 19 used use VERB bracis-28377 92 20 and and CCONJ bracis-28377 92 21 effective effective ADJ bracis-28377 92 22 for for ADP bracis-28377 92 23 many many ADJ bracis-28377 92 24 computer computer NOUN bracis-28377 92 25 vision vision NOUN bracis-28377 92 26 tasks task NOUN bracis-28377 92 27 is be AUX bracis-28377 92 28 the the DET bracis-28377 92 29 resnet-50 resnet-50 PROPN bracis-28377 92 30 ( ( PUNCT bracis-28377 92 31 with with ADP bracis-28377 92 32 50 50 NUM bracis-28377 92 33 layers layer NOUN bracis-28377 92 34 ) ) PUNCT bracis-28377 92 35 . . PUNCT bracis-28377 93 1 3.2 3.2 NUM bracis-28377 93 2 vision vision NOUN bracis-28377 93 3 transformers transformer NOUN bracis-28377 93 4 and and CCONJ bracis-28377 93 5     SPACE bracis-28377 93 6 the the DET bracis-28377 93 7     SPACE bracis-28377 93 8 efficientformer efficientformer NOUN bracis-28377 93 9 architecture architecture NOUN bracis-28377 93 10 transformers transformer NOUN bracis-28377 93 11 were be AUX bracis-28377 93 12 proposed propose VERB bracis-28377 93 13 by by ADP bracis-28377 93 14 vaswani vaswani PROPN bracis-28377 93 15 et et PROPN bracis-28377 93 16 al al PROPN bracis-28377 93 17 . . PUNCT bracis-28377 94 1 [ [ X bracis-28377 94 2 20 20 NUM bracis-28377 94 3 ] ] PUNCT bracis-28377 94 4 to to PART bracis-28377 94 5 solve solve VERB bracis-28377 94 6 a a DET bracis-28377 94 7 problem problem NOUN bracis-28377 94 8 involving involve VERB bracis-28377 94 9 natural natural ADJ bracis-28377 94 10 language language NOUN bracis-28377 94 11 processing processing NOUN bracis-28377 94 12 ( ( PUNCT bracis-28377 94 13 nlp nlp NOUN bracis-28377 94 14 ) ) PUNCT bracis-28377 94 15 and and CCONJ bracis-28377 94 16 have have AUX bracis-28377 94 17 become become VERB bracis-28377 94 18 state state NOUN bracis-28377 94 19 - - PUNCT bracis-28377 94 20 of of ADP bracis-28377 94 21 - - PUNCT bracis-28377 94 22 the the DET bracis-28377 94 23 - - PUNCT bracis-28377 94 24 art art NOUN bracis-28377 94 25 for for ADP bracis-28377 94 26 many many ADJ bracis-28377 94 27 types type NOUN bracis-28377 94 28 of of ADP bracis-28377 94 29 applications application NOUN bracis-28377 94 30 . . PUNCT bracis-28377 95 1 for for ADP bracis-28377 95 2 image image NOUN bracis-28377 95 3 processing processing NOUN bracis-28377 95 4 applications application NOUN bracis-28377 95 5 , , PUNCT bracis-28377 95 6 dosovitskiy dosovitskiy NOUN bracis-28377 95 7 et et NOUN bracis-28377 95 8 al al PROPN bracis-28377 95 9 . . PUNCT bracis-28377 96 1 [ [ X bracis-28377 96 2 5 5 NUM bracis-28377 96 3 ] ] PUNCT bracis-28377 96 4 proposed propose VERB bracis-28377 96 5 the the DET bracis-28377 96 6 vision vision NOUN bracis-28377 96 7 transformers transformer NOUN bracis-28377 96 8 ( ( PUNCT bracis-28377 96 9 vit vit NOUN bracis-28377 96 10 ) ) PUNCT bracis-28377 96 11 models model NOUN bracis-28377 96 12 , , PUNCT bracis-28377 96 13 in in ADP bracis-28377 96 14 which which PRON bracis-28377 96 15 the the DET bracis-28377 96 16 input input NOUN bracis-28377 96 17 image image NOUN bracis-28377 96 18 is be AUX bracis-28377 96 19 divided divide VERB bracis-28377 96 20 into into ADP bracis-28377 96 21 patches patch NOUN bracis-28377 96 22 of of ADP bracis-28377 96 23 fixed fix VERB bracis-28377 96 24 size size NOUN bracis-28377 96 25 , , PUNCT bracis-28377 96 26 transforming transform VERB bracis-28377 96 27 it it PRON bracis-28377 96 28 into into ADP bracis-28377 96 29 an an DET bracis-28377 96 30 input input NOUN bracis-28377 96 31 sequence sequence NOUN bracis-28377 96 32 of of ADP bracis-28377 96 33 image image NOUN bracis-28377 96 34 patches patch NOUN bracis-28377 96 35 . . PUNCT bracis-28377 97 1 continuing continue VERB bracis-28377 97 2 through through ADP bracis-28377 97 3 the the DET bracis-28377 97 4 initial initial ADJ bracis-28377 97 5 patches patch NOUN bracis-28377 97 6 , , PUNCT bracis-28377 97 7 feature feature NOUN bracis-28377 97 8 vectors vector NOUN bracis-28377 97 9 are be AUX bracis-28377 97 10 projected project VERB bracis-28377 97 11 across across ADP bracis-28377 97 12 an an DET bracis-28377 97 13 initial initial ADJ bracis-28377 97 14 linear linear NOUN bracis-28377 97 15 layer layer NOUN bracis-28377 97 16 and and CCONJ bracis-28377 97 17 position position NOUN bracis-28377 97 18 references reference NOUN bracis-28377 97 19 are be AUX bracis-28377 97 20 added add VERB bracis-28377 97 21 directly directly ADV bracis-28377 97 22 to to ADP bracis-28377 97 23 a a DET bracis-28377 97 24 standard standard ADJ bracis-28377 97 25 transformer transformer NOUN bracis-28377 97 26 . . PUNCT bracis-28377 98 1 in in ADP bracis-28377 98 2 fig fig NOUN bracis-28377 98 3 . . PUNCT bracis-28377 98 4     SPACE bracis-28377 98 5 8 8 NUM bracis-28377 98 6 there there PRON bracis-28377 98 7 is be VERB bracis-28377 98 8 an an DET bracis-28377 98 9 example example NOUN bracis-28377 98 10 of of ADP bracis-28377 98 11 the the DET bracis-28377 98 12 architecture architecture NOUN bracis-28377 98 13 of of ADP bracis-28377 98 14 a a DET bracis-28377 98 15 vision vision NOUN bracis-28377 98 16 transformer transformer NOUN bracis-28377 98 17 and and CCONJ bracis-28377 98 18 its its PRON bracis-28377 98 19 operations operation NOUN bracis-28377 98 20 . . PUNCT bracis-28377 99 1 according accord VERB bracis-28377 99 2 to to ADP bracis-28377 99 3 [ [ X bracis-28377 99 4 5 5 NUM bracis-28377 99 5 ] ] PUNCT bracis-28377 99 6 , , PUNCT bracis-28377 99 7 in in ADP bracis-28377 99 8 the the DET bracis-28377 99 9 benchmarks benchmark NOUN bracis-28377 99 10 performed perform VERB bracis-28377 99 11 , , PUNCT bracis-28377 99 12 the the DET bracis-28377 99 13 vision vision NOUN bracis-28377 99 14 transformers transformer NOUN bracis-28377 99 15 architecture architecture NOUN bracis-28377 99 16 outperformed outperform VERB bracis-28377 99 17 state state NOUN bracis-28377 99 18 - - PUNCT bracis-28377 99 19 of of ADP bracis-28377 99 20 - - PUNCT bracis-28377 99 21 the the DET bracis-28377 99 22 - - PUNCT bracis-28377 99 23 art art NOUN bracis-28377 99 24 image image NOUN bracis-28377 99 25 recognition recognition NOUN bracis-28377 99 26 architectures architecture NOUN bracis-28377 99 27 , , PUNCT bracis-28377 99 28 specifically specifically ADV bracis-28377 99 29 with with ADP bracis-28377 99 30 the the DET bracis-28377 99 31 use use NOUN bracis-28377 99 32 of of ADP bracis-28377 99 33 convolutional convolutional ADJ bracis-28377 99 34 neural neural ADJ bracis-28377 99 35 networks network NOUN bracis-28377 99 36 . . PUNCT bracis-28377 100 1 for for ADP bracis-28377 100 2 example example NOUN bracis-28377 100 3 , , PUNCT bracis-28377 100 4 in in ADP bracis-28377 100 5 the the DET bracis-28377 100 6 imagenet imagenet NOUN bracis-28377 100 7 database database NOUN bracis-28377 100 8 , , PUNCT bracis-28377 100 9 the the DET bracis-28377 100 10 vision vision NOUN bracis-28377 100 11 transformer transformer NOUN bracis-28377 100 12 architecture architecture NOUN bracis-28377 100 13 had have VERB bracis-28377 100 14 an an DET bracis-28377 100 15 accuracy accuracy NOUN bracis-28377 100 16 of of ADP bracis-28377 100 17 88.5 88.5 NUM bracis-28377 100 18 % % NOUN bracis-28377 100 19 , , PUNCT bracis-28377 100 20 and and CCONJ bracis-28377 100 21 the the DET bracis-28377 100 22 others other NOUN bracis-28377 100 23 , , PUNCT bracis-28377 100 24 using use VERB bracis-28377 100 25 cnns cnn NOUN bracis-28377 100 26 , , PUNCT bracis-28377 100 27 were be AUX bracis-28377 100 28 below below ADP bracis-28377 100 29 88 88 NUM bracis-28377 100 30 % % NOUN bracis-28377 100 31 . . PUNCT bracis-28377 101 1 fig fig NOUN bracis-28377 101 2 . . PUNCT bracis-28377 102 1 8 8 NUM bracis-28377 102 2 . . PUNCT bracis-28377 102 3 vision vision NOUN bracis-28377 102 4 transformer transformer NOUN bracis-28377 102 5 architecture architecture NOUN bracis-28377 102 6 proposed propose VERB bracis-28377 102 7 by by ADP bracis-28377 102 8 dosovitskiy dosovitskiy NOUN bracis-28377 103 1 [ [ X bracis-28377 103 2 5 5 NUM bracis-28377 103 3 ] ] PUNCT bracis-28377 103 4 . . PUNCT bracis-28377 104 1 full full ADJ bracis-28377 104 2 size size NOUN bracis-28377 104 3 image image NOUN bracis-28377 104 4 even even ADV bracis-28377 104 5 with with ADP bracis-28377 104 6 the the DET bracis-28377 104 7 great great ADJ bracis-28377 104 8 progress progress NOUN bracis-28377 104 9 caused cause VERB bracis-28377 104 10 by by ADP bracis-28377 104 11 vits vit NOUN bracis-28377 104 12 in in ADP bracis-28377 104 13 the the DET bracis-28377 104 14 computer computer NOUN bracis-28377 104 15 vision vision NOUN bracis-28377 104 16 area area NOUN bracis-28377 104 17 , , PUNCT bracis-28377 104 18 models model NOUN bracis-28377 104 19 based base VERB bracis-28377 104 20 on on ADP bracis-28377 104 21 transformers transformer NOUN bracis-28377 104 22 are be AUX bracis-28377 104 23 still still ADV bracis-28377 104 24 very very ADV bracis-28377 104 25 heavy heavy ADJ bracis-28377 104 26 and and CCONJ bracis-28377 104 27 present present VERB bracis-28377 104 28 a a DET bracis-28377 104 29 problem problem NOUN bracis-28377 104 30 of of ADP bracis-28377 104 31 high high ADJ bracis-28377 104 32 latency latency NOUN bracis-28377 104 33 , , PUNCT bracis-28377 104 34 which which PRON bracis-28377 104 35 hinders hinder VERB bracis-28377 104 36 both both CCONJ bracis-28377 104 37 the the DET bracis-28377 104 38 learning learning NOUN bracis-28377 104 39 and and CCONJ bracis-28377 104 40 the the DET bracis-28377 104 41 use use NOUN bracis-28377 104 42 of of ADP bracis-28377 104 43 this this DET bracis-28377 104 44 type type NOUN bracis-28377 104 45 of of ADP bracis-28377 104 46 model model NOUN bracis-28377 104 47 in in ADP bracis-28377 104 48 less less ADV bracis-28377 104 49 capable capable ADJ bracis-28377 104 50 environments environment NOUN bracis-28377 104 51 . . PUNCT bracis-28377 105 1 efficientformer efficientformer NOUN bracis-28377 105 2 - - PUNCT bracis-28377 105 3 type type NOUN bracis-28377 105 4 networks network NOUN bracis-28377 105 5 arise arise VERB bracis-28377 105 6 to to PART bracis-28377 105 7 mitigate mitigate VERB bracis-28377 105 8 this this DET bracis-28377 105 9 problem problem NOUN bracis-28377 105 10 through through ADP bracis-28377 105 11 a a DET bracis-28377 105 12 pure pure ADJ bracis-28377 105 13 transformer transformer NOUN bracis-28377 105 14 , , PUNCT bracis-28377 105 15 with with ADP bracis-28377 105 16 consistent consistent ADJ bracis-28377 105 17 data datum NOUN bracis-28377 105 18 dimensionality dimensionality NOUN bracis-28377 105 19 , , PUNCT bracis-28377 105 20 capable capable ADJ bracis-28377 105 21 of of ADP bracis-28377 105 22 running run VERB bracis-28377 105 23 in in ADP bracis-28377 105 24 more more ADV bracis-28377 105 25 diverse diverse ADJ bracis-28377 105 26 environments environment NOUN bracis-28377 105 27 . . PUNCT bracis-28377 106 1 figure figure NOUN bracis-28377 106 2     SPACE bracis-28377 106 3 9 9 NUM bracis-28377 106 4 shows show VERB bracis-28377 106 5 an an DET bracis-28377 106 6 overview overview NOUN bracis-28377 106 7 of of ADP bracis-28377 106 8 the the DET bracis-28377 106 9 efficentformer efficentformer NOUN bracis-28377 106 10 proposed propose VERB bracis-28377 106 11 by by ADP bracis-28377 106 12 lie lie NOUN bracis-28377 106 13 et et PROPN bracis-28377 106 14 al al PROPN bracis-28377 106 15 . . PUNCT bracis-28377 106 16     SPACE bracis-28377 107 1 [ [ X bracis-28377 107 2 16 16 NUM bracis-28377 107 3 ] ] PUNCT bracis-28377 107 4 . . PUNCT bracis-28377 108 1 the the DET bracis-28377 108 2 network network NOUN bracis-28377 108 3 starts start VERB bracis-28377 108 4 with with ADP bracis-28377 108 5 a a DET bracis-28377 108 6 convolution convolution NOUN bracis-28377 108 7 stem stem NOUN bracis-28377 108 8 as as ADP bracis-28377 108 9 patch patch NOUN bracis-28377 108 10 embedding embed VERB bracis-28377 108 11 , , PUNCT bracis-28377 108 12 followed follow VERB bracis-28377 108 13 by by ADP bracis-28377 108 14 metablock metablock NOUN bracis-28377 108 15 ( ( PUNCT bracis-28377 108 16 mb mb NOUN bracis-28377 108 17 ) ) PUNCT bracis-28377 108 18 . . PUNCT bracis-28377 109 1 the the DET bracis-28377 109 2 \(mb^{4d}\ \(mb^{4d}\ NOUN bracis-28377 109 3 ) ) PUNCT bracis-28377 109 4 and and CCONJ bracis-28377 109 5 \(mb^{3d}\ \(mb^{3d}\ NOUN bracis-28377 109 6 ) ) PUNCT bracis-28377 109 7 contain contain VERB bracis-28377 109 8 different different ADJ bracis-28377 109 9 token token ADJ bracis-28377 109 10 mixer mixer NOUN bracis-28377 109 11 configurations configuration NOUN bracis-28377 109 12 , , PUNCT bracis-28377 109 13 that that ADV bracis-28377 109 14 is is ADV bracis-28377 109 15 , , PUNCT bracis-28377 109 16 local local ADJ bracis-28377 109 17 pooling pooling NOUN bracis-28377 109 18 or or CCONJ bracis-28377 109 19 global global ADJ bracis-28377 109 20 multi multi ADJ bracis-28377 109 21 - - ADJ bracis-28377 109 22 head head ADJ bracis-28377 109 23 self self NOUN bracis-28377 109 24 - - PUNCT bracis-28377 109 25 attention attention NOUN bracis-28377 109 26 , , PUNCT bracis-28377 109 27 arranged arrange VERB bracis-28377 109 28 in in ADP bracis-28377 109 29 a a DET bracis-28377 109 30 dimension dimension NOUN bracis-28377 109 31 - - PUNCT bracis-28377 109 32 consistent consistent ADJ bracis-28377 109 33 manner manner NOUN bracis-28377 109 34 . . PUNCT bracis-28377 110 1 fig fig NOUN bracis-28377 110 2 . . PUNCT bracis-28377 111 1 9 9 X bracis-28377 111 2 . . X bracis-28377 111 3 overview overview NOUN bracis-28377 111 4 of of ADP bracis-28377 111 5 the the DET bracis-28377 111 6 efficientformer efficientformer NOUN bracis-28377 111 7     SPACE bracis-28377 112 1 [ [ X bracis-28377 112 2 16 16 NUM bracis-28377 112 3 ] ] PUNCT bracis-28377 112 4 . . PUNCT bracis-28377 113 1 full full ADJ bracis-28377 113 2 size size NOUN bracis-28377 113 3 image image NOUN bracis-28377 113 4 3.3 3.3 NUM bracis-28377 113 5 arcface arcface NOUN bracis-28377 113 6 in in ADP bracis-28377 113 7 the the DET bracis-28377 113 8 arcface arcface NOUN bracis-28377 113 9     SPACE bracis-28377 113 10 [ [ X bracis-28377 113 11 4 4 NUM bracis-28377 113 12 ] ] PUNCT bracis-28377 113 13 error error NOUN bracis-28377 113 14 function function NOUN bracis-28377 113 15 , , PUNCT bracis-28377 113 16 the the DET bracis-28377 113 17 objective objective NOUN bracis-28377 113 18 of of ADP bracis-28377 113 19 learning learning NOUN bracis-28377 113 20 becomes becomes AUX bracis-28377 113 21 jointly jointly ADV bracis-28377 113 22 minimizing minimize VERB bracis-28377 113 23 the the DET bracis-28377 113 24 angular angular ADJ bracis-28377 113 25 distance distance NOUN bracis-28377 113 26 between between ADP bracis-28377 113 27 elements element NOUN bracis-28377 113 28 of of ADP bracis-28377 113 29 the the DET bracis-28377 113 30 same same ADJ bracis-28377 113 31 class class NOUN bracis-28377 113 32 while while SCONJ bracis-28377 113 33 maximizing maximize VERB bracis-28377 113 34 the the DET bracis-28377 113 35 intra intra ADJ bracis-28377 113 36 - - ADJ bracis-28377 113 37 class class ADJ bracis-28377 113 38 similarity similarity NOUN bracis-28377 113 39 through through ADP bracis-28377 113 40 eq eq PROPN bracis-28377 113 41 . . PUNCT bracis-28377 113 42     SPACE bracis-28377 113 43 1 1 NUM bracis-28377 113 44 . . PUNCT bracis-28377 113 45 $ $ SYM bracis-28377 113 46 $ $ SYM bracis-28377 113 47 \begin{aligned \begin{aligne VERB bracis-28377 113 48 } } PUNCT bracis-28377 113 49 l l NOUN bracis-28377 113 50 = = SYM bracis-28377 113 51 -\frac{1}{n -\frac{1}{n ADJ bracis-28377 113 52 } } PUNCT bracis-28377 113 53 \sum \sum NOUN bracis-28377 113 54 \limits \limit NOUN bracis-28377 113 55 _ _ PRON bracis-28377 113 56 { { PUNCT bracis-28377 113 57 i=1}^{n i=1}^{n ADJ bracis-28377 113 58 } } PUNCT bracis-28377 113 59 log log NOUN bracis-28377 113 60 \frac{e^{s(\cos \frac{e^{s(\cos PUNCT bracis-28377 113 61 ( ( PUNCT bracis-28377 113 62 { { PUNCT bracis-28377 113 63 \theta \theta PROPN bracis-28377 113 64 _ _ PROPN bracis-28377 113 65 { { PUNCT bracis-28377 113 66 y_i y_i NUM bracis-28377 113 67 } } PUNCT bracis-28377 113 68 } } PUNCT bracis-28377 114 1 + + NUM bracis-28377 114 2 m))}}{e^{s(\cos m))}}{e^{s(\co NOUN bracis-28377 114 3 ( ( PUNCT bracis-28377 114 4 { { PUNCT bracis-28377 114 5 \theta \theta PROPN bracis-28377 114 6 _ _ PROPN bracis-28377 114 7 { { PUNCT bracis-28377 114 8 y_i y_i NUM bracis-28377 114 9 } } PUNCT bracis-28377 114 10 } } PUNCT bracis-28377 114 11 + + NUM bracis-28377 114 12 m m NOUN bracis-28377 114 13 ) ) PUNCT bracis-28377 114 14 ) ) PUNCT bracis-28377 114 15 } } PUNCT bracis-28377 115 1 + + ADJ bracis-28377 115 2 \sum \sum NOUN bracis-28377 115 3 _ _ PUNCT bracis-28377 115 4 { { PUNCT bracis-28377 115 5 j=1,j\ne j=1,j\ne PROPN bracis-28377 115 6 y_i}^n y_i}^n PROPN bracis-28377 115 7 e^{s e^{s NOUN bracis-28377 115 8 \cos \cos PROPN bracis-28377 115 9 ( ( PUNCT bracis-28377 115 10 \theta \theta PROPN bracis-28377 115 11 _ _ PROPN bracis-28377 115 12 j j PROPN bracis-28377 115 13 ) ) PUNCT bracis-28377 115 14 } } PUNCT bracis-28377 115 15 } } PUNCT bracis-28377 115 16 \end{aligned}$$ \end{aligned}$$ X bracis-28377 115 17 ( ( PUNCT bracis-28377 115 18 1 1 NUM bracis-28377 115 19 ) ) PUNCT bracis-28377 115 20 in in ADP bracis-28377 115 21 this this DET bracis-28377 115 22 case case NOUN bracis-28377 115 23 , , PUNCT bracis-28377 115 24 the the DET bracis-28377 115 25 class class NOUN bracis-28377 115 26 samples sample NOUN bracis-28377 115 27 are be AUX bracis-28377 115 28 positioned position VERB bracis-28377 115 29 in in ADP bracis-28377 115 30 a a DET bracis-28377 115 31 hypersphere hypersphere NOUN bracis-28377 115 32 of of ADP bracis-28377 115 33 radius radius NOUN bracis-28377 115 34 s s PROPN bracis-28377 115 35 , , PUNCT bracis-28377 115 36 and and CCONJ bracis-28377 115 37 the the DET bracis-28377 115 38 objective objective NOUN bracis-28377 115 39 of of ADP bracis-28377 115 40 the the DET bracis-28377 115 41 error error NOUN bracis-28377 115 42 function function NOUN bracis-28377 115 43 becomes become VERB bracis-28377 115 44 to to PART bracis-28377 115 45 minimize minimize VERB bracis-28377 115 46 the the DET bracis-28377 115 47 angular angular ADJ bracis-28377 115 48 distance distance NOUN bracis-28377 115 49 between between ADP bracis-28377 115 50 elements element NOUN bracis-28377 115 51 of of ADP bracis-28377 115 52 this this DET bracis-28377 115 53 hypersphere hypersphere NOUN bracis-28377 115 54 , , PUNCT bracis-28377 115 55 in in ADP bracis-28377 115 56 such such DET bracis-28377 115 57 a a DET bracis-28377 115 58 way way NOUN bracis-28377 115 59 that that PRON bracis-28377 115 60 the the DET bracis-28377 115 61 features feature NOUN bracis-28377 115 62 extracted extract VERB bracis-28377 115 63 in in ADP bracis-28377 115 64 the the DET bracis-28377 115 65 last last ADJ bracis-28377 115 66 layer layer NOUN bracis-28377 115 67 before before ADP bracis-28377 115 68 the the DET bracis-28377 115 69 softmax softmax NOUN bracis-28377 115 70 , , PUNCT bracis-28377 115 71 represent represent VERB bracis-28377 115 72 a a DET bracis-28377 115 73 sample sample NOUN bracis-28377 115 74 in in ADP bracis-28377 115 75 this this DET bracis-28377 115 76 space space NOUN bracis-28377 115 77 , , PUNCT bracis-28377 115 78 whose whose DET bracis-28377 115 79 elements element NOUN bracis-28377 115 80 similar similar ADJ bracis-28377 115 81 to to ADP bracis-28377 115 82 this this DET bracis-28377 115 83 sample sample NOUN bracis-28377 115 84 always always ADV bracis-28377 115 85 end end VERB bracis-28377 115 86 up up ADP bracis-28377 115 87 being be AUX bracis-28377 115 88 positioned position VERB bracis-28377 115 89 at at ADP bracis-28377 115 90 an an DET bracis-28377 115 91 angular angular ADJ bracis-28377 115 92 distance distance NOUN bracis-28377 115 93 in in ADP bracis-28377 115 94 the the DET bracis-28377 115 95 interval interval NOUN bracis-28377 115 96 m. m. NOUN bracis-28377 115 97 the the DET bracis-28377 115 98 error error NOUN bracis-28377 115 99 function function NOUN bracis-28377 115 100 can can AUX bracis-28377 115 101 be be AUX bracis-28377 115 102 extended extend VERB bracis-28377 115 103 to to PART bracis-28377 115 104 perform perform VERB bracis-28377 115 105 an an DET bracis-28377 115 106 analysis analysis NOUN bracis-28377 115 107 on on ADP bracis-28377 115 108 subspaces subspace NOUN bracis-28377 115 109 by by ADP bracis-28377 115 110 adding add VERB bracis-28377 115 111 an an DET bracis-28377 115 112 additional additional ADJ bracis-28377 115 113 parameter parameter NOUN bracis-28377 115 114 of of ADP bracis-28377 115 115 subcenter subcenter NOUN bracis-28377 115 116 when when SCONJ bracis-28377 115 117 extracting extract VERB bracis-28377 115 118 features feature NOUN bracis-28377 115 119 , , PUNCT bracis-28377 115 120 this this DET bracis-28377 115 121 parameter parameter NOUN bracis-28377 115 122 consists consist VERB bracis-28377 115 123 of of ADP bracis-28377 115 124 a a DET bracis-28377 115 125 matrix matrix NOUN bracis-28377 115 126 of of ADP bracis-28377 115 127 weights weight NOUN bracis-28377 115 128 of of ADP bracis-28377 115 129 dimensions dimension NOUN bracis-28377 115 130 \(c \(c NOUN bracis-28377 115 131 \times \time NOUN bracis-28377 115 132 k\ k\ PROPN bracis-28377 115 133 ) ) PUNCT bracis-28377 115 134 , , PUNCT bracis-28377 115 135 where where SCONJ bracis-28377 115 136 c c PROPN bracis-28377 115 137 is be AUX bracis-28377 115 138 the the DET bracis-28377 115 139 number number NOUN bracis-28377 115 140 of of ADP bracis-28377 115 141 classes class NOUN bracis-28377 115 142 and and CCONJ bracis-28377 115 143 k k NOUN bracis-28377 115 144 the the DET bracis-28377 115 145 number number NOUN bracis-28377 115 146 of of ADP bracis-28377 115 147 subspaces subspace NOUN bracis-28377 115 148 , , PUNCT bracis-28377 115 149 such such ADJ bracis-28377 115 150 that that SCONJ bracis-28377 115 151 the the DET bracis-28377 115 152 features feature NOUN bracis-28377 115 153 are be AUX bracis-28377 115 154 mapped map VERB bracis-28377 115 155 to to ADP bracis-28377 115 156 the the DET bracis-28377 115 157 k k PROPN bracis-28377 115 158 subspaces subspace NOUN bracis-28377 115 159 , , PUNCT bracis-28377 115 160 so so SCONJ bracis-28377 115 161 that that SCONJ bracis-28377 115 162 the the DET bracis-28377 115 163 arcface arcface NOUN bracis-28377 115 164 error error NOUN bracis-28377 115 165 function function NOUN bracis-28377 115 166 can can AUX bracis-28377 115 167 be be AUX bracis-28377 115 168 applied apply VERB bracis-28377 115 169 to to ADP bracis-28377 115 170 each each PRON bracis-28377 115 171 of of ADP bracis-28377 115 172 them they PRON bracis-28377 115 173     SPACE bracis-28377 116 1 [ [ X bracis-28377 116 2 3 3 NUM bracis-28377 116 3 ] ] PUNCT bracis-28377 116 4 . . PUNCT bracis-28377 117 1 4 4 NUM bracis-28377 117 2 proposed propose VERB bracis-28377 117 3 method method NOUN bracis-28377 117 4 in in ADP bracis-28377 117 5 order order NOUN bracis-28377 117 6 to to PART bracis-28377 117 7 improve improve VERB bracis-28377 117 8 the the DET bracis-28377 117 9 state state NOUN bracis-28377 117 10 - - PUNCT bracis-28377 117 11 of of ADP bracis-28377 117 12 - - PUNCT bracis-28377 117 13 the the DET bracis-28377 117 14 - - PUNCT bracis-28377 117 15 art art NOUN bracis-28377 117 16 results result NOUN bracis-28377 117 17 in in ADP bracis-28377 117 18 the the DET bracis-28377 117 19 task task NOUN bracis-28377 117 20 of of ADP bracis-28377 117 21 dog dog NOUN bracis-28377 117 22 facial facial ADJ bracis-28377 117 23 recognition recognition NOUN bracis-28377 117 24 , , PUNCT bracis-28377 117 25 in in ADP bracis-28377 117 26 this this DET bracis-28377 117 27 work work NOUN bracis-28377 117 28 we we PRON bracis-28377 117 29 propose propose VERB bracis-28377 117 30 a a DET bracis-28377 117 31 method method NOUN bracis-28377 117 32 that that PRON bracis-28377 117 33 uses use VERB bracis-28377 117 34 efficientformer efficientformer NOUN bracis-28377 117 35 - - PUNCT bracis-28377 117 36 l1 l1 PROPN bracis-28377 117 37 , , PUNCT bracis-28377 117 38 a a DET bracis-28377 117 39 vision vision NOUN bracis-28377 117 40 transformer transformer NOUN bracis-28377 117 41 architecture architecture NOUN bracis-28377 117 42 proposed propose VERB bracis-28377 117 43 by by ADP bracis-28377 117 44 lie lie NOUN bracis-28377 117 45 et et PROPN bracis-28377 117 46 al al PROPN bracis-28377 117 47 . . PUNCT bracis-28377 117 48     SPACE bracis-28377 118 1 [ [ X bracis-28377 118 2 16 16 NUM bracis-28377 118 3 ] ] PUNCT bracis-28377 118 4 . . PUNCT bracis-28377 119 1 in in ADP bracis-28377 119 2 the the DET bracis-28377 119 3 proposed propose VERB bracis-28377 119 4 method method NOUN bracis-28377 119 5 , , PUNCT bracis-28377 119 6 the the DET bracis-28377 119 7 efficientformer efficientformer NOUN bracis-28377 119 8 - - PUNCT bracis-28377 119 9 l1 l1 PROPN bracis-28377 119 10 is be AUX bracis-28377 119 11 coupled couple VERB bracis-28377 119 12 with with ADP bracis-28377 119 13 the the DET bracis-28377 119 14 arcface arcface NOUN bracis-28377 119 15 error error NOUN bracis-28377 119 16 function function NOUN bracis-28377 119 17     SPACE bracis-28377 120 1 [ [ X bracis-28377 120 2 4 4 NUM bracis-28377 120 3 ] ] PUNCT bracis-28377 120 4 , , PUNCT bracis-28377 120 5 a a DET bracis-28377 120 6 popular popular ADJ bracis-28377 120 7 loss loss NOUN bracis-28377 120 8 function function NOUN bracis-28377 120 9 in in ADP bracis-28377 120 10 biometrics biometric NOUN bracis-28377 120 11 which which PRON bracis-28377 120 12 aims aim VERB bracis-28377 120 13 to to PART bracis-28377 120 14 minimize minimize VERB bracis-28377 120 15 the the DET bracis-28377 120 16 interclass interclass NOUN bracis-28377 120 17 angular angular ADJ bracis-28377 120 18 distance distance NOUN bracis-28377 120 19 while while SCONJ bracis-28377 120 20 the the DET bracis-28377 120 21 intraclass intraclass NOUN bracis-28377 120 22 angular angular ADJ bracis-28377 120 23 distance distance NOUN bracis-28377 120 24 is be AUX bracis-28377 120 25 maximized maximize VERB bracis-28377 120 26 . . PUNCT bracis-28377 121 1 in in ADP bracis-28377 121 2 addition addition NOUN bracis-28377 121 3 , , PUNCT bracis-28377 121 4 an an DET bracis-28377 121 5 svm svm NOUN bracis-28377 121 6 classifier classifier NOUN bracis-28377 121 7 will will AUX bracis-28377 121 8 be be AUX bracis-28377 121 9 used use VERB bracis-28377 121 10 in in ADP bracis-28377 121 11 the the DET bracis-28377 121 12 output output NOUN bracis-28377 121 13 of of ADP bracis-28377 121 14 both both DET bracis-28377 121 15 architectures architecture NOUN bracis-28377 121 16 , , PUNCT bracis-28377 121 17 since since SCONJ bracis-28377 121 18 in in ADP bracis-28377 121 19 biometrics biometric NOUN bracis-28377 121 20 calculation calculation NOUN bracis-28377 121 21 is be AUX bracis-28377 121 22 used use VERB bracis-28377 121 23 using use VERB bracis-28377 121 24 distances distance NOUN bracis-28377 121 25 , , PUNCT bracis-28377 121 26 such such ADJ bracis-28377 121 27 as as ADP bracis-28377 121 28 the the DET bracis-28377 121 29 cosine cosine NOUN bracis-28377 121 30 distance distance NOUN bracis-28377 121 31 . . PUNCT bracis-28377 122 1 figure figure NOUN bracis-28377 122 2     SPACE bracis-28377 122 3 10 10 NUM bracis-28377 122 4 demonstrates demonstrate VERB bracis-28377 122 5 the the DET bracis-28377 122 6 architecture architecture NOUN bracis-28377 122 7 of of ADP bracis-28377 122 8 the the DET bracis-28377 122 9 proposed propose VERB bracis-28377 122 10 method method NOUN bracis-28377 122 11 with with ADP bracis-28377 122 12 the the DET bracis-28377 122 13 architecture architecture NOUN bracis-28377 122 14 of of ADP bracis-28377 122 15 efficientformer efficientformer NOUN bracis-28377 122 16 - - PUNCT bracis-28377 122 17 l1 l1 PROPN bracis-28377 122 18 . . PUNCT bracis-28377 123 1 fig fig NOUN bracis-28377 123 2 . . PUNCT bracis-28377 124 1 10 10 NUM bracis-28377 124 2 . . X bracis-28377 125 1 diagram diagram NOUN bracis-28377 125 2 of of ADP bracis-28377 125 3 the the DET bracis-28377 125 4 proposed propose VERB bracis-28377 125 5 method method NOUN bracis-28377 125 6 , , PUNCT bracis-28377 125 7 describing describe VERB bracis-28377 125 8 both both CCONJ bracis-28377 125 9 the the DET bracis-28377 125 10 transformer transformer ADJ bracis-28377 125 11 step step NOUN bracis-28377 125 12 and and CCONJ bracis-28377 125 13 the the DET bracis-28377 125 14 arcface arcface NOUN bracis-28377 125 15 technique technique NOUN bracis-28377 125 16 used use VERB bracis-28377 125 17 to to PART bracis-28377 125 18 position position VERB bracis-28377 125 19 the the DET bracis-28377 125 20 face face NOUN bracis-28377 125 21 descriptors descriptor NOUN bracis-28377 125 22 from from ADP bracis-28377 125 23 the the DET bracis-28377 125 24 same same ADJ bracis-28377 125 25 dog dog NOUN bracis-28377 125 26 near near ADP bracis-28377 125 27 each each DET bracis-28377 125 28 other other ADJ bracis-28377 125 29 and and CCONJ bracis-28377 125 30 the the DET bracis-28377 125 31 face face NOUN bracis-28377 125 32 descriptors descriptor NOUN bracis-28377 125 33 from from ADP bracis-28377 125 34 distinct distinct ADJ bracis-28377 125 35 dogs dog NOUN bracis-28377 125 36 far far ADV bracis-28377 125 37 away away ADV bracis-28377 125 38 , , PUNCT bracis-28377 125 39 in in ADP bracis-28377 125 40 the the DET bracis-28377 125 41 latent latent NOUN bracis-28377 125 42 space space NOUN bracis-28377 125 43 . . PUNCT bracis-28377 126 1 full full ADJ bracis-28377 126 2 size size NOUN bracis-28377 126 3 image image NOUN bracis-28377 126 4 5 5 NUM bracis-28377 126 5 results result NOUN bracis-28377 126 6 and and CCONJ bracis-28377 126 7     SPACE bracis-28377 126 8 discussion discussion NOUN bracis-28377 126 9 in in ADP bracis-28377 126 10 order order NOUN bracis-28377 126 11 to to PART bracis-28377 126 12 assess assess VERB bracis-28377 126 13 the the DET bracis-28377 126 14 proposed propose VERB bracis-28377 126 15 method method NOUN bracis-28377 126 16 for for ADP bracis-28377 126 17 dog dog NOUN bracis-28377 126 18 face face NOUN bracis-28377 126 19 recognition recognition NOUN bracis-28377 126 20 , , PUNCT bracis-28377 126 21 using use VERB bracis-28377 126 22 visual visual ADJ bracis-28377 126 23 transformers transformer NOUN bracis-28377 126 24 , , PUNCT bracis-28377 126 25 more more ADV bracis-28377 126 26 specifically specifically ADV bracis-28377 126 27 by by ADP bracis-28377 126 28 using use VERB bracis-28377 126 29 the the DET bracis-28377 126 30 efficientformer efficientformer NOUN bracis-28377 126 31 - - PUNCT bracis-28377 126 32 l1 l1 PROPN bracis-28377 126 33 architecture architecture NOUN bracis-28377 126 34 , , PUNCT bracis-28377 126 35 and and CCONJ bracis-28377 126 36 arcface arcface NOUN bracis-28377 126 37 , , PUNCT bracis-28377 126 38 two two NUM bracis-28377 126 39 experiments experiment NOUN bracis-28377 126 40 were be AUX bracis-28377 126 41 carried carry VERB bracis-28377 126 42 out out ADP bracis-28377 126 43 on on ADP bracis-28377 126 44 the the DET bracis-28377 126 45 dogfacenet dogfacenet NOUN bracis-28377 126 46 dataset dataset NOUN bracis-28377 126 47 . . PUNCT bracis-28377 127 1 as as ADP bracis-28377 127 2 the the DET bracis-28377 127 3 subset subset NOUN bracis-28377 127 4 of of ADP bracis-28377 127 5 data datum NOUN bracis-28377 127 6 used use VERB bracis-28377 127 7 by by ADP bracis-28377 127 8 mougeot mougeot NOUN bracis-28377 127 9 , , PUNCT bracis-28377 127 10 li li PROPN bracis-28377 127 11 and and CCONJ bracis-28377 127 12 jia jia PROPN bracis-28377 128 1 [ [ X bracis-28377 128 2 17 17 NUM bracis-28377 128 3 ] ] X bracis-28377 128 4 is be AUX bracis-28377 128 5 different different ADJ bracis-28377 128 6 than than SCONJ bracis-28377 128 7 the the DET bracis-28377 128 8 one one NOUN bracis-28377 128 9 publicly publicly ADV bracis-28377 128 10 shared share VERB bracis-28377 128 11 , , PUNCT bracis-28377 128 12 a a DET bracis-28377 128 13 protocol protocol NOUN bracis-28377 128 14 , , PUNCT bracis-28377 128 15 inspired inspire VERB bracis-28377 128 16 by by ADP bracis-28377 128 17 them they PRON bracis-28377 128 18 and and CCONJ bracis-28377 128 19 using use VERB bracis-28377 128 20 standard standard ADJ bracis-28377 128 21 practices practice NOUN bracis-28377 128 22 for for ADP bracis-28377 128 23 open open ADJ bracis-28377 128 24 - - PUNCT bracis-28377 128 25 set set VERB bracis-28377 128 26 biometric biometric ADJ bracis-28377 128 27 authentication authentication NOUN bracis-28377 128 28 was be AUX bracis-28377 128 29 used use VERB bracis-28377 128 30 . . PUNCT bracis-28377 129 1 the the DET bracis-28377 129 2 results result NOUN bracis-28377 129 3 obtained obtain VERB bracis-28377 129 4 were be AUX bracis-28377 129 5 compared compare VERB bracis-28377 129 6 with with ADP bracis-28377 129 7 a a DET bracis-28377 129 8 baseline baseline NOUN bracis-28377 129 9 method method NOUN bracis-28377 129 10 based base VERB bracis-28377 129 11 on on ADP bracis-28377 129 12 the the DET bracis-28377 129 13 resnet-50 resnet-50 PROPN bracis-28377 129 14 architecture architecture NOUN bracis-28377 129 15     SPACE bracis-28377 129 16 [ [ X bracis-28377 129 17 8 8 NUM bracis-28377 129 18 ] ] PUNCT bracis-28377 129 19 proposed propose VERB bracis-28377 129 20 by by ADP bracis-28377 129 21 mougeot mougeot PROPN bracis-28377 129 22 , , PUNCT bracis-28377 129 23 li li PROPN bracis-28377 129 24 and and CCONJ bracis-28377 129 25 jia jia PROPN bracis-28377 130 1 [ [ X bracis-28377 130 2 17 17 NUM bracis-28377 130 3 ] ] PUNCT bracis-28377 130 4 . . PUNCT bracis-28377 131 1 it it PRON bracis-28377 131 2 is be AUX bracis-28377 131 3 important important ADJ bracis-28377 131 4 to to PART bracis-28377 131 5 point point VERB bracis-28377 131 6 out out ADP bracis-28377 131 7 that that SCONJ bracis-28377 131 8 the the DET bracis-28377 131 9 experiments experiment NOUN bracis-28377 131 10 are be AUX bracis-28377 131 11 not not PART bracis-28377 131 12 identical identical ADJ bracis-28377 131 13 to to ADP bracis-28377 131 14 those those PRON bracis-28377 131 15 carried carry VERB bracis-28377 131 16 out out ADP bracis-28377 131 17 by by ADP bracis-28377 131 18 mougeot mougeot NOUN bracis-28377 131 19 , , PUNCT bracis-28377 131 20 li li PROPN bracis-28377 131 21 and and CCONJ bracis-28377 131 22 jia jia PROPN bracis-28377 132 1 [ [ X bracis-28377 132 2 17 17 NUM bracis-28377 132 3 ] ] PUNCT bracis-28377 132 4 , , PUNCT bracis-28377 132 5 as as SCONJ bracis-28377 132 6 the the DET bracis-28377 132 7 publicly publicly ADV bracis-28377 132 8 disclosed disclose VERB bracis-28377 132 9 base base NOUN bracis-28377 132 10 is be AUX bracis-28377 132 11 different different ADJ bracis-28377 132 12 from from ADP bracis-28377 132 13 the the DET bracis-28377 132 14 original original ADJ bracis-28377 132 15 base base NOUN bracis-28377 132 16 . . PUNCT bracis-28377 133 1 in in ADP bracis-28377 133 2 addiction addiction NOUN bracis-28377 133 3 , , PUNCT bracis-28377 133 4 in in ADP bracis-28377 133 5 table table NOUN bracis-28377 133 6     SPACE bracis-28377 133 7 1 1 NUM bracis-28377 133 8 there there PRON bracis-28377 133 9 are be VERB bracis-28377 133 10 details detail NOUN bracis-28377 133 11 of of ADP bracis-28377 133 12 the the DET bracis-28377 133 13 servers server NOUN bracis-28377 133 14 that that PRON bracis-28377 133 15 were be AUX bracis-28377 133 16 used use VERB bracis-28377 133 17 to to PART bracis-28377 133 18 run run VERB bracis-28377 133 19 the the DET bracis-28377 133 20 experiments experiment NOUN bracis-28377 133 21 . . PUNCT bracis-28377 134 1 being be AUX bracis-28377 134 2 the the DET bracis-28377 134 3 implementation implementation NOUN bracis-28377 134 4 of of ADP bracis-28377 134 5 the the DET bracis-28377 134 6 resnet resnet ADJ bracis-28377 134 7 architecture architecture NOUN bracis-28377 134 8 on on ADP bracis-28377 134 9 server server NOUN bracis-28377 134 10 1 1 NUM bracis-28377 134 11 and and CCONJ bracis-28377 134 12 the the DET bracis-28377 134 13 efficientformer efficientformer NOUN bracis-28377 134 14 - - PUNCT bracis-28377 134 15 l1 l1 PROPN bracis-28377 134 16 architecture architecture NOUN bracis-28377 134 17 on on ADP bracis-28377 134 18 server server NOUN bracis-28377 134 19 2 2 NUM bracis-28377 134 20 . . PUNCT bracis-28377 134 21 table table NOUN bracis-28377 134 22 1 1 NUM bracis-28377 134 23 . . PUNCT bracis-28377 134 24 server server NOUN bracis-28377 134 25 settingsfull settingsfull NOUN bracis-28377 134 26 size size NOUN bracis-28377 134 27 table table NOUN bracis-28377 134 28 5.1 5.1 NUM bracis-28377 134 29 experiment experiment NOUN bracis-28377 134 30 verification verification NOUN bracis-28377 134 31 task task NOUN bracis-28377 134 32 in in ADP bracis-28377 134 33 the the DET bracis-28377 134 34 verification verification NOUN bracis-28377 134 35 task task NOUN bracis-28377 134 36 , , PUNCT bracis-28377 134 37 two two NUM bracis-28377 134 38 architectures architecture NOUN bracis-28377 134 39 , , PUNCT bracis-28377 134 40 efficientformer efficientformer NOUN bracis-28377 134 41 - - PUNCT bracis-28377 134 42 l1 l1 PROPN bracis-28377 134 43 and and CCONJ bracis-28377 134 44 resnet-50 resnet-50 NUM bracis-28377 134 45 , , PUNCT bracis-28377 134 46 were be AUX bracis-28377 134 47 employed employ VERB bracis-28377 134 48 in in ADP bracis-28377 134 49 a a DET bracis-28377 134 50 standard standard ADJ bracis-28377 134 51 facial facial ADJ bracis-28377 134 52 biometrics biometric NOUN bracis-28377 134 53 evaluation evaluation NOUN bracis-28377 134 54 protocol protocol NOUN bracis-28377 134 55 . . PUNCT bracis-28377 135 1 this this DET bracis-28377 135 2 protocol protocol NOUN bracis-28377 135 3 involves involve VERB bracis-28377 135 4 using use VERB bracis-28377 135 5 all all DET bracis-28377 135 6 the the DET bracis-28377 135 7 comparisons comparison NOUN bracis-28377 135 8 within within ADP bracis-28377 135 9 the the DET bracis-28377 135 10 dataset dataset NOUN bracis-28377 135 11 in in ADP bracis-28377 135 12 order order NOUN bracis-28377 135 13 to to PART bracis-28377 135 14 obtain obtain VERB bracis-28377 135 15 the the DET bracis-28377 135 16 genuine genuine ADJ bracis-28377 135 17 and and CCONJ bracis-28377 135 18 impostor impostor NOUN bracis-28377 135 19 score score NOUN bracis-28377 135 20 distributions distribution NOUN bracis-28377 135 21 and and CCONJ bracis-28377 135 22 , , PUNCT bracis-28377 135 23 then then ADV bracis-28377 135 24 , , PUNCT bracis-28377 135 25 construct construct VERB bracis-28377 135 26 the the DET bracis-28377 135 27 receiver receiver NOUN bracis-28377 135 28 operating operate VERB bracis-28377 135 29 characteristic characteristic NOUN bracis-28377 135 30 ( ( PUNCT bracis-28377 135 31 roc roc PROPN bracis-28377 135 32 ) ) PUNCT bracis-28377 135 33 curve curve NOUN bracis-28377 135 34 . . PUNCT bracis-28377 136 1 furthermore furthermore ADV bracis-28377 136 2 , , PUNCT bracis-28377 136 3 to to PART bracis-28377 136 4 ensure ensure VERB bracis-28377 136 5 balanced balanced ADJ bracis-28377 136 6 evaluation evaluation NOUN bracis-28377 136 7 , , PUNCT bracis-28377 136 8 we we PRON bracis-28377 136 9 selected select VERB bracis-28377 136 10 all all DET bracis-28377 136 11 positive positive ADJ bracis-28377 136 12 comparisons comparison NOUN bracis-28377 136 13 ( ( PUNCT bracis-28377 136 14 a a DET bracis-28377 136 15 total total NOUN bracis-28377 136 16 of of ADP bracis-28377 136 17 2561 2561 NUM bracis-28377 136 18 pairs pair NOUN bracis-28377 136 19 of of ADP bracis-28377 136 20 dogs dog NOUN bracis-28377 136 21 ) ) PUNCT bracis-28377 136 22 and and CCONJ bracis-28377 136 23 an an DET bracis-28377 136 24 equal equal ADJ bracis-28377 136 25 number number NOUN bracis-28377 136 26 of of ADP bracis-28377 136 27 negative negative ADJ bracis-28377 136 28 comparisons comparison NOUN bracis-28377 136 29 . . PUNCT bracis-28377 137 1 from from ADP bracis-28377 137 2 these these DET bracis-28377 137 3 pairs pair NOUN bracis-28377 137 4 , , PUNCT bracis-28377 137 5 we we PRON bracis-28377 137 6 computed compute VERB bracis-28377 137 7 additional additional ADJ bracis-28377 137 8 evaluation evaluation NOUN bracis-28377 137 9 metrics metric NOUN bracis-28377 137 10 , , PUNCT bracis-28377 137 11 including include VERB bracis-28377 137 12 accuracy accuracy NOUN bracis-28377 137 13 , , PUNCT bracis-28377 137 14 precision precision NOUN bracis-28377 137 15 , , PUNCT bracis-28377 137 16 recall recall NOUN bracis-28377 137 17 , , PUNCT bracis-28377 137 18 and and CCONJ bracis-28377 137 19 f-1 f-1 NOUN bracis-28377 137 20 score score NOUN bracis-28377 137 21 . . PUNCT bracis-28377 138 1 initially initially ADV bracis-28377 138 2 , , PUNCT bracis-28377 138 3 an an DET bracis-28377 138 4 experiment experiment NOUN bracis-28377 138 5 was be AUX bracis-28377 138 6 carried carry VERB bracis-28377 138 7 out out ADP bracis-28377 138 8 in in ADP bracis-28377 138 9 relation relation NOUN bracis-28377 138 10 to to ADP bracis-28377 138 11 the the DET bracis-28377 138 12 verification verification NOUN bracis-28377 138 13 task task NOUN bracis-28377 138 14 , , PUNCT bracis-28377 138 15 an an DET bracis-28377 138 16 efficientformer efficientformer NOUN bracis-28377 138 17 - - PUNCT bracis-28377 138 18 l1 l1 NOUN bracis-28377 138 19 architecture architecture NOUN bracis-28377 138 20 was be AUX bracis-28377 138 21 used use VERB bracis-28377 138 22 for for ADP bracis-28377 138 23 feature feature NOUN bracis-28377 138 24 extraction extraction NOUN bracis-28377 138 25 combined combine VERB bracis-28377 138 26 with with ADP bracis-28377 138 27 the the DET bracis-28377 138 28 arcface arcface NOUN bracis-28377 138 29 error error NOUN bracis-28377 138 30 function function NOUN bracis-28377 138 31 with with ADP bracis-28377 138 32 3 3 NUM bracis-28377 138 33 subcenters subcenter NOUN bracis-28377 138 34 to to PART bracis-28377 138 35 perform perform VERB bracis-28377 138 36 the the DET bracis-28377 138 37 learning learning NOUN bracis-28377 138 38 . . PUNCT bracis-28377 139 1 the the DET bracis-28377 139 2 hyperparameters hyperparameter NOUN bracis-28377 139 3 in in ADP bracis-28377 139 4 relation relation NOUN bracis-28377 139 5 to to ADP bracis-28377 139 6 the the DET bracis-28377 139 7 arcface arcface NOUN bracis-28377 139 8 error error NOUN bracis-28377 139 9 function function NOUN bracis-28377 139 10 were be AUX bracis-28377 139 11 obtained obtain VERB bracis-28377 139 12 using use VERB bracis-28377 139 13 as as ADP bracis-28377 139 14 reference reference NOUN bracis-28377 139 15 the the DET bracis-28377 139 16 analysis analysis NOUN bracis-28377 139 17 proposed propose VERB bracis-28377 139 18 by by ADP bracis-28377 139 19 the the DET bracis-28377 139 20 fixed fix VERB bracis-28377 139 21 adacos adacos PROPN bracis-28377 139 22 method method NOUN bracis-28377 139 23 , , PUNCT bracis-28377 139 24 proposed propose VERB bracis-28377 139 25 by by ADP bracis-28377 139 26 [ [ X bracis-28377 139 27 23 23 NUM bracis-28377 139 28 ] ] PUNCT bracis-28377 139 29 , , PUNCT bracis-28377 139 30 in in ADP bracis-28377 139 31 which which PRON bracis-28377 139 32 the the DET bracis-28377 139 33 margin margin NOUN bracis-28377 139 34 parameter parameter NOUN bracis-28377 139 35 is be AUX bracis-28377 139 36 fixed fix VERB bracis-28377 139 37 at at ADP bracis-28377 139 38 \(m \(m NOUN bracis-28377 139 39 = = SYM bracis-28377 139 40 0.5\ 0.5\ NUM bracis-28377 139 41 ) ) PUNCT bracis-28377 139 42 and and CCONJ bracis-28377 139 43 the the DET bracis-28377 139 44 scale scale NOUN bracis-28377 139 45 parameter parameter NOUN bracis-28377 139 46 is be AUX bracis-28377 139 47 obtained obtain VERB bracis-28377 139 48 dynamically dynamically ADV bracis-28377 139 49 by by ADP bracis-28377 139 50 \(s \( NOUN bracis-28377 139 51 = = SYM bracis-28377 139 52 \sqrt{2}\cdot \sqrt{2}\cdot NOUN bracis-28377 139 53 \log \log PROPN bracis-28377 139 54 ( ( PUNCT bracis-28377 139 55 c-1)\ c-1)\ PROPN bracis-28377 139 56 ) ) PUNCT bracis-28377 139 57 , , PUNCT bracis-28377 139 58 where where SCONJ bracis-28377 139 59 c c PROPN bracis-28377 139 60 is be AUX bracis-28377 139 61 the the DET bracis-28377 139 62 number number NOUN bracis-28377 139 63 of of ADP bracis-28377 139 64 classes class NOUN bracis-28377 139 65 . . PUNCT bracis-28377 140 1 figure figure VERB bracis-28377 140 2     SPACE bracis-28377 140 3 11 11 NUM bracis-28377 140 4 shows show VERB bracis-28377 140 5 the the DET bracis-28377 140 6 genuine genuine ADJ bracis-28377 140 7 comparison comparison NOUN bracis-28377 140 8 score score NOUN bracis-28377 140 9 distributions distribution NOUN bracis-28377 140 10 and and CCONJ bracis-28377 140 11 the the DET bracis-28377 140 12 imposter impost ADJ bracis-28377 140 13 comparison comparison NOUN bracis-28377 140 14 score score NOUN bracis-28377 140 15 distributions distribution NOUN bracis-28377 140 16 obtained obtain VERB bracis-28377 140 17 using use VERB bracis-28377 140 18 resnet-50 resnet-50 PROPN bracis-28377 140 19 and and CCONJ bracis-28377 140 20 efficientformer efficientformer NOUN bracis-28377 140 21 - - PUNCT bracis-28377 140 22 l1 l1 PROPN bracis-28377 140 23 features feature NOUN bracis-28377 140 24 . . PUNCT bracis-28377 141 1 one one PRON bracis-28377 141 2 can can AUX bracis-28377 141 3 observe observe VERB bracis-28377 141 4 that that SCONJ bracis-28377 141 5 the the DET bracis-28377 141 6 efficientformer efficientformer NOUN bracis-28377 141 7 - - PUNCT bracis-28377 141 8 l1 l1 PROPN bracis-28377 141 9 features feature NOUN bracis-28377 141 10 generated generate VERB bracis-28377 141 11 more more ADV bracis-28377 141 12 separated separated ADJ bracis-28377 141 13 distributions distribution NOUN bracis-28377 141 14 than than ADP bracis-28377 141 15 the the DET bracis-28377 141 16 resnet-50 resnet-50 PROPN bracis-28377 141 17 features feature NOUN bracis-28377 141 18 , , PUNCT bracis-28377 141 19 leading lead VERB bracis-28377 141 20 to to ADP bracis-28377 141 21 a a DET bracis-28377 141 22 lower low ADJ bracis-28377 141 23 intersection intersection NOUN bracis-28377 141 24 between between ADP bracis-28377 141 25 the the DET bracis-28377 141 26 distributions distribution NOUN bracis-28377 141 27 and and CCONJ bracis-28377 141 28 , , PUNCT bracis-28377 141 29 consequently consequently ADV bracis-28377 141 30 , , PUNCT bracis-28377 141 31 lower low ADJ bracis-28377 141 32 error error NOUN bracis-28377 141 33 rates rate NOUN bracis-28377 141 34 . . PUNCT bracis-28377 142 1 efficientformer efficientformer NOUN bracis-28377 142 2 - - PUNCT bracis-28377 142 3 l1 l1 PROPN bracis-28377 142 4 obtained obtain VERB bracis-28377 142 5 a a DET bracis-28377 142 6 better well ADJ bracis-28377 142 7 auc auc NOUN bracis-28377 142 8 value value NOUN bracis-28377 142 9 , , PUNCT bracis-28377 142 10 equal equal ADJ bracis-28377 142 11 to to ADP bracis-28377 142 12 0.989319 0.989319 NUM bracis-28377 142 13 , , PUNCT bracis-28377 142 14 while while SCONJ bracis-28377 142 15 the the DET bracis-28377 142 16 resnet-50 resnet-50 PROPN bracis-28377 142 17 obtained obtain VERB bracis-28377 142 18 an an DET bracis-28377 142 19 auc auc NOUN bracis-28377 142 20 value value NOUN bracis-28377 142 21 equal equal ADJ bracis-28377 142 22 to to ADP bracis-28377 142 23 0.952105 0.952105 NUM bracis-28377 142 24 , , PUNCT bracis-28377 142 25 as as SCONJ bracis-28377 142 26 shown show VERB bracis-28377 142 27 in in ADP bracis-28377 142 28 fig fig NOUN bracis-28377 142 29 . . PUNCT bracis-28377 142 30     SPACE bracis-28377 143 1 12 12 NUM bracis-28377 143 2 . . PUNCT bracis-28377 144 1 fig fig NOUN bracis-28377 144 2 . . PUNCT bracis-28377 145 1 11 11 NUM bracis-28377 145 2 . . PUNCT bracis-28377 146 1 distributions distribution NOUN bracis-28377 146 2 of of ADP bracis-28377 146 3 similarity similarity NOUN bracis-28377 146 4 scores score NOUN bracis-28377 146 5 obtained obtain VERB bracis-28377 146 6 through through ADP bracis-28377 146 7 features feature NOUN bracis-28377 146 8 extracted extract VERB bracis-28377 146 9 using use VERB bracis-28377 146 10 resnet-50 resnet-50 PROPN bracis-28377 146 11 and and CCONJ bracis-28377 146 12 efficientformer efficientformer NOUN bracis-28377 146 13 - - PUNCT bracis-28377 146 14 l1 l1 PROPN bracis-28377 146 15 . . PUNCT bracis-28377 147 1 full full ADJ bracis-28377 147 2 size size NOUN bracis-28377 147 3 image image NOUN bracis-28377 147 4 fig fig NOUN bracis-28377 147 5 . . PUNCT bracis-28377 148 1 12 12 NUM bracis-28377 148 2 . . PUNCT bracis-28377 149 1 roc roc PROPN bracis-28377 149 2 curves curve NOUN bracis-28377 149 3 obtained obtain VERB bracis-28377 149 4 on on ADP bracis-28377 149 5 dogfacenet dogfacenet NOUN bracis-28377 149 6 dataset dataset VERB bracis-28377 149 7 by by ADP bracis-28377 149 8 using use VERB bracis-28377 149 9 features feature NOUN bracis-28377 149 10 extracted extract VERB bracis-28377 149 11 from from ADP bracis-28377 149 12 the the DET bracis-28377 149 13 resnet-50 resnet-50 PROPN bracis-28377 149 14 and and CCONJ bracis-28377 149 15 efficientformer efficientformer NOUN bracis-28377 149 16 - - PUNCT bracis-28377 149 17 l1 l1 PROPN bracis-28377 149 18 . . PUNCT bracis-28377 150 1 full full ADJ bracis-28377 150 2 size size NOUN bracis-28377 150 3 image image NOUN bracis-28377 150 4 table table NOUN bracis-28377 150 5     SPACE bracis-28377 150 6 2 2 NUM bracis-28377 150 7 presents present VERB bracis-28377 150 8 the the DET bracis-28377 150 9 values value NOUN bracis-28377 150 10 of of ADP bracis-28377 150 11 area area NOUN bracis-28377 150 12 under under ADP bracis-28377 150 13 the the DET bracis-28377 150 14 roc roc PROPN bracis-28377 150 15 curve curve NOUN bracis-28377 150 16 ( ( PUNCT bracis-28377 150 17 auc auc NOUN bracis-28377 150 18 ) ) PUNCT bracis-28377 150 19 , , PUNCT bracis-28377 150 20 equal equal ADJ bracis-28377 150 21 error error NOUN bracis-28377 150 22 rate rate NOUN bracis-28377 150 23 ( ( PUNCT bracis-28377 150 24 eer eer PROPN bracis-28377 150 25 ) ) PUNCT bracis-28377 150 26 , , PUNCT bracis-28377 150 27 accuracy accuracy NOUN bracis-28377 150 28 , , PUNCT bracis-28377 150 29 f1 f1 NOUN bracis-28377 150 30 - - PUNCT bracis-28377 150 31 score score NOUN bracis-28377 150 32 , , PUNCT bracis-28377 150 33 precision precision NOUN bracis-28377 150 34 and and CCONJ bracis-28377 150 35 recall recall NOUN bracis-28377 150 36 , , PUNCT bracis-28377 150 37 obtained obtain VERB bracis-28377 150 38 in in ADP bracis-28377 150 39 the the DET bracis-28377 150 40 verification verification NOUN bracis-28377 150 41 task task NOUN bracis-28377 150 42 , , PUNCT bracis-28377 150 43 on on ADP bracis-28377 150 44 the the DET bracis-28377 150 45 dogfacenet dogfacenet NOUN bracis-28377 150 46 dataset dataset NOUN bracis-28377 150 47 , , PUNCT bracis-28377 150 48 using use VERB bracis-28377 150 49 face face NOUN bracis-28377 150 50 features feature NOUN bracis-28377 150 51 extracted extract VERB bracis-28377 150 52 from from ADP bracis-28377 150 53 the the DET bracis-28377 150 54 efficientformer efficientformer NOUN bracis-28377 150 55 - - PUNCT bracis-28377 150 56 l1 l1 PROPN bracis-28377 150 57 and and CCONJ bracis-28377 150 58 resnet-50 resnet-50 PROPN bracis-28377 150 59 . . PUNCT bracis-28377 151 1 one one PRON bracis-28377 151 2 can can AUX bracis-28377 151 3 observe observe VERB bracis-28377 151 4 that that SCONJ bracis-28377 151 5 efficientformer efficientformer NOUN bracis-28377 151 6 - - PUNCT bracis-28377 151 7 l1 l1 PROPN bracis-28377 151 8 features feature NOUN bracis-28377 151 9 performed perform VERB bracis-28377 151 10 better well ADV bracis-28377 151 11 than than ADP bracis-28377 151 12 the the DET bracis-28377 151 13 resnet-50 resnet-50 PROPN bracis-28377 151 14 features feature NOUN bracis-28377 151 15 in in ADP bracis-28377 151 16 all all DET bracis-28377 151 17 metrics metric NOUN bracis-28377 151 18 . . PUNCT bracis-28377 152 1 table table NOUN bracis-28377 152 2 2 2 NUM bracis-28377 152 3 . . PUNCT bracis-28377 152 4 results result NOUN bracis-28377 152 5 obtained obtain VERB bracis-28377 152 6 in in ADP bracis-28377 152 7 the the DET bracis-28377 152 8 verification verification NOUN bracis-28377 152 9 task task NOUN bracis-28377 152 10 , , PUNCT bracis-28377 152 11 on on ADP bracis-28377 152 12 dogfacenet dogfacenet NOUN bracis-28377 152 13 dataset dataset NOUN bracis-28377 152 14 , , PUNCT bracis-28377 152 15 using use VERB bracis-28377 152 16 face face NOUN bracis-28377 152 17 features feature NOUN bracis-28377 152 18 extracted extract VERB bracis-28377 152 19 from from ADP bracis-28377 152 20 the the DET bracis-28377 152 21 efficientformer efficientformer NOUN bracis-28377 152 22 - - PUNCT bracis-28377 152 23 l1 l1 PROPN bracis-28377 152 24 and and CCONJ bracis-28377 152 25 resnet-50.full resnet-50.full NOUN bracis-28377 152 26 size size NOUN bracis-28377 152 27 table table NOUN bracis-28377 152 28 5.2 5.2 NUM bracis-28377 152 29 experiment experiment NOUN bracis-28377 152 30 identification identification NOUN bracis-28377 152 31 task task NOUN bracis-28377 152 32 in in ADP bracis-28377 152 33 the the DET bracis-28377 152 34 identification identification NOUN bracis-28377 152 35 task task NOUN bracis-28377 152 36 experiment experiment NOUN bracis-28377 152 37 , , PUNCT bracis-28377 152 38 the the DET bracis-28377 152 39 two two NUM bracis-28377 152 40 architectures architecture NOUN bracis-28377 152 41 , , PUNCT bracis-28377 152 42 efficientformer efficientformer NOUN bracis-28377 152 43 - - PUNCT bracis-28377 152 44 l1 l1 PROPN bracis-28377 152 45 and and CCONJ bracis-28377 152 46 resnet-50 resnet-50 NUM bracis-28377 152 47 , , PUNCT bracis-28377 152 48 were be AUX bracis-28377 152 49 used use VERB bracis-28377 152 50 again again ADV bracis-28377 152 51 . . PUNCT bracis-28377 153 1 to to PART bracis-28377 153 2 evaluate evaluate VERB bracis-28377 153 3 their their PRON bracis-28377 153 4 performance performance NOUN bracis-28377 153 5 , , PUNCT bracis-28377 153 6 we we PRON bracis-28377 153 7 employed employ VERB bracis-28377 153 8 two two NUM bracis-28377 153 9 distinct distinct ADJ bracis-28377 153 10 settings setting NOUN bracis-28377 153 11 with with ADP bracis-28377 153 12 two two NUM bracis-28377 153 13 different different ADJ bracis-28377 153 14 types type NOUN bracis-28377 153 15 of of ADP bracis-28377 153 16 data datum NOUN bracis-28377 153 17 split split NOUN bracis-28377 153 18 . . PUNCT bracis-28377 154 1 in in ADP bracis-28377 154 2 the the DET bracis-28377 154 3 first first ADJ bracis-28377 154 4 setting setting NOUN bracis-28377 154 5 , , PUNCT bracis-28377 154 6 we we PRON bracis-28377 154 7 computed compute VERB bracis-28377 154 8 the the DET bracis-28377 154 9 cumulative cumulative ADJ bracis-28377 154 10 matching matching NOUN bracis-28377 154 11 characteristic characteristic ADJ bracis-28377 154 12 ( ( PUNCT bracis-28377 154 13 cmc cmc NOUN bracis-28377 154 14 ) ) PUNCT bracis-28377 154 15 curve curve NOUN bracis-28377 154 16 by by ADP bracis-28377 154 17 comparing compare VERB bracis-28377 154 18 the the DET bracis-28377 154 19 cosine cosine NOUN bracis-28377 154 20 distance distance NOUN bracis-28377 154 21 between between ADP bracis-28377 154 22 every every DET bracis-28377 154 23 element element NOUN bracis-28377 154 24 in in ADP bracis-28377 154 25 the the DET bracis-28377 154 26 probe probe NOUN bracis-28377 154 27 set set NOUN bracis-28377 154 28 and and CCONJ bracis-28377 154 29 every every DET bracis-28377 154 30 element element NOUN bracis-28377 154 31 in in ADP bracis-28377 154 32 the the DET bracis-28377 154 33 gallery gallery NOUN bracis-28377 154 34 set set NOUN bracis-28377 154 35 . . PUNCT bracis-28377 155 1 this this DET bracis-28377 155 2 analysis analysis NOUN bracis-28377 155 3 provides provide VERB bracis-28377 155 4 insights insight NOUN bracis-28377 155 5 into into ADP bracis-28377 155 6 the the DET bracis-28377 155 7 model model NOUN bracis-28377 155 8 ’s ’s PART bracis-28377 155 9 ability ability NOUN bracis-28377 155 10 to to PART bracis-28377 155 11 rank rank VERB bracis-28377 155 12 the the DET bracis-28377 155 13 correct correct ADJ bracis-28377 155 14 matches match NOUN bracis-28377 155 15 in in ADP bracis-28377 155 16 the the DET bracis-28377 155 17 gallery gallery NOUN bracis-28377 155 18 . . PUNCT bracis-28377 156 1 the the DET bracis-28377 156 2 data datum NOUN bracis-28377 156 3 was be AUX bracis-28377 156 4 split split VERB bracis-28377 156 5 by by ADP bracis-28377 156 6 using use VERB bracis-28377 156 7 a a DET bracis-28377 156 8 k k ADJ bracis-28377 156 9 - - ADJ bracis-28377 156 10 fold fold ADJ bracis-28377 156 11 cross cross ADJ bracis-28377 156 12 - - ADJ bracis-28377 156 13 validation validation ADJ bracis-28377 156 14 strategy strategy NOUN bracis-28377 156 15 , , PUNCT bracis-28377 156 16 with with ADP bracis-28377 156 17 k k PROPN bracis-28377 156 18 set set VERB bracis-28377 156 19 to to ADP bracis-28377 156 20 10 10 NUM bracis-28377 156 21 , , PUNCT bracis-28377 156 22 where where SCONJ bracis-28377 156 23 90 90 NUM bracis-28377 156 24 % % NOUN bracis-28377 156 25 of of ADP bracis-28377 156 26 the the DET bracis-28377 156 27 open open ADJ bracis-28377 156 28 - - PUNCT bracis-28377 156 29 set set VERB bracis-28377 156 30 data datum NOUN bracis-28377 156 31 was be AUX bracis-28377 156 32 used use VERB bracis-28377 156 33 for for ADP bracis-28377 156 34 gallery gallery NOUN bracis-28377 156 35 creation creation NOUN bracis-28377 156 36 , , PUNCT bracis-28377 156 37 and and CCONJ bracis-28377 156 38 the the DET bracis-28377 156 39 remaining remain VERB bracis-28377 156 40 10 10 NUM bracis-28377 156 41 % % NOUN bracis-28377 156 42 was be AUX bracis-28377 156 43 reserved reserve VERB bracis-28377 156 44 for for ADP bracis-28377 156 45 evaluation evaluation NOUN bracis-28377 156 46 . . PUNCT bracis-28377 157 1 to to PART bracis-28377 157 2 further far ADV bracis-28377 157 3 evaluate evaluate VERB bracis-28377 157 4 the the DET bracis-28377 157 5 models model NOUN bracis-28377 157 6 , , PUNCT bracis-28377 157 7 we we PRON bracis-28377 157 8 trained train VERB bracis-28377 157 9 an an DET bracis-28377 157 10 svm svm NOUN bracis-28377 157 11 using use VERB bracis-28377 157 12 the the DET bracis-28377 157 13 k k ADJ bracis-28377 157 14 - - ADJ bracis-28377 157 15 fold fold ADJ bracis-28377 157 16 cross cross NOUN bracis-28377 157 17 - - NOUN bracis-28377 157 18 validations validation NOUN bracis-28377 157 19 with with ADP bracis-28377 157 20 k k PROPN bracis-28377 157 21 set set VERB bracis-28377 157 22 to to ADP bracis-28377 157 23 10 10 NUM bracis-28377 157 24 . . PUNCT bracis-28377 158 1 table table NOUN bracis-28377 158 2     SPACE bracis-28377 158 3 3 3 NUM bracis-28377 158 4 shows show VERB bracis-28377 158 5 the the DET bracis-28377 158 6 results result NOUN bracis-28377 158 7 obtained obtain VERB bracis-28377 158 8 . . PUNCT bracis-28377 158 9 table table NOUN bracis-28377 158 10 3 3 NUM bracis-28377 158 11 . . PUNCT bracis-28377 159 1 accuracy accuracy NOUN bracis-28377 159 2 results result NOUN bracis-28377 159 3 in in ADP bracis-28377 159 4 the the DET bracis-28377 159 5 identification identification NOUN bracis-28377 159 6 task task NOUN bracis-28377 159 7 using use VERB bracis-28377 159 8 face face NOUN bracis-28377 159 9 features feature NOUN bracis-28377 159 10 extracted extract VERB bracis-28377 159 11 from from ADP bracis-28377 159 12 efficientformer efficientformer NOUN bracis-28377 159 13 - - PUNCT bracis-28377 159 14 l1 l1 PROPN bracis-28377 159 15 and and CCONJ bracis-28377 159 16 resnet-50 resnet-50 PROPN bracis-28377 159 17 and and CCONJ bracis-28377 159 18 an an DET bracis-28377 159 19 svm svm ADJ bracis-28377 159 20 classifier classifier NOUN bracis-28377 159 21 in in ADP bracis-28377 159 22 the the DET bracis-28377 159 23 k k ADJ bracis-28377 159 24 - - ADJ bracis-28377 159 25 fold fold ADJ bracis-28377 159 26 cross cross ADJ bracis-28377 159 27 - - ADJ bracis-28377 159 28 validation validation ADJ bracis-28377 159 29 protocol.full protocol.full NOUN bracis-28377 159 30 size size NOUN bracis-28377 159 31 table table NOUN bracis-28377 159 32 for for ADP bracis-28377 159 33 the the DET bracis-28377 159 34 second second ADJ bracis-28377 159 35 split split NOUN bracis-28377 159 36 , , PUNCT bracis-28377 159 37 a a DET bracis-28377 159 38 random random ADJ bracis-28377 159 39 sample sample NOUN bracis-28377 159 40 approach approach NOUN bracis-28377 159 41 was be AUX bracis-28377 159 42 employed employ VERB bracis-28377 159 43 , , PUNCT bracis-28377 159 44 as as SCONJ bracis-28377 159 45 proposed propose VERB bracis-28377 159 46 by by ADP bracis-28377 159 47 mougeot mougeot PROPN bracis-28377 159 48 , , PUNCT bracis-28377 159 49 li li PROPN bracis-28377 159 50 and and CCONJ bracis-28377 159 51 jia jia PROPN bracis-28377 160 1 [ [ X bracis-28377 160 2 17 17 NUM bracis-28377 160 3 ] ] PUNCT bracis-28377 160 4 , , PUNCT bracis-28377 160 5 where where SCONJ bracis-28377 160 6 m m PRON bracis-28377 160 7 samples sample NOUN bracis-28377 160 8 for for ADP bracis-28377 160 9 each each DET bracis-28377 160 10 class class NOUN bracis-28377 160 11 are be AUX bracis-28377 160 12 randomly randomly ADV bracis-28377 160 13 selected select VERB bracis-28377 160 14 from from ADP bracis-28377 160 15 the the DET bracis-28377 160 16 dataset dataset NOUN bracis-28377 160 17 and and CCONJ bracis-28377 160 18 included include VERB bracis-28377 160 19 in in ADP bracis-28377 160 20 the the DET bracis-28377 160 21 gallery gallery NOUN bracis-28377 160 22 set set NOUN bracis-28377 160 23 , , PUNCT bracis-28377 160 24 while while SCONJ bracis-28377 160 25 the the DET bracis-28377 160 26 rest rest NOUN bracis-28377 160 27 of of ADP bracis-28377 160 28 the the DET bracis-28377 160 29 samples sample NOUN bracis-28377 160 30 are be AUX bracis-28377 160 31 used use VERB bracis-28377 160 32 for for ADP bracis-28377 160 33 evaluation evaluation NOUN bracis-28377 160 34 . . PUNCT bracis-28377 161 1 for for ADP bracis-28377 161 2 this this DET bracis-28377 161 3 data datum NOUN bracis-28377 161 4 split split NOUN bracis-28377 161 5 approach approach NOUN bracis-28377 161 6 , , PUNCT bracis-28377 161 7 a a DET bracis-28377 161 8 k k PROPN bracis-28377 161 9 - - PUNCT bracis-28377 161 10 nn nn PROPN bracis-28377 161 11 classifier classifier NOUN bracis-28377 161 12 was be AUX bracis-28377 161 13 used use VERB bracis-28377 161 14 , , PUNCT bracis-28377 161 15 with with ADP bracis-28377 161 16 the the DET bracis-28377 161 17 k k PROPN bracis-28377 161 18 value value NOUN bracis-28377 161 19 being be AUX bracis-28377 161 20 chosen choose VERB bracis-28377 161 21 also also ADV bracis-28377 161 22 as as SCONJ bracis-28377 161 23 proposed propose VERB bracis-28377 161 24 by by ADP bracis-28377 161 25 mougeot mougeot PROPN bracis-28377 161 26 , , PUNCT bracis-28377 161 27 li li PROPN bracis-28377 161 28 and and CCONJ bracis-28377 161 29 jia jia PROPN bracis-28377 162 1 [ [ X bracis-28377 162 2 17 17 NUM bracis-28377 162 3 ] ] PUNCT bracis-28377 162 4 . . PUNCT bracis-28377 163 1 figure figure NOUN bracis-28377 163 2     SPACE bracis-28377 163 3 3 3 NUM bracis-28377 163 4 presents present VERB bracis-28377 163 5 the the DET bracis-28377 163 6 rank-10 rank-10 PROPN bracis-28377 163 7 cumulative cumulative ADJ bracis-28377 163 8 match match NOUN bracis-28377 163 9 characteristic characteristic ADJ bracis-28377 163 10 ( ( PUNCT bracis-28377 163 11 cmc cmc NOUN bracis-28377 163 12 ) ) PUNCT bracis-28377 163 13 curve curve NOUN bracis-28377 163 14 in in ADP bracis-28377 163 15 each each PRON bracis-28377 163 16 of of ADP bracis-28377 163 17 the the DET bracis-28377 163 18 data datum NOUN bracis-28377 163 19 separation separation NOUN bracis-28377 163 20 protocols protocol NOUN bracis-28377 163 21 for for ADP bracis-28377 163 22 the the DET bracis-28377 163 23 identification identification NOUN bracis-28377 163 24 task task NOUN bracis-28377 163 25 . . PUNCT bracis-28377 164 1 one one PRON bracis-28377 164 2 can can AUX bracis-28377 164 3 observe observe VERB bracis-28377 164 4 that that SCONJ bracis-28377 164 5 the the DET bracis-28377 164 6 efficientformer efficientformer NOUN bracis-28377 164 7 - - PUNCT bracis-28377 164 8 l1 l1 PROPN bracis-28377 164 9 features feature NOUN bracis-28377 164 10 obtained obtain VERB bracis-28377 164 11 very very ADV bracis-28377 164 12 superior superior ADJ bracis-28377 164 13 performances performance NOUN bracis-28377 164 14 when when SCONJ bracis-28377 164 15 compared compare VERB bracis-28377 164 16 with with ADP bracis-28377 164 17 the the DET bracis-28377 164 18 resnet-50 resnet-50 PROPN bracis-28377 164 19 features feature NOUN bracis-28377 164 20 , , PUNCT bracis-28377 164 21 in in ADP bracis-28377 164 22 the the DET bracis-28377 164 23 identification identification NOUN bracis-28377 164 24 task task NOUN bracis-28377 164 25 . . PUNCT bracis-28377 165 1 the the DET bracis-28377 165 2 best good ADJ bracis-28377 165 3 rank-1 rank-1 NUM bracis-28377 165 4 accuracy accuracy NOUN bracis-28377 165 5 value value NOUN bracis-28377 165 6 , , PUNCT bracis-28377 165 7 \(91.4\%\ \(91.4\%\ PROPN bracis-28377 165 8 ) ) PUNCT bracis-28377 165 9 , , PUNCT bracis-28377 165 10 was be AUX bracis-28377 165 11 obtained obtain VERB bracis-28377 165 12 by by ADP bracis-28377 165 13 efficientformer efficientformer NOUN bracis-28377 165 14 , , PUNCT bracis-28377 165 15 with with ADP bracis-28377 165 16 \(m=1\ \(m=1\ NOUN bracis-28377 165 17 ) ) PUNCT bracis-28377 165 18 , , PUNCT bracis-28377 165 19 while while SCONJ bracis-28377 165 20 the the DET bracis-28377 165 21 best good ADJ bracis-28377 165 22 rank-1 rank-1 NUM bracis-28377 165 23 accuracy accuracy NOUN bracis-28377 165 24 value value NOUN bracis-28377 165 25 obtained obtain VERB bracis-28377 165 26 by by ADP bracis-28377 165 27 resnet-50 resnet-50 PROPN bracis-28377 165 28 was be AUX bracis-28377 165 29 only only ADV bracis-28377 165 30 \(63.3\%\ \(63.3\%\ NOUN bracis-28377 165 31 ) ) PUNCT bracis-28377 165 32 , , PUNCT bracis-28377 165 33 also also ADV bracis-28377 165 34 with with ADP bracis-28377 165 35 \(m=1\ \(m=1\ NOUN bracis-28377 165 36 ) ) PUNCT bracis-28377 165 37 . . PUNCT bracis-28377 166 1 regarding regard VERB bracis-28377 166 2 the the DET bracis-28377 166 3 related relate VERB bracis-28377 166 4 works work NOUN bracis-28377 166 5 presented present VERB bracis-28377 166 6 in in ADP bracis-28377 166 7 sect sect NOUN bracis-28377 166 8 . . PUNCT bracis-28377 166 9     SPACE bracis-28377 166 10 2.2 2.2 NUM bracis-28377 166 11 , , PUNCT bracis-28377 166 12 using use VERB bracis-28377 166 13 the the DET bracis-28377 166 14 smaller small ADJ bracis-28377 166 15 initial initial ADJ bracis-28377 166 16 version version NOUN bracis-28377 166 17 of of ADP bracis-28377 166 18 the the DET bracis-28377 166 19 dogfacenet dogfacenet NOUN bracis-28377 166 20 , , PUNCT bracis-28377 166 21 mougeot mougeot NOUN bracis-28377 166 22 , , PUNCT bracis-28377 166 23 li li PROPN bracis-28377 166 24 and and CCONJ bracis-28377 166 25 jia jia PROPN bracis-28377 167 1 [ [ X bracis-28377 167 2 17 17 NUM bracis-28377 167 3 ] ] PUNCT bracis-28377 167 4 , , PUNCT bracis-28377 167 5 obtained obtain VERB bracis-28377 167 6 \(92\%\ \(92\%\ ADP bracis-28377 167 7 ) ) PUNCT bracis-28377 167 8 of of ADP bracis-28377 167 9 accuracy accuracy NOUN bracis-28377 167 10 in in ADP bracis-28377 167 11 the the DET bracis-28377 167 12 verification verification NOUN bracis-28377 167 13 task task NOUN bracis-28377 167 14 , , PUNCT bracis-28377 167 15 and and CCONJ bracis-28377 167 16 \(60.44\%\ \(60.44\%\ NOUN bracis-28377 167 17 ) ) PUNCT bracis-28377 167 18 of of ADP bracis-28377 167 19 rank-1 rank-1 NUM bracis-28377 167 20 accuracy accuracy NOUN bracis-28377 167 21 ( ( PUNCT bracis-28377 167 22 best good ADJ bracis-28377 167 23 case case NOUN bracis-28377 167 24 ) ) PUNCT bracis-28377 167 25 in in ADP bracis-28377 167 26 the the DET bracis-28377 167 27 identification identification NOUN bracis-28377 167 28 task task NOUN bracis-28377 167 29 , , PUNCT bracis-28377 167 30 while while SCONJ bracis-28377 167 31 yoon yoon PROPN bracis-28377 167 32 , , PUNCT bracis-28377 167 33 so so ADV bracis-28377 167 34 and and CCONJ bracis-28377 167 35 rhee rhee NOUN bracis-28377 168 1 [ [ X bracis-28377 168 2 21 21 NUM bracis-28377 168 3 ] ] PUNCT bracis-28377 168 4 , , PUNCT bracis-28377 168 5 using use VERB bracis-28377 168 6 the the DET bracis-28377 168 7 same same ADJ bracis-28377 168 8 dataset dataset NOUN bracis-28377 168 9 that that PRON bracis-28377 168 10 we we PRON bracis-28377 168 11 have have AUX bracis-28377 168 12 used use VERB bracis-28377 168 13 , , PUNCT bracis-28377 168 14 obtained obtain VERB bracis-28377 168 15 a a DET bracis-28377 168 16 mean mean ADJ bracis-28377 168 17 accuracy accuracy NOUN bracis-28377 168 18 of of ADP bracis-28377 168 19 \(88.8\%\ \(88.8\%\ NOUN bracis-28377 168 20 ) ) PUNCT bracis-28377 168 21 with with ADP bracis-28377 168 22 arcface+vl arcface+vl PRON bracis-28377 168 23 in in ADP bracis-28377 168 24 the the DET bracis-28377 168 25 verification verification NOUN bracis-28377 168 26 task task NOUN bracis-28377 168 27 ( ( PUNCT bracis-28377 168 28 fig.13 fig.13 NOUN bracis-28377 168 29 ) ) PUNCT bracis-28377 168 30 . . PUNCT bracis-28377 169 1 fig fig NOUN bracis-28377 169 2 . . PUNCT bracis-28377 170 1 13 13 NUM bracis-28377 170 2 . . PUNCT bracis-28377 171 1 cmc cmc NOUN bracis-28377 171 2 curves curve NOUN bracis-28377 171 3 obtained obtain VERB bracis-28377 171 4 on on ADP bracis-28377 171 5 dogfacenet dogfacenet NOUN bracis-28377 171 6 dataset dataset VERB bracis-28377 171 7 by by ADP bracis-28377 171 8 using use VERB bracis-28377 171 9 face face NOUN bracis-28377 171 10 features feature NOUN bracis-28377 171 11 extracted extract VERB bracis-28377 171 12 from from ADP bracis-28377 171 13 efficientformer efficientformer NOUN bracis-28377 171 14 - - PUNCT bracis-28377 171 15 l1 l1 PROPN bracis-28377 171 16 and and CCONJ bracis-28377 171 17 resnet-50 resnet-50 PROPN bracis-28377 171 18 . . PROPN bracis-28377 171 19 full full ADJ bracis-28377 171 20 size size NOUN bracis-28377 171 21 image image NOUN bracis-28377 171 22 6 6 NUM bracis-28377 171 23 conclusion conclusion NOUN bracis-28377 171 24 with with ADP bracis-28377 171 25 the the DET bracis-28377 171 26 advancement advancement NOUN bracis-28377 171 27 of of ADP bracis-28377 171 28 technology technology NOUN bracis-28377 171 29 , , PUNCT bracis-28377 171 30 especially especially ADV bracis-28377 171 31 in in ADP bracis-28377 171 32 the the DET bracis-28377 171 33 scenario scenario NOUN bracis-28377 171 34 of of ADP bracis-28377 171 35 smart smart ADJ bracis-28377 171 36 cities city NOUN bracis-28377 171 37 and and CCONJ bracis-28377 171 38 internet internet NOUN bracis-28377 171 39 of of ADP bracis-28377 171 40 things thing NOUN bracis-28377 171 41 devices device NOUN bracis-28377 171 42 , , PUNCT bracis-28377 171 43 there there PRON bracis-28377 171 44 is be VERB bracis-28377 171 45 an an DET bracis-28377 171 46 improvement improvement NOUN bracis-28377 171 47 and and CCONJ bracis-28377 171 48 , , PUNCT bracis-28377 171 49 consequently consequently ADV bracis-28377 171 50 , , PUNCT bracis-28377 171 51 sophistication sophistication NOUN bracis-28377 171 52 in in ADP bracis-28377 171 53 biometric biometric ADJ bracis-28377 171 54 applications application NOUN bracis-28377 171 55 . . PUNCT bracis-28377 172 1 with with ADP bracis-28377 172 2 this this PRON bracis-28377 172 3 , , PUNCT bracis-28377 172 4 it it PRON bracis-28377 172 5 is be AUX bracis-28377 172 6 possible possible ADJ bracis-28377 172 7 to to PART bracis-28377 172 8 explore explore VERB bracis-28377 172 9 the the DET bracis-28377 172 10 field field NOUN bracis-28377 172 11 of of ADP bracis-28377 172 12 animal animal NOUN bracis-28377 172 13 biometrics biometric NOUN bracis-28377 172 14 , , PUNCT bracis-28377 172 15 as as SCONJ bracis-28377 172 16 there there PRON bracis-28377 172 17 are be VERB bracis-28377 172 18 still still ADV bracis-28377 172 19 few few ADJ bracis-28377 172 20 applications application NOUN bracis-28377 172 21 and and CCONJ bracis-28377 172 22 studies study NOUN bracis-28377 172 23 in in ADP bracis-28377 172 24 the the DET bracis-28377 172 25 area area NOUN bracis-28377 172 26 applied apply VERB bracis-28377 172 27 to to ADP bracis-28377 172 28 dogs dog NOUN bracis-28377 172 29 , , PUNCT bracis-28377 172 30 according accord VERB bracis-28377 172 31 to to ADP bracis-28377 172 32 the the DET bracis-28377 172 33 study study NOUN bracis-28377 172 34 by by ADP bracis-28377 172 35 mougeot mougeot PROPN bracis-28377 172 36 , , PUNCT bracis-28377 172 37 li li PROPN bracis-28377 172 38 and and CCONJ bracis-28377 172 39 jia jia PROPN bracis-28377 173 1 [ [ X bracis-28377 173 2 17 17 NUM bracis-28377 173 3 ] ] PUNCT bracis-28377 173 4 . . PUNCT bracis-28377 174 1 in in ADP bracis-28377 174 2 the the DET bracis-28377 174 3 case case NOUN bracis-28377 174 4 of of ADP bracis-28377 174 5 dogs dog NOUN bracis-28377 174 6 specifically specifically ADV bracis-28377 174 7 , , PUNCT bracis-28377 174 8 biometric biometric ADJ bracis-28377 174 9 identification identification NOUN bracis-28377 174 10 through through ADP bracis-28377 174 11 faces face NOUN bracis-28377 174 12 has have AUX bracis-28377 174 13 shown show VERB bracis-28377 174 14 to to PART bracis-28377 174 15 be be AUX bracis-28377 174 16 very very ADV bracis-28377 174 17 promising promising ADJ bracis-28377 174 18 , , PUNCT bracis-28377 174 19 as as SCONJ bracis-28377 174 20 it it PRON bracis-28377 174 21 does do AUX bracis-28377 174 22 not not PART bracis-28377 174 23 require require VERB bracis-28377 174 24 specific specific ADJ bracis-28377 174 25 hardware hardware NOUN bracis-28377 174 26 resources resource NOUN bracis-28377 174 27 and and CCONJ bracis-28377 174 28 can can AUX bracis-28377 174 29 be be AUX bracis-28377 174 30 used use VERB bracis-28377 174 31 through through ADP bracis-28377 174 32 smartphone smartphone NOUN bracis-28377 174 33 cameras camera NOUN bracis-28377 174 34 . . PUNCT bracis-28377 175 1 as as ADP bracis-28377 175 2 for for ADP bracis-28377 175 3 the the DET bracis-28377 175 4 recognition recognition NOUN bracis-28377 175 5 techniques technique NOUN bracis-28377 175 6 , , PUNCT bracis-28377 175 7 it it PRON bracis-28377 175 8 is be AUX bracis-28377 175 9 noted note VERB bracis-28377 175 10 that that SCONJ bracis-28377 175 11 the the DET bracis-28377 175 12 application application NOUN bracis-28377 175 13 of of ADP bracis-28377 175 14 architectures architecture NOUN bracis-28377 175 15 based base VERB bracis-28377 175 16 on on ADP bracis-28377 175 17 transformers transformer NOUN bracis-28377 175 18 , , PUNCT bracis-28377 175 19 especially especially ADV bracis-28377 175 20 with with ADP bracis-28377 175 21 efficientformer efficientformer NOUN bracis-28377 175 22 - - PUNCT bracis-28377 175 23 l1 l1 PROPN bracis-28377 175 24 used use VERB bracis-28377 175 25 in in ADP bracis-28377 175 26 this this DET bracis-28377 175 27 work work NOUN bracis-28377 175 28 , , PUNCT bracis-28377 175 29 brings bring VERB bracis-28377 175 30 excellent excellent ADJ bracis-28377 175 31 results result NOUN bracis-28377 175 32 compared compare VERB bracis-28377 175 33 to to ADP bracis-28377 175 34 the the DET bracis-28377 175 35 state state NOUN bracis-28377 175 36 of of ADP bracis-28377 175 37 the the DET bracis-28377 175 38 art art NOUN bracis-28377 175 39 in in ADP bracis-28377 175 40 the the DET bracis-28377 175 41 area area NOUN bracis-28377 175 42 of of ADP bracis-28377 175 43 pattern pattern NOUN bracis-28377 175 44 recognition recognition NOUN bracis-28377 175 45 in in ADP bracis-28377 175 46 images image NOUN bracis-28377 175 47 that that PRON bracis-28377 175 48 are be AUX bracis-28377 175 49 the the DET bracis-28377 175 50 convolutional convolutional ADJ bracis-28377 175 51 neural neural ADJ bracis-28377 175 52 networks network NOUN bracis-28377 175 53 , , PUNCT bracis-28377 175 54 in in ADP bracis-28377 175 55 this this DET bracis-28377 175 56 case case NOUN bracis-28377 175 57 from from ADP bracis-28377 175 58 resnet-50 resnet-50 PROPN bracis-28377 175 59 . . PUNCT bracis-28377 176 1 still still ADV bracis-28377 176 2 , , PUNCT bracis-28377 176 3 the the DET bracis-28377 176 4 application application NOUN bracis-28377 176 5 of of ADP bracis-28377 176 6 biometric biometric ADJ bracis-28377 176 7 techniques technique NOUN bracis-28377 176 8 in in ADP bracis-28377 176 9 dogs dog NOUN bracis-28377 176 10 is be AUX bracis-28377 176 11 a a DET bracis-28377 176 12 vast vast ADJ bracis-28377 176 13 area area NOUN bracis-28377 176 14 to to PART bracis-28377 176 15 be be AUX bracis-28377 176 16 explored explore VERB bracis-28377 176 17 and and CCONJ bracis-28377 176 18 its its PRON bracis-28377 176 19 results result NOUN bracis-28377 176 20 add add VERB bracis-28377 176 21 positively positively ADV bracis-28377 176 22 to to ADP bracis-28377 176 23 the the DET bracis-28377 176 24 population population NOUN bracis-28377 176 25 , , PUNCT bracis-28377 176 26 because because SCONJ bracis-28377 176 27 , , PUNCT bracis-28377 176 28 through through ADP bracis-28377 176 29 this this PRON bracis-28377 176 30 , , PUNCT bracis-28377 176 31 it it PRON bracis-28377 176 32 provides provide VERB bracis-28377 176 33 ways way NOUN bracis-28377 176 34 of of ADP bracis-28377 176 35 monitoring monitor VERB bracis-28377 176 36 these these DET bracis-28377 176 37 animals animal NOUN bracis-28377 176 38 in in ADP bracis-28377 176 39 cities city NOUN bracis-28377 176 40 , , PUNCT bracis-28377 176 41 searching search VERB bracis-28377 176 42 for for ADP bracis-28377 176 43 lost lose VERB bracis-28377 176 44 animals animal NOUN bracis-28377 176 45 , , PUNCT bracis-28377 176 46 reducing reduce VERB bracis-28377 176 47 abandonment abandonment NOUN bracis-28377 176 48 , , PUNCT bracis-28377 176 49 carrying carry VERB bracis-28377 176 50 out out ADP bracis-28377 176 51 disease disease NOUN bracis-28377 176 52 control control NOUN bracis-28377 176 53 more more ADV bracis-28377 176 54 effectively effectively ADV bracis-28377 176 55 and and CCONJ bracis-28377 176 56 help help VERB bracis-28377 176 57 prevent prevent VERB bracis-28377 176 58 or or CCONJ bracis-28377 176 59 detect detect VERB bracis-28377 176 60 fraud fraud NOUN bracis-28377 176 61 . . PUNCT bracis-28377 177 1 references reference NOUN bracis-28377 177 2 chollet chollet VERB bracis-28377 177 3 , , PUNCT bracis-28377 177 4 f. f. PROPN bracis-28377 177 5 : : PUNCT bracis-28377 177 6 how how SCONJ bracis-28377 177 7 convolutional convolutional ADJ bracis-28377 177 8 neural neural ADJ bracis-28377 177 9 networks network NOUN bracis-28377 177 10 see see VERB bracis-28377 177 11 the the DET bracis-28377 177 12 world world NOUN bracis-28377 177 13 . . PUNCT bracis-28377 178 1 the the DET bracis-28377 178 2 keras keras PROPN bracis-28377 178 3 blog blog NOUN bracis-28377 178 4 30 30 NUM bracis-28377 178 5 ( ( PUNCT bracis-28377 178 6 2016 2016 NUM bracis-28377 178 7 ) ) PUNCT bracis-28377 178 8 google google PROPN bracis-28377 178 9 scholar scholar NOUN bracis-28377 178 10     SPACE bracis-28377 178 11 de de X bracis-28377 178 12 souza souza PROPN bracis-28377 178 13 , , PUNCT bracis-28377 178 14 g.b g.b PROPN bracis-28377 178 15 . . PROPN bracis-28377 178 16 , , PUNCT bracis-28377 178 17 da da PROPN bracis-28377 178 18 silva silva PROPN bracis-28377 178 19 santos santos PROPN bracis-28377 178 20 , , PUNCT bracis-28377 178 21 d.f d.f PROPN bracis-28377 178 22 . . PROPN bracis-28377 178 23 , , PUNCT bracis-28377 178 24 pires pire NOUN bracis-28377 178 25 , , PUNCT bracis-28377 178 26 r.g r.g PROPN bracis-28377 178 27 . . PROPN bracis-28377 178 28 , , PUNCT bracis-28377 178 29 marana marana PROPN bracis-28377 178 30 , , PUNCT bracis-28377 178 31 a.n a.n PROPN bracis-28377 178 32 . . PROPN bracis-28377 178 33 , , PUNCT bracis-28377 178 34 papa papa PROPN bracis-28377 178 35 , , PUNCT bracis-28377 178 36 j.p j.p PROPN bracis-28377 178 37 . . PROPN bracis-28377 178 38 : : PUNCT bracis-28377 179 1 deep deep ADJ bracis-28377 179 2 texture texture NOUN bracis-28377 179 3 features feature VERB bracis-28377 179 4 for for ADP bracis-28377 179 5 robust robust ADJ bracis-28377 179 6 face face NOUN bracis-28377 179 7 spoofing spoof VERB bracis-28377 179 8 detection detection NOUN bracis-28377 179 9 . . PUNCT bracis-28377 180 1 ieee ieee PROPN bracis-28377 180 2 trans trans PROPN bracis-28377 180 3 . . PUNCT bracis-28377 181 1 circuits circuit NOUN bracis-28377 181 2 syst syst PROPN bracis-28377 181 3 . . PUNCT bracis-28377 182 1 ii ii PROPN bracis-28377 182 2 express express VERB bracis-28377 182 3 briefs brief NOUN bracis-28377 182 4 64(12 64(12 NUM bracis-28377 182 5 ) ) PUNCT bracis-28377 182 6 , , PUNCT bracis-28377 182 7 1397–1401 1397–1401 NUM bracis-28377 182 8 ( ( PUNCT bracis-28377 182 9 2017 2017 NUM bracis-28377 182 10 ) ) PUNCT bracis-28377 182 11 google google PROPN bracis-28377 182 12 scholar scholar NOUN bracis-28377 182 13     SPACE bracis-28377 182 14 deng deng PROPN bracis-28377 182 15 , , PUNCT bracis-28377 182 16 j. j. PROPN bracis-28377 182 17 , , PUNCT bracis-28377 182 18 guo guo PROPN bracis-28377 182 19 , , PUNCT bracis-28377 182 20 j. j. PROPN bracis-28377 182 21 , , PUNCT bracis-28377 182 22 liu liu PROPN bracis-28377 182 23 , , PUNCT bracis-28377 182 24 t. t. PROPN bracis-28377 182 25 , , PUNCT bracis-28377 182 26 gong gong PROPN bracis-28377 182 27 , , PUNCT bracis-28377 182 28 m. m. NOUN bracis-28377 182 29 , , PUNCT bracis-28377 182 30 zafeiriou zafeiriou NOUN bracis-28377 182 31 , , PUNCT bracis-28377 182 32 s. s. PROPN bracis-28377 182 33 : : PUNCT bracis-28377 182 34 sub sub ADJ bracis-28377 182 35 - - ADJ bracis-28377 182 36 center center ADJ bracis-28377 182 37 arcface arcface NOUN bracis-28377 182 38 : : PUNCT bracis-28377 182 39 boosting boost VERB bracis-28377 182 40 face face NOUN bracis-28377 182 41 recognition recognition NOUN bracis-28377 182 42 by by ADP bracis-28377 182 43 large large ADJ bracis-28377 182 44 - - PUNCT bracis-28377 182 45 scale scale NOUN bracis-28377 182 46 noisy noisy ADJ bracis-28377 182 47 web web NOUN bracis-28377 182 48 faces face NOUN bracis-28377 182 49 . . PUNCT bracis-28377 183 1 in in ADP bracis-28377 183 2 : : PUNCT bracis-28377 183 3 vedaldi vedaldi PROPN bracis-28377 183 4 , , PUNCT bracis-28377 183 5 a. a. NOUN bracis-28377 183 6 , , PUNCT bracis-28377 183 7 bischof bischof NOUN bracis-28377 183 8 , , PUNCT bracis-28377 183 9 h. h. PROPN bracis-28377 183 10 , , PUNCT bracis-28377 183 11 brox brox PROPN bracis-28377 183 12 , , PUNCT bracis-28377 183 13 t. t. PROPN bracis-28377 183 14 , , PUNCT bracis-28377 183 15 frahm frahm NOUN bracis-28377 183 16 , , PUNCT bracis-28377 183 17 j.-m j.-m NOUN bracis-28377 183 18 . . PUNCT bracis-28377 184 1 ( ( PUNCT bracis-28377 184 2 eds ed NOUN bracis-28377 184 3 . . PUNCT bracis-28377 184 4 ) ) PUNCT bracis-28377 184 5 eccv eccv ADV bracis-28377 184 6 2020 2020 NUM bracis-28377 184 7 . . PUNCT bracis-28377 185 1 lncs lncs PROPN bracis-28377 185 2 , , PUNCT bracis-28377 185 3 vol vol NOUN bracis-28377 185 4 . . PROPN bracis-28377 185 5 12356 12356 NUM bracis-28377 185 6 , , PUNCT bracis-28377 185 7 pp pp ADP bracis-28377 185 8 . . PUNCT bracis-28377 186 1 741–757 741–757 NOUN bracis-28377 186 2 . . PUNCT bracis-28377 186 3 springer springer NOUN bracis-28377 186 4 , , PUNCT bracis-28377 186 5 cham cham PROPN bracis-28377 186 6 ( ( PUNCT bracis-28377 186 7 2020 2020 NUM bracis-28377 186 8 ) ) PUNCT bracis-28377 186 9 . . PUNCT bracis-28377 187 1 https://doi.org/10.1007/978-3-030-58621-8_43 https://doi.org/10.1007/978-3-030-58621-8_43 PROPN bracis-28377 187 2 chapter chapter NOUN bracis-28377 187 3     SPACE bracis-28377 187 4 google google PROPN bracis-28377 187 5 scholar scholar NOUN bracis-28377 187 6     SPACE bracis-28377 187 7 deng deng PROPN bracis-28377 187 8 , , PUNCT bracis-28377 187 9 j. j. PROPN bracis-28377 187 10 , , PUNCT bracis-28377 187 11 guo guo PROPN bracis-28377 187 12 , , PUNCT bracis-28377 187 13 j. j. PROPN bracis-28377 187 14 , , PUNCT bracis-28377 187 15 xue xue PROPN bracis-28377 187 16 , , PUNCT bracis-28377 187 17 n. n. NOUN bracis-28377 187 18 , , PUNCT bracis-28377 187 19 zafeiriou zafeiriou NOUN bracis-28377 187 20 , , PUNCT bracis-28377 187 21 s. s. PROPN bracis-28377 187 22 : : PUNCT bracis-28377 187 23 arcface arcface NOUN bracis-28377 187 24 : : PUNCT bracis-28377 187 25 additive additive ADJ bracis-28377 187 26 angular angular ADJ bracis-28377 187 27 margin margin NOUN bracis-28377 187 28 loss loss NOUN bracis-28377 187 29 for for ADP bracis-28377 187 30 deep deep ADJ bracis-28377 187 31 face face NOUN bracis-28377 187 32 recognition recognition NOUN bracis-28377 187 33 . . PUNCT bracis-28377 188 1 in in ADP bracis-28377 188 2 : : PUNCT bracis-28377 188 3 proceedings proceeding NOUN bracis-28377 188 4 of of ADP bracis-28377 188 5 the the DET bracis-28377 188 6 ieee ieee NOUN bracis-28377 188 7 / / SYM bracis-28377 188 8 cvf cvf NOUN bracis-28377 188 9 conference conference NOUN bracis-28377 188 10 on on ADP bracis-28377 188 11 computer computer NOUN bracis-28377 188 12 vision vision NOUN bracis-28377 188 13 and and CCONJ bracis-28377 188 14 pattern pattern NOUN bracis-28377 188 15 recognition recognition NOUN bracis-28377 188 16 , , PUNCT bracis-28377 188 17 pp pp X bracis-28377 188 18 . . PUNCT bracis-28377 189 1 4690–4699 4690–4699 NUM bracis-28377 189 2 ( ( PUNCT bracis-28377 189 3 2019 2019 NUM bracis-28377 189 4 ) ) PUNCT bracis-28377 189 5 google google NOUN bracis-28377 189 6 scholar scholar NOUN bracis-28377 189 7     SPACE bracis-28377 189 8 dosovitskiy dosovitskiy NOUN bracis-28377 189 9 , , PUNCT bracis-28377 189 10 a. a. NOUN bracis-28377 189 11 , , PUNCT bracis-28377 189 12 et et PROPN bracis-28377 189 13 al al PROPN bracis-28377 189 14 . . PUNCT bracis-28377 189 15 : : PUNCT bracis-28377 190 1 an an DET bracis-28377 190 2 image image NOUN bracis-28377 190 3 is be AUX bracis-28377 190 4 worth worth ADJ bracis-28377 190 5 16x16 16x16 NUM bracis-28377 190 6 words word NOUN bracis-28377 190 7 : : PUNCT bracis-28377 190 8 transformers transformer NOUN bracis-28377 190 9 for for ADP bracis-28377 190 10 image image NOUN bracis-28377 190 11 recognition recognition NOUN bracis-28377 190 12 at at ADP bracis-28377 190 13 scale scale NOUN bracis-28377 190 14 . . PUNCT bracis-28377 191 1 arxiv arxiv PROPN bracis-28377 191 2 preprint preprint VERB bracis-28377 191 3 arxiv:2010.11929 arxiv:2010.11929 NOUN bracis-28377 191 4 ( ( PUNCT bracis-28377 191 5 2020 2020 NUM bracis-28377 191 6 ) ) PUNCT bracis-28377 191 7 forsyth forsyth NOUN bracis-28377 191 8 , , PUNCT bracis-28377 191 9 d.a d.a PROPN bracis-28377 191 10 . . PROPN bracis-28377 191 11 , , PUNCT bracis-28377 191 12 et et PROPN bracis-28377 191 13 al al PROPN bracis-28377 191 14 . . PUNCT bracis-28377 191 15 : : PUNCT bracis-28377 192 1 object object VERB bracis-28377 192 2 recognition recognition NOUN bracis-28377 192 3 with with ADP bracis-28377 192 4 gradient gradient NOUN bracis-28377 192 5 - - PUNCT bracis-28377 192 6 based base VERB bracis-28377 192 7 learning learning NOUN bracis-28377 192 8 . . PUNCT bracis-28377 193 1 shape shape NOUN bracis-28377 193 2 , , PUNCT bracis-28377 193 3 contour contour NOUN bracis-28377 193 4 and and CCONJ bracis-28377 193 5 grouping group VERB bracis-28377 193 6 in in ADP bracis-28377 193 7 computer computer NOUN bracis-28377 193 8 vision vision NOUN bracis-28377 193 9 , , PUNCT bracis-28377 193 10 pp pp ADV bracis-28377 193 11 . . PUNCT bracis-28377 194 1 319–345 319–345 NUM bracis-28377 194 2 ( ( PUNCT bracis-28377 194 3 1999 1999 NUM bracis-28377 194 4 ) ) PUNCT bracis-28377 194 5 google google PROPN bracis-28377 194 6 scholar scholar NOUN bracis-28377 194 7     SPACE bracis-28377 194 8 geeksforgeeks geeksforgeek NOUN bracis-28377 194 9 : : PUNCT bracis-28377 194 10 residual residual ADJ bracis-28377 194 11 networks network NOUN bracis-28377 194 12 ( ( PUNCT bracis-28377 194 13 resnet resnet NOUN bracis-28377 194 14 ) ) PUNCT bracis-28377 194 15 deep deep ADJ bracis-28377 194 16 learning learning NOUN bracis-28377 194 17 . . PUNCT bracis-28377 195 1 https://www.geeksforgeeks.org/residual-networks-resnet-deep-learning/. https://www.geeksforgeeks.org/residual-networks-resnet-deep-learning/. PROPN bracis-28377 195 2 accessed access VERB bracis-28377 195 3 18 18 NUM bracis-28377 195 4 june june PROPN bracis-28377 195 5 2022 2022 NUM bracis-28377 195 6 he he PRON bracis-28377 195 7 , , PUNCT bracis-28377 195 8 k. k. PROPN bracis-28377 195 9 , , PUNCT bracis-28377 195 10 zhang zhang PROPN bracis-28377 195 11 , , PUNCT bracis-28377 195 12 x. x. PROPN bracis-28377 195 13 , , PUNCT bracis-28377 195 14 ren ren PROPN bracis-28377 195 15 , , PUNCT bracis-28377 195 16 s. s. PROPN bracis-28377 195 17 , , PUNCT bracis-28377 195 18 sun sun PROPN bracis-28377 195 19 , , PUNCT bracis-28377 195 20 j. j. PROPN bracis-28377 195 21 : : PUNCT bracis-28377 195 22 identity identity NOUN bracis-28377 195 23 mappings mapping NOUN bracis-28377 195 24 in in ADP bracis-28377 195 25 deep deep ADJ bracis-28377 195 26 residual residual ADJ bracis-28377 195 27 networks network NOUN bracis-28377 195 28 . . PUNCT bracis-28377 196 1 in in ADP bracis-28377 196 2 : : PUNCT bracis-28377 196 3 leibe leibe ADJ bracis-28377 196 4 , , PUNCT bracis-28377 196 5 b. b. PROPN bracis-28377 196 6 , , PUNCT bracis-28377 196 7 matas matas PROPN bracis-28377 196 8 , , PUNCT bracis-28377 196 9 j. j. PROPN bracis-28377 196 10 , , PUNCT bracis-28377 196 11 sebe sebe PROPN bracis-28377 196 12 , , PUNCT bracis-28377 196 13 n. n. NOUN bracis-28377 196 14 , , PUNCT bracis-28377 196 15 welling well VERB bracis-28377 196 16 , , PUNCT bracis-28377 196 17 m. m. NOUN bracis-28377 196 18 ( ( PUNCT bracis-28377 196 19 eds ed NOUN bracis-28377 196 20 . . PUNCT bracis-28377 196 21 ) ) PUNCT bracis-28377 196 22 eccv eccv ADV bracis-28377 196 23 2016 2016 NUM bracis-28377 196 24 . . PUNCT bracis-28377 197 1 lncs lncs PROPN bracis-28377 197 2 , , PUNCT bracis-28377 197 3 vol vol NOUN bracis-28377 197 4 . . PROPN bracis-28377 197 5 9908 9908 NUM bracis-28377 197 6 , , PUNCT bracis-28377 197 7 pp pp ADP bracis-28377 197 8 . . PUNCT bracis-28377 198 1 630–645 630–645 NUM bracis-28377 198 2 . . PUNCT bracis-28377 198 3 springer springer NOUN bracis-28377 198 4 , , PUNCT bracis-28377 198 5 cham cham PROPN bracis-28377 198 6 ( ( PUNCT bracis-28377 198 7 2016 2016 NUM bracis-28377 198 8 ) ) PUNCT bracis-28377 198 9 . . PUNCT bracis-28377 199 1 https://doi.org/10.1007/978-3-319-46493-0_38 https://doi.org/10.1007/978-3-319-46493-0_38 PROPN bracis-28377 199 2 chapter chapter NOUN bracis-28377 199 3     SPACE bracis-28377 199 4 google google PROPN bracis-28377 199 5 scholar scholar NOUN bracis-28377 199 6     SPACE bracis-28377 199 7 institute institute NOUN bracis-28377 199 8 , , PUNCT bracis-28377 199 9 p.b p.b PROPN bracis-28377 199 10 . . PROPN bracis-28377 199 11 : : PUNCT bracis-28377 199 12 pet pet ADJ bracis-28377 199 13 census census NOUN bracis-28377 199 14 ( ( PUNCT bracis-28377 199 15 2019 2019 NUM bracis-28377 199 16 ) ) PUNCT bracis-28377 199 17 . . PUNCT bracis-28377 200 1 https://institutopetbrasil.com/imprensa/censo-pet-1393-milhoes-de-animais-de-estimacao-no-brasil https://institutopetbrasil.com/imprensa/censo-pet-1393-milhoes-de-animais-de-estimacao-no-brasil NOUN bracis-28377 200 2 . . PUNCT bracis-28377 201 1 accessed access VERB bracis-28377 201 2 18 18 NUM bracis-28377 201 3 june june PROPN bracis-28377 201 4 2022 2022 NUM bracis-28377 201 5 jang jang PROPN bracis-28377 201 6 , , PUNCT bracis-28377 201 7 d.h d.h PROPN bracis-28377 201 8 . . PROPN bracis-28377 201 9 , , PUNCT bracis-28377 201 10 kwon kwon PROPN bracis-28377 201 11 , , PUNCT bracis-28377 201 12 k.s k.s PROPN bracis-28377 201 13 . . PROPN bracis-28377 201 14 , , PUNCT bracis-28377 201 15 kim kim PROPN bracis-28377 201 16 , , PUNCT bracis-28377 201 17 j.k j.k PROPN bracis-28377 201 18 . . PROPN bracis-28377 201 19 , , PUNCT bracis-28377 201 20 yang yang PROPN bracis-28377 201 21 , , PUNCT bracis-28377 201 22 k.y k.y PROPN bracis-28377 201 23 . . PROPN bracis-28377 201 24 , , PUNCT bracis-28377 201 25 kim kim PROPN bracis-28377 201 26 , , PUNCT bracis-28377 201 27 j.b j.b PROPN bracis-28377 201 28 . . PUNCT bracis-28377 201 29 : : PUNCT bracis-28377 202 1 dog dog NOUN bracis-28377 202 2 identification identification NOUN bracis-28377 202 3 method method NOUN bracis-28377 202 4 based base VERB bracis-28377 202 5 on on ADP bracis-28377 202 6 muzzle muzzle NOUN bracis-28377 202 7 pattern pattern NOUN bracis-28377 202 8 image image NOUN bracis-28377 202 9 . . PUNCT bracis-28377 203 1 appl appl PROPN bracis-28377 203 2 . . PUNCT bracis-28377 204 1 sci sci PROPN bracis-28377 204 2 . . PROPN bracis-28377 204 3 10(24 10(24 NUM bracis-28377 204 4 ) ) PUNCT bracis-28377 204 5 , , PUNCT bracis-28377 204 6 8994 8994 NUM bracis-28377 204 7 ( ( PUNCT bracis-28377 204 8 2020 2020 NUM bracis-28377 204 9 ) ) PUNCT bracis-28377 204 10 article article NOUN bracis-28377 204 11     SPACE bracis-28377 204 12 google google PROPN bracis-28377 204 13 scholar scholar NOUN bracis-28377 204 14     SPACE bracis-28377 204 15 kumar kumar PROPN bracis-28377 204 16 , , PUNCT bracis-28377 204 17 s. s. PROPN bracis-28377 204 18 , , PUNCT bracis-28377 204 19 singh singh PROPN bracis-28377 204 20 , , PUNCT bracis-28377 204 21 s.k s.k PROPN bracis-28377 204 22 . . PROPN bracis-28377 204 23 : : PUNCT bracis-28377 204 24 visual visual ADJ bracis-28377 204 25 animal animal NOUN bracis-28377 204 26 biometrics biometric NOUN bracis-28377 204 27 : : PUNCT bracis-28377 204 28 survey survey NOUN bracis-28377 204 29 . . PUNCT bracis-28377 205 1 iet iet PROPN bracis-28377 205 2 . . PUNCT bracis-28377 206 1 biometrics biometrics PROPN bracis-28377 206 2 6(3 6(3 NUM bracis-28377 206 3 ) ) PUNCT bracis-28377 206 4 , , PUNCT bracis-28377 206 5 139–156 139–156 NUM bracis-28377 206 6 ( ( PUNCT bracis-28377 206 7 2017 2017 NUM bracis-28377 206 8 ) ) PUNCT bracis-28377 206 9 google google PROPN bracis-28377 206 10 scholar scholar NOUN bracis-28377 206 11     SPACE bracis-28377 206 12 lai lai PROPN bracis-28377 206 13 , , PUNCT bracis-28377 206 14 k. k. PROPN bracis-28377 206 15 , , PUNCT bracis-28377 206 16 tu tu PROPN bracis-28377 206 17 , , PUNCT bracis-28377 206 18 x. x. PROPN bracis-28377 206 19 , , PUNCT bracis-28377 206 20 yanushkevich yanushkevich PROPN bracis-28377 206 21 , , PUNCT bracis-28377 206 22 s. s. PROPN bracis-28377 206 23 : : PUNCT bracis-28377 206 24 dog dog NOUN bracis-28377 206 25 identification identification NOUN bracis-28377 206 26 using use VERB bracis-28377 206 27 soft soft ADJ bracis-28377 206 28 biometrics biometric NOUN bracis-28377 206 29 and and CCONJ bracis-28377 206 30 neural neural ADJ bracis-28377 206 31 networks network NOUN bracis-28377 206 32 . . PUNCT bracis-28377 207 1 in in ADP bracis-28377 207 2 : : PUNCT bracis-28377 207 3 2019 2019 NUM bracis-28377 207 4 international international ADJ bracis-28377 207 5 joint joint ADJ bracis-28377 207 6 conference conference NOUN bracis-28377 207 7 on on ADP bracis-28377 207 8 neural neural ADJ bracis-28377 207 9 networks network NOUN bracis-28377 207 10 ( ( PUNCT bracis-28377 207 11 ijcnn ijcnn PROPN bracis-28377 207 12 ) ) PUNCT bracis-28377 207 13 , , PUNCT bracis-28377 207 14 pp pp ADP bracis-28377 207 15 . . PUNCT bracis-28377 208 1 1–8 1–8 X bracis-28377 208 2 . . PUNCT bracis-28377 208 3 ieee ieee PROPN bracis-28377 208 4 ( ( PUNCT bracis-28377 208 5 2019 2019 NUM bracis-28377 208 6 ) ) PUNCT bracis-28377 208 7 google google PROPN bracis-28377 208 8 scholar scholar NOUN bracis-28377 208 9     SPACE bracis-28377 208 10 lecun lecun ADJ bracis-28377 208 11 , , PUNCT bracis-28377 208 12 y. y. PROPN bracis-28377 208 13 , , PUNCT bracis-28377 208 14 bottou bottou PROPN bracis-28377 208 15 , , PUNCT bracis-28377 208 16 l. l. PROPN bracis-28377 208 17 , , PUNCT bracis-28377 208 18 bengio bengio PROPN bracis-28377 208 19 , , PUNCT bracis-28377 208 20 y. y. PROPN bracis-28377 208 21 , , PUNCT bracis-28377 208 22 haffner haffner NOUN bracis-28377 208 23 , , PUNCT bracis-28377 208 24 p. p. NOUN bracis-28377 208 25 : : PUNCT bracis-28377 208 26 gradient gradient NOUN bracis-28377 208 27 - - PUNCT bracis-28377 208 28 based base VERB bracis-28377 208 29 learning learning NOUN bracis-28377 208 30 applied apply VERB bracis-28377 208 31 to to ADP bracis-28377 208 32 document document NOUN bracis-28377 208 33 recognition recognition NOUN bracis-28377 208 34 . . PUNCT bracis-28377 209 1 proc proc PROPN bracis-28377 209 2 . . PUNCT bracis-28377 210 1 ieee ieee PROPN bracis-28377 210 2 86(11 86(11 NUM bracis-28377 210 3 ) ) PUNCT bracis-28377 210 4 , , PUNCT bracis-28377 210 5 2278–2324 2278–2324 NUM bracis-28377 210 6 ( ( PUNCT bracis-28377 210 7 1998 1998 NUM bracis-28377 210 8 ) ) PUNCT bracis-28377 210 9 article article NOUN bracis-28377 210 10     SPACE bracis-28377 210 11 google google PROPN bracis-28377 210 12 scholar scholar NOUN bracis-28377 210 13     SPACE bracis-28377 210 14 lemos lemos PROPN bracis-28377 210 15 , , PUNCT bracis-28377 210 16 s. s. PROPN bracis-28377 210 17 : : PUNCT bracis-28377 210 18 the the DET bracis-28377 210 19 number number NOUN bracis-28377 210 20 of of ADP bracis-28377 210 21 adoptions adoption NOUN bracis-28377 210 22 and and CCONJ bracis-28377 210 23 abandonment abandonment NOUN bracis-28377 210 24 of of ADP bracis-28377 210 25 animals animal NOUN bracis-28377 210 26 in in ADP bracis-28377 210 27 the the DET bracis-28377 210 28 pandemic pandemic ADJ bracis-28377 210 29 ( ( PUNCT bracis-28377 210 30 2021 2021 NUM bracis-28377 210 31 ) ) PUNCT bracis-28377 210 32 . . PUNCT bracis-28377 211 1 https://jornal.usp.br/atualidades/cresce-o-numero-de-adocoes-e-de-abandono-de-animais-na-pandemia https://jornal.usp.br/atualidades/cresce-o-numero-de-adocoes-e-de-abandono-de-animais-na-pandemia NOUN bracis-28377 211 2 . . PUNCT bracis-28377 212 1 accessed access VERB bracis-28377 212 2 18 18 NUM bracis-28377 212 3 june june PROPN bracis-28377 212 4 2022 2022 NUM bracis-28377 212 5 li li PROPN bracis-28377 212 6 , , PUNCT bracis-28377 212 7 s. s. PROPN bracis-28377 212 8 , , PUNCT bracis-28377 212 9 jiao jiao PROPN bracis-28377 212 10 , , PUNCT bracis-28377 212 11 j. j. PROPN bracis-28377 212 12 , , PUNCT bracis-28377 212 13 han han PROPN bracis-28377 212 14 , , PUNCT bracis-28377 212 15 y. y. PROPN bracis-28377 212 16 , , PUNCT bracis-28377 212 17 weissman weissman PROPN bracis-28377 212 18 , , PUNCT bracis-28377 212 19 t. t. PROPN bracis-28377 212 20 : : PUNCT bracis-28377 212 21 demystifying demystify VERB bracis-28377 212 22 resnet resnet NOUN bracis-28377 212 23 . . PUNCT bracis-28377 213 1 arxiv arxiv PROPN bracis-28377 213 2 preprint preprint VERB bracis-28377 213 3 arxiv:1611.01186 arxiv:1611.01186 ADJ bracis-28377 213 4 ( ( PUNCT bracis-28377 213 5 2016 2016 NUM bracis-28377 213 6 ) ) PUNCT bracis-28377 213 7 li li PROPN bracis-28377 213 8 , , PUNCT bracis-28377 213 9 y. y. PROPN bracis-28377 213 10 , , PUNCT bracis-28377 213 11 yuan yuan PROPN bracis-28377 213 12 , , PUNCT bracis-28377 213 13 g. g. PROPN bracis-28377 213 14 , , PUNCT bracis-28377 213 15 wen wen PROPN bracis-28377 213 16 , , PUNCT bracis-28377 213 17 y. y. PROPN bracis-28377 213 18 , , PUNCT bracis-28377 213 19 hu hu PROPN bracis-28377 213 20 , , PUNCT bracis-28377 213 21 j. j. PROPN bracis-28377 213 22 , , PUNCT bracis-28377 213 23 evangelidis evangelidis PROPN bracis-28377 213 24 , , PUNCT bracis-28377 213 25 g. g. NOUN bracis-28377 213 26 , , PUNCT bracis-28377 213 27 tulyakov tulyakov PROPN bracis-28377 213 28 , , PUNCT bracis-28377 213 29 s. s. PROPN bracis-28377 213 30 , , PUNCT bracis-28377 213 31 wang wang PROPN bracis-28377 213 32 , , PUNCT bracis-28377 213 33 y. y. PROPN bracis-28377 213 34 , , PUNCT bracis-28377 213 35 ren ren PROPN bracis-28377 213 36 , , PUNCT bracis-28377 213 37 j. j. PROPN bracis-28377 213 38 : : PUNCT bracis-28377 213 39 efficientformer efficientformer NOUN bracis-28377 213 40 : : PUNCT bracis-28377 213 41 vision vision NOUN bracis-28377 213 42 transformers transformer NOUN bracis-28377 213 43 at at ADP bracis-28377 213 44 mobilenet mobilenet NOUN bracis-28377 213 45 speed speed NOUN bracis-28377 213 46 . . PUNCT bracis-28377 214 1 adv adv PROPN bracis-28377 214 2 . . PUNCT bracis-28377 214 3 neural neural PROPN bracis-28377 214 4 . . PUNCT bracis-28377 214 5 inf inf PROPN bracis-28377 214 6 . . PROPN bracis-28377 214 7 process process NOUN bracis-28377 214 8 . . PUNCT bracis-28377 215 1 syst syst PROPN bracis-28377 215 2 . . PROPN bracis-28377 216 1 35 35 NUM bracis-28377 216 2 , , PUNCT bracis-28377 216 3 12934–12949 12934–12949 NUM bracis-28377 216 4 ( ( PUNCT bracis-28377 216 5 2022 2022 NUM bracis-28377 216 6 ) ) PUNCT bracis-28377 216 7 google google PROPN bracis-28377 216 8 scholar scholar NOUN bracis-28377 216 9     SPACE bracis-28377 216 10 mougeot mougeot NOUN bracis-28377 216 11 , , PUNCT bracis-28377 216 12 g. g. PROPN bracis-28377 216 13 , , PUNCT bracis-28377 216 14 li li PROPN bracis-28377 216 15 , , PUNCT bracis-28377 216 16 d. d. PROPN bracis-28377 216 17 , , PUNCT bracis-28377 216 18 jia jia PROPN bracis-28377 216 19 , , PUNCT bracis-28377 216 20 s. s. PROPN bracis-28377 216 21 : : PUNCT bracis-28377 216 22 a a DET bracis-28377 216 23 deep deep ADJ bracis-28377 216 24 learning learning NOUN bracis-28377 216 25 approach approach NOUN bracis-28377 216 26 for for ADP bracis-28377 216 27 dog dog NOUN bracis-28377 216 28 face face NOUN bracis-28377 216 29 verification verification NOUN bracis-28377 216 30 and and CCONJ bracis-28377 216 31 recognition recognition NOUN bracis-28377 216 32 . . PUNCT bracis-28377 217 1 in in ADP bracis-28377 217 2 : : PUNCT bracis-28377 217 3 nayak nayak PROPN bracis-28377 217 4 , , PUNCT bracis-28377 217 5 a.c a.c PROPN bracis-28377 217 6 . . PROPN bracis-28377 217 7 , , PUNCT bracis-28377 217 8 sharma sharma PROPN bracis-28377 217 9 , , PUNCT bracis-28377 217 10 a. a. NOUN bracis-28377 217 11 ( ( PUNCT bracis-28377 217 12 eds eds PROPN bracis-28377 217 13 . . PUNCT bracis-28377 217 14 ) ) PUNCT bracis-28377 217 15 pricai pricai PROPN bracis-28377 217 16 2019 2019 NUM bracis-28377 217 17 . . PUNCT bracis-28377 218 1 lncs lncs PROPN bracis-28377 218 2 ( ( PUNCT bracis-28377 218 3 lnai lnai ADJ bracis-28377 218 4 ) ) PUNCT bracis-28377 218 5 , , PUNCT bracis-28377 218 6 vol vol NOUN bracis-28377 218 7 . . PROPN bracis-28377 218 8 11672 11672 NUM bracis-28377 218 9 , , PUNCT bracis-28377 218 10 pp pp ADV bracis-28377 218 11 . . PUNCT bracis-28377 219 1 418–430 418–430 NUM bracis-28377 219 2 . . PUNCT bracis-28377 219 3 springer springer NOUN bracis-28377 219 4 , , PUNCT bracis-28377 219 5 cham cham PROPN bracis-28377 219 6 ( ( PUNCT bracis-28377 219 7 2019 2019 NUM bracis-28377 219 8 ) ) PUNCT bracis-28377 219 9 . . PUNCT bracis-28377 220 1 https://doi.org/10.1007/978-3-030-29894-4_34 https://doi.org/10.1007/978-3-030-29894-4_34 PROPN bracis-28377 220 2 chapter chapter NOUN bracis-28377 220 3     SPACE bracis-28377 220 4 google google PROPN bracis-28377 220 5 scholar scholar NOUN bracis-28377 220 6     SPACE bracis-28377 220 7 software software NOUN bracis-28377 220 8 , , PUNCT bracis-28377 220 9 a. a. NOUN bracis-28377 220 10 : : PUNCT bracis-28377 220 11 pet pet ADJ bracis-28377 220 12 insurance insurance NOUN bracis-28377 220 13 fraud fraud NOUN bracis-28377 220 14 increases increase NOUN bracis-28377 220 15 ( ( PUNCT bracis-28377 220 16 2018 2018 NUM bracis-28377 220 17 ) ) PUNCT bracis-28377 220 18 . . PUNCT bracis-28377 221 1 https://youtalk-insurance.com/broker-news/400-rise-in-pet-insurance-fraud-highlights-need-for-new-approach https://youtalk-insurance.com/broker-news/400-rise-in-pet-insurance-fraud-highlights-need-for-new-approach PROPN bracis-28377 221 2 . . PUNCT bracis-28377 222 1 accessed access VERB bracis-28377 222 2 18 18 NUM bracis-28377 222 3 june june PROPN bracis-28377 222 4 2022 2022 NUM bracis-28377 222 5 targ targ PROPN bracis-28377 222 6 , , PUNCT bracis-28377 222 7 s. s. PROPN bracis-28377 222 8 , , PUNCT bracis-28377 222 9 almeida almeida PROPN bracis-28377 222 10 , , PUNCT bracis-28377 222 11 d. d. PROPN bracis-28377 222 12 , , PUNCT bracis-28377 222 13 lyman lyman PROPN bracis-28377 222 14 , , PUNCT bracis-28377 222 15 k. k. PROPN bracis-28377 222 16 : : PUNCT bracis-28377 222 17 resnet resnet NOUN bracis-28377 222 18 in in ADP bracis-28377 222 19 resnet resnet NOUN bracis-28377 222 20 : : PUNCT bracis-28377 222 21 generalizing generalize VERB bracis-28377 222 22 residual residual ADJ bracis-28377 222 23 architectures architecture NOUN bracis-28377 222 24 ( ( PUNCT bracis-28377 222 25 2016 2016 NUM bracis-28377 222 26 ) ) PUNCT bracis-28377 222 27 . . PUNCT bracis-28377 223 1 arxiv arxiv PROPN bracis-28377 223 2 preprint preprint VERB bracis-28377 223 3 arxiv:1603.08029 arxiv:1603.08029 ADJ bracis-28377 223 4 vaswani vaswani NOUN bracis-28377 223 5 , , PUNCT bracis-28377 223 6 a. a. NOUN bracis-28377 223 7 , , PUNCT bracis-28377 223 8 et et PROPN bracis-28377 223 9 al al PROPN bracis-28377 223 10 . . PUNCT bracis-28377 223 11 : : PUNCT bracis-28377 224 1 attention attention NOUN bracis-28377 224 2 is be AUX bracis-28377 224 3 all all PRON bracis-28377 224 4 you you PRON bracis-28377 224 5 need need VERB bracis-28377 224 6 . . PUNCT bracis-28377 225 1 advances advance NOUN bracis-28377 225 2 in in ADP bracis-28377 225 3 neural neural ADJ bracis-28377 225 4 information information NOUN bracis-28377 225 5 processing processing NOUN bracis-28377 225 6 systems system NOUN bracis-28377 225 7 30 30 NUM bracis-28377 225 8 ( ( PUNCT bracis-28377 225 9 2017 2017 NUM bracis-28377 225 10 ) ) PUNCT bracis-28377 225 11 google google PROPN bracis-28377 225 12 scholar scholar NOUN bracis-28377 225 13     SPACE bracis-28377 225 14 yoon yoon PROPN bracis-28377 225 15 , , PUNCT bracis-28377 225 16 b. b. PROPN bracis-28377 225 17 , , PUNCT bracis-28377 225 18 so so ADV bracis-28377 225 19 , , PUNCT bracis-28377 225 20 h. h. PROPN bracis-28377 225 21 , , PUNCT bracis-28377 225 22 rhee rhee PROPN bracis-28377 225 23 , , PUNCT bracis-28377 225 24 j. j. PROPN bracis-28377 225 25 : : PUNCT bracis-28377 225 26 a a DET bracis-28377 225 27 methodology methodology NOUN bracis-28377 225 28 for for ADP bracis-28377 225 29 utilizing utilize VERB bracis-28377 225 30 vector vector NOUN bracis-28377 225 31 space space NOUN bracis-28377 225 32 to to PART bracis-28377 225 33 improve improve VERB bracis-28377 225 34 the the DET bracis-28377 225 35 performance performance NOUN bracis-28377 225 36 of of ADP bracis-28377 225 37 a a DET bracis-28377 225 38 dog dog NOUN bracis-28377 225 39 face face NOUN bracis-28377 225 40 identification identification NOUN bracis-28377 225 41 model model NOUN bracis-28377 225 42 . . PUNCT bracis-28377 226 1 appl appl PROPN bracis-28377 226 2 . . PUNCT bracis-28377 227 1 sci sci PROPN bracis-28377 227 2 . . PROPN bracis-28377 227 3 11(5 11(5 NUM bracis-28377 227 4 ) ) PUNCT bracis-28377 227 5 , , PUNCT bracis-28377 227 6 2074 2074 NUM bracis-28377 227 7 ( ( PUNCT bracis-28377 227 8 2021 2021 NUM bracis-28377 227 9 ) ) PUNCT bracis-28377 227 10 article article NOUN bracis-28377 227 11     SPACE bracis-28377 227 12 google google PROPN bracis-28377 227 13 scholar scholar PROPN bracis-28377 227 14     SPACE bracis-28377 227 15 zhang zhang PROPN bracis-28377 227 16 , , PUNCT bracis-28377 227 17 k. k. PROPN bracis-28377 227 18 , , PUNCT bracis-28377 227 19 sun sun PROPN bracis-28377 227 20 , , PUNCT bracis-28377 227 21 m. m. NOUN bracis-28377 227 22 , , PUNCT bracis-28377 227 23 han han PROPN bracis-28377 227 24 , , PUNCT bracis-28377 227 25 t.x t.x PROPN bracis-28377 227 26 . . PROPN bracis-28377 227 27 , , PUNCT bracis-28377 227 28 yuan yuan PROPN bracis-28377 227 29 , , PUNCT bracis-28377 227 30 x. x. PROPN bracis-28377 227 31 , , PUNCT bracis-28377 227 32 guo guo PROPN bracis-28377 227 33 , , PUNCT bracis-28377 227 34 l. l. PROPN bracis-28377 227 35 , , PUNCT bracis-28377 227 36 liu liu PROPN bracis-28377 227 37 , , PUNCT bracis-28377 227 38 t. t. PROPN bracis-28377 227 39 : : PUNCT bracis-28377 227 40 residual residual ADJ bracis-28377 227 41 networks network NOUN bracis-28377 227 42 of of ADP bracis-28377 227 43 residual residual ADJ bracis-28377 227 44 networks network NOUN bracis-28377 227 45 : : PUNCT bracis-28377 227 46 multilevel multilevel VERB bracis-28377 227 47 residual residual ADJ bracis-28377 227 48 networks network NOUN bracis-28377 227 49 . . PUNCT bracis-28377 228 1 ieee ieee PROPN bracis-28377 228 2 trans trans PROPN bracis-28377 228 3 . . PUNCT bracis-28377 229 1 circuits circuit NOUN bracis-28377 229 2 syst syst PROPN bracis-28377 229 3 . . PUNCT bracis-28377 229 4 video video NOUN bracis-28377 229 5 technol technol NOUN bracis-28377 229 6 . . PUNCT bracis-28377 229 7 28(6 28(6 NOUN bracis-28377 229 8 ) ) PUNCT bracis-28377 229 9 , , PUNCT bracis-28377 229 10 1303–1314 1303–1314 NUM bracis-28377 229 11 ( ( PUNCT bracis-28377 229 12 2017 2017 NUM bracis-28377 229 13 ) ) PUNCT bracis-28377 229 14 article article NOUN bracis-28377 229 15     SPACE bracis-28377 229 16 google google PROPN bracis-28377 229 17 scholar scholar NOUN bracis-28377 229 18     SPACE bracis-28377 229 19 zhang zhang PROPN bracis-28377 229 20 , , PUNCT bracis-28377 229 21 x. x. PROPN bracis-28377 229 22 , , PUNCT bracis-28377 229 23 zhao zhao PROPN bracis-28377 229 24 , , PUNCT bracis-28377 229 25 r. r. PROPN bracis-28377 229 26 , , PUNCT bracis-28377 229 27 qiao qiao PROPN bracis-28377 229 28 , , PUNCT bracis-28377 229 29 y. y. PROPN bracis-28377 229 30 , , PUNCT bracis-28377 229 31 wang wang PROPN bracis-28377 229 32 , , PUNCT bracis-28377 229 33 x. x. PROPN bracis-28377 229 34 , , PUNCT bracis-28377 229 35 li li PROPN bracis-28377 229 36 , , PUNCT bracis-28377 229 37 h. h. PROPN bracis-28377 229 38 : : PUNCT bracis-28377 229 39 adacos adaco NOUN bracis-28377 229 40 : : PUNCT bracis-28377 229 41 adaptively adaptively ADV bracis-28377 229 42 scaling scale VERB bracis-28377 229 43 cosine cosine NOUN bracis-28377 229 44 logits logit NOUN bracis-28377 229 45 for for ADP bracis-28377 229 46 effectively effectively ADV bracis-28377 229 47 learning learn VERB bracis-28377 229 48 deep deep ADJ bracis-28377 229 49 face face NOUN bracis-28377 229 50 representations representation NOUN bracis-28377 229 51 . . PUNCT bracis-28377 230 1 in in ADP bracis-28377 230 2 : : PUNCT bracis-28377 230 3 proceedings proceeding NOUN bracis-28377 230 4 of of ADP bracis-28377 230 5 the the DET bracis-28377 230 6 ieee ieee NOUN bracis-28377 230 7 / / SYM bracis-28377 230 8 cvf cvf NOUN bracis-28377 230 9 conference conference NOUN bracis-28377 230 10 on on ADP bracis-28377 230 11 computer computer NOUN bracis-28377 230 12 vision vision NOUN bracis-28377 230 13 and and CCONJ bracis-28377 230 14 pattern pattern NOUN bracis-28377 230 15 recognition recognition NOUN bracis-28377 230 16 , , PUNCT bracis-28377 230 17 pp pp PROPN bracis-28377 230 18 . . PUNCT bracis-28377 230 19 10823–10832 10823–10832 NUM bracis-28377 230 20 ( ( PUNCT bracis-28377 230 21 2019 2019 NUM bracis-28377 230 22 ) ) PUNCT bracis-28377 230 23 google google NOUN bracis-28377 230 24 scholar scholar NOUN bracis-28377 230 25     SPACE bracis-28377 230 26 download download NOUN bracis-28377 230 27 references reference NOUN bracis-28377 230 28 author author NOUN bracis-28377 230 29 information information NOUN bracis-28377 230 30 authors author NOUN bracis-28377 230 31 and and CCONJ bracis-28377 230 32 affiliations affiliation NOUN bracis-28377 230 33 sao sao PROPN bracis-28377 230 34 paulo paulo PROPN bracis-28377 230 35 state state PROPN bracis-28377 230 36 university university PROPN bracis-28377 230 37 ( ( PUNCT bracis-28377 230 38 unesp unesp NOUN bracis-28377 230 39 ) ) PUNCT bracis-28377 230 40 , , PUNCT bracis-28377 230 41 bauru bauru PROPN bracis-28377 230 42 , , PUNCT bracis-28377 230 43 sao sao PROPN bracis-28377 230 44 paulo paulo PROPN bracis-28377 230 45 , , PUNCT bracis-28377 230 46 brazil brazil PROPN bracis-28377 230 47 victor victor PROPN bracis-28377 230 48 hugo hugo PROPN bracis-28377 230 49 braguim braguim PROPN bracis-28377 230 50 canto canto PROPN bracis-28377 230 51 , , PUNCT bracis-28377 230 52     SPACE bracis-28377 230 53 joão joão PROPN bracis-28377 230 54 renato renato PROPN bracis-28377 230 55 ribeiro ribeiro PROPN bracis-28377 230 56 manesco manesco PROPN bracis-28377 230 57     SPACE bracis-28377 230 58 & & CCONJ bracis-28377 230 59     SPACE bracis-28377 230 60 aparecido aparecido PROPN bracis-28377 230 61 nilceu nilceu PROPN bracis-28377 230 62 marana marana PROPN bracis-28377 230 63 bank bank PROPN bracis-28377 230 64 of of ADP bracis-28377 230 65 brazil brazil PROPN bracis-28377 230 66 brasilia brasilia PROPN bracis-28377 230 67 , , PUNCT bracis-28377 230 68 federal federal ADJ bracis-28377 230 69 district district PROPN bracis-28377 230 70 , , PUNCT bracis-28377 230 71 brazil brazil PROPN bracis-28377 230 72 gustavo gustavo PROPN bracis-28377 230 73 botelho botelho PROPN bracis-28377 230 74 de de PROPN bracis-28377 230 75 souza souza PROPN bracis-28377 230 76 authors authors PROPN bracis-28377 230 77 victor victor PROPN bracis-28377 230 78 hugo hugo PROPN bracis-28377 230 79 braguim braguim PROPN bracis-28377 230 80 cantoview cantoview NOUN bracis-28377 230 81 author author NOUN bracis-28377 230 82 publications publication NOUN bracis-28377 230 83 search search NOUN bracis-28377 230 84 author author NOUN bracis-28377 230 85 on on ADP bracis-28377 230 86 : : PUNCT bracis-28377 230 87 pubmed pubmed PROPN bracis-28377 230 88     SPACE bracis-28377 230 89 google google PROPN bracis-28377 230 90 scholar scholar PROPN bracis-28377 230 91 joão joão PROPN bracis-28377 230 92 renato renato PROPN bracis-28377 230 93 ribeiro ribeiro PROPN bracis-28377 230 94 manescoview manescoview PROPN bracis-28377 230 95 author author NOUN bracis-28377 230 96 publications publication NOUN bracis-28377 230 97 search search NOUN bracis-28377 230 98 author author NOUN bracis-28377 230 99 on on ADP bracis-28377 230 100 : : PUNCT bracis-28377 230 101 pubmed pubmed PROPN bracis-28377 230 102     SPACE bracis-28377 230 103 google google PROPN bracis-28377 230 104 scholar scholar NOUN bracis-28377 230 105 gustavo gustavo PROPN bracis-28377 230 106 botelho botelho PROPN bracis-28377 230 107 de de ADP bracis-28377 230 108 souzaview souzaview NOUN bracis-28377 230 109 author author NOUN bracis-28377 230 110 publications publication NOUN bracis-28377 230 111 search search NOUN bracis-28377 230 112 author author NOUN bracis-28377 230 113 on on ADP bracis-28377 230 114 : : PUNCT bracis-28377 230 115 pubmed pubmed PROPN bracis-28377 230 116     SPACE bracis-28377 230 117 google google PROPN bracis-28377 230 118 scholar scholar PROPN bracis-28377 230 119 aparecido aparecido PROPN bracis-28377 230 120 nilceu nilceu PROPN bracis-28377 230 121 maranaview maranaview PROPN bracis-28377 230 122 author author NOUN bracis-28377 230 123 publications publication VERB bracis-28377 230 124 search search NOUN bracis-28377 230 125 author author NOUN bracis-28377 230 126 on on ADP bracis-28377 230 127 : : PUNCT bracis-28377 230 128 pubmed pubmed PROPN bracis-28377 230 129     SPACE bracis-28377 230 130 google google PROPN bracis-28377 230 131 scholar scholar NOUN bracis-28377 230 132 corresponding correspond VERB bracis-28377 230 133 author author NOUN bracis-28377 230 134 correspondence correspondence NOUN bracis-28377 230 135 to to ADP bracis-28377 230 136 victor victor PROPN bracis-28377 230 137 hugo hugo PROPN bracis-28377 230 138 braguim braguim PROPN bracis-28377 230 139 canto canto PROPN bracis-28377 230 140 . . PUNCT bracis-28377 231 1 editor editor NOUN bracis-28377 231 2 information information NOUN bracis-28377 231 3 editors editor NOUN bracis-28377 231 4 and and CCONJ bracis-28377 231 5 affiliations affiliation NOUN bracis-28377 231 6 federal federal PROPN bracis-28377 231 7 university university PROPN bracis-28377 231 8 of of ADP bracis-28377 231 9 são são PROPN bracis-28377 231 10 carlos carlos PROPN bracis-28377 231 11 , , PUNCT bracis-28377 231 12 são são PROPN bracis-28377 231 13 carlos carlos PROPN bracis-28377 231 14 , , PUNCT bracis-28377 231 15 brazil brazil PROPN bracis-28377 231 16 murilo murilo PROPN bracis-28377 231 17 c. c. PROPN bracis-28377 231 18 naldi naldi PROPN bracis-28377 231 19 centro centro PROPN bracis-28377 231 20 universitario universitario PROPN bracis-28377 231 21 da da PROPN bracis-28377 231 22 fei fei PROPN bracis-28377 231 23 , , PUNCT bracis-28377 231 24 são são PROPN bracis-28377 231 25 bernardo bernardo PROPN bracis-28377 231 26 do do AUX bracis-28377 231 27 campo campo PROPN bracis-28377 231 28 , , PUNCT bracis-28377 231 29 brazil brazil PROPN bracis-28377 231 30 reinaldo reinaldo PROPN bracis-28377 231 31 a. a. PROPN bracis-28377 231 32 c. c. PROPN bracis-28377 231 33 bianchi bianchi PROPN bracis-28377 231 34 rights right NOUN bracis-28377 231 35 and and CCONJ bracis-28377 231 36 permissions permission NOUN bracis-28377 231 37 reprints reprint NOUN bracis-28377 231 38 and and CCONJ bracis-28377 231 39 permissions permission VERB bracis-28377 231 40 copyright copyright NOUN bracis-28377 231 41 information information NOUN bracis-28377 231 42 © © ADP bracis-28377 231 43 2023 2023 NUM bracis-28377 231 44 the the DET bracis-28377 231 45 author(s author(s NOUN bracis-28377 231 46 ) ) PUNCT bracis-28377 231 47 , , PUNCT bracis-28377 231 48 under under ADP bracis-28377 231 49 exclusive exclusive ADJ bracis-28377 231 50 license license NOUN bracis-28377 231 51 to to ADP bracis-28377 231 52 springer springer NOUN bracis-28377 231 53 nature nature PROPN bracis-28377 231 54 switzerland switzerland PROPN bracis-28377 231 55 ag ag PROPN bracis-28377 231 56 about about ADP bracis-28377 231 57 this this DET bracis-28377 231 58 paper paper NOUN bracis-28377 231 59 cite cite VERB bracis-28377 231 60 this this DET bracis-28377 231 61 paper paper NOUN bracis-28377 231 62 canto canto NOUN bracis-28377 231 63 , , PUNCT bracis-28377 231 64 v.h.b v.h.b ADJ bracis-28377 231 65 . . PROPN bracis-28377 231 66 , , PUNCT bracis-28377 231 67 manesco manesco PROPN bracis-28377 231 68 , , PUNCT bracis-28377 231 69 j.r.r j.r.r PROPN bracis-28377 231 70 . . PROPN bracis-28377 231 71 , , PUNCT bracis-28377 231 72 de de PROPN bracis-28377 231 73 souza souza PROPN bracis-28377 231 74 , , PUNCT bracis-28377 231 75 g.b g.b PROPN bracis-28377 231 76 . . PROPN bracis-28377 231 77 , , PUNCT bracis-28377 231 78 marana marana PROPN bracis-28377 231 79 , , PUNCT bracis-28377 231 80 a.n a.n PROPN bracis-28377 231 81 . . PROPN bracis-28377 231 82 ( ( PUNCT bracis-28377 231 83 2023 2023 NUM bracis-28377 231 84 ) ) PUNCT bracis-28377 231 85 . . PUNCT bracis-28377 232 1 dog dog NOUN bracis-28377 232 2 face face NOUN bracis-28377 232 3 recognition recognition NOUN bracis-28377 232 4 using use VERB bracis-28377 232 5 vision vision NOUN bracis-28377 232 6 transformer transformer NOUN bracis-28377 232 7 . . PUNCT bracis-28377 233 1 in in ADP bracis-28377 233 2 : : PUNCT bracis-28377 233 3 naldi naldi PROPN bracis-28377 233 4 , , PUNCT bracis-28377 233 5 m.c m.c PROPN bracis-28377 233 6 . . PROPN bracis-28377 233 7 , , PUNCT bracis-28377 233 8 bianchi bianchi PROPN bracis-28377 233 9 , , PUNCT bracis-28377 233 10 r.a.c r.a.c ADP bracis-28377 233 11 . . PUNCT bracis-28377 233 12 ( ( PUNCT bracis-28377 233 13 eds ed NOUN bracis-28377 233 14 ) ) PUNCT bracis-28377 233 15 intelligent intelligent ADJ bracis-28377 233 16 systems system NOUN bracis-28377 233 17 . . PUNCT bracis-28377 234 1 bracis bracis PROPN bracis-28377 234 2 2023 2023 NUM bracis-28377 234 3 . . PUNCT bracis-28377 235 1 lecture lecture NOUN bracis-28377 235 2 notes note NOUN bracis-28377 235 3 in in ADP bracis-28377 235 4 computer computer NOUN bracis-28377 235 5 science science NOUN bracis-28377 235 6 ( ( PUNCT bracis-28377 235 7 ) ) PUNCT bracis-28377 235 8 , , PUNCT bracis-28377 235 9 vol vol NOUN bracis-28377 235 10 14196 14196 NUM bracis-28377 235 11 . . PUNCT bracis-28377 236 1 springer springer NOUN bracis-28377 236 2 , , PUNCT bracis-28377 236 3 cham cham PROPN bracis-28377 236 4 . . PUNCT bracis-28377 237 1 https://doi.org/10.1007/978-3-031-45389-2_3 https://doi.org/10.1007/978-3-031-45389-2_3 PROPN bracis-28377 237 2 download download PROPN bracis-28377 237 3 citation citation NOUN bracis-28377 237 4 .ris .ris PUNCT bracis-28377 238 1 .enw .enw PROPN bracis-28377 238 2 .bib .bib PUNCT bracis-28377 239 1 doi doi PROPN bracis-28377 239 2 : : PUNCT bracis-28377 239 3 https://doi.org/10.1007/978-3-031-45389-2_3 https://doi.org/10.1007/978-3-031-45389-2_3 X bracis-28377 239 4 published publish VERB bracis-28377 239 5 : : PUNCT bracis-28377 239 6 12 12 NUM bracis-28377 239 7 october october PROPN bracis-28377 239 8 2023 2023 NUM bracis-28377 239 9 publisher publisher NOUN bracis-28377 239 10 name name NOUN bracis-28377 239 11 : : PUNCT bracis-28377 239 12 springer springer NOUN bracis-28377 239 13 , , PUNCT bracis-28377 239 14 cham cham PROPN bracis-28377 239 15 print print PROPN bracis-28377 239 16 isbn isbn PROPN bracis-28377 239 17 : : PUNCT bracis-28377 239 18 978 978 NUM bracis-28377 239 19 - - SYM bracis-28377 239 20 3 3 NUM bracis-28377 239 21 - - PUNCT bracis-28377 239 22 031 031 NUM bracis-28377 239 23 - - PUNCT bracis-28377 239 24 45388 45388 NUM bracis-28377 239 25 - - SYM bracis-28377 239 26 5 5 NUM bracis-28377 239 27 online online ADJ bracis-28377 239 28 isbn isbn NOUN bracis-28377 239 29 : : PUNCT bracis-28377 239 30 978 978 NUM bracis-28377 239 31 - - SYM bracis-28377 239 32 3 3 NUM bracis-28377 239 33 - - PUNCT bracis-28377 239 34 031 031 NUM bracis-28377 239 35 - - PUNCT bracis-28377 239 36 45389 45389 NUM bracis-28377 239 37 - - SYM bracis-28377 239 38 2 2 NUM bracis-28377 239 39 ebook ebook NOUN bracis-28377 239 40 packages package NOUN bracis-28377 239 41 : : PUNCT bracis-28377 239 42 computer computer NOUN bracis-28377 239 43 sciencecomputer sciencecomputer NOUN bracis-28377 239 44 science science NOUN bracis-28377 239 45 ( ( PUNCT bracis-28377 239 46 r0 r0 NOUN bracis-28377 239 47 ) ) PUNCT bracis-28377 239 48 share share VERB bracis-28377 239 49 this this DET bracis-28377 239 50 paper paper NOUN bracis-28377 239 51 anyone anyone PRON bracis-28377 239 52 you you PRON bracis-28377 239 53 share share VERB bracis-28377 239 54 the the DET bracis-28377 239 55 following follow VERB bracis-28377 239 56 link link NOUN bracis-28377 239 57 with with ADP bracis-28377 239 58 will will AUX bracis-28377 239 59 be be AUX bracis-28377 239 60 able able ADJ bracis-28377 239 61 to to PART bracis-28377 239 62 read read VERB bracis-28377 239 63 this this DET bracis-28377 239 64 content content NOUN bracis-28377 239 65 : : PUNCT bracis-28377 239 66 get get VERB bracis-28377 239 67 shareable shareable ADJ bracis-28377 239 68 linksorry linksorry NOUN bracis-28377 239 69 , , PUNCT bracis-28377 239 70 a a DET bracis-28377 239 71 shareable shareable ADJ bracis-28377 239 72 link link NOUN bracis-28377 239 73 is be AUX bracis-28377 239 74 not not PART bracis-28377 239 75 currently currently ADV bracis-28377 239 76 available available ADJ bracis-28377 239 77 for for ADP bracis-28377 239 78 this this DET bracis-28377 239 79 article article NOUN bracis-28377 239 80 . . PUNCT bracis-28377 240 1 copy copy VERB bracis-28377 240 2 shareable shareable ADJ bracis-28377 240 3 link link NOUN bracis-28377 240 4 to to PART bracis-28377 240 5 clipboard clipboard NOUN bracis-28377 240 6 provided provide VERB bracis-28377 240 7 by by ADP bracis-28377 240 8 the the DET bracis-28377 240 9 springer springer NOUN bracis-28377 240 10 nature nature PROPN bracis-28377 240 11 sharedit sharedit PROPN bracis-28377 240 12 content content NOUN bracis-28377 240 13 - - PUNCT bracis-28377 240 14 sharing share VERB bracis-28377 240 15 initiative initiative NOUN bracis-28377 240 16 keywords keyword NOUN bracis-28377 240 17 animal animal NOUN bracis-28377 240 18 biometrics biometric NOUN bracis-28377 240 19 dog dog NOUN bracis-28377 240 20 identification identification NOUN bracis-28377 240 21 facial facial ADJ bracis-28377 240 22 recognition recognition NOUN bracis-28377 240 23 convolutional convolutional ADJ bracis-28377 240 24 neural neural ADJ bracis-28377 240 25 network network NOUN bracis-28377 240 26 visual visual ADJ bracis-28377 240 27 transformer transformer NOUN bracis-28377 240 28 arcface arcface NOUN bracis-28377 240 29 publish publish VERB bracis-28377 240 30 with with ADP bracis-28377 240 31 us us PROPN bracis-28377 240 32 policies policy NOUN bracis-28377 240 33 and and CCONJ bracis-28377 240 34 ethics ethic NOUN bracis-28377 240 35 profiles profile NOUN bracis-28377 240 36 joão joão PROPN bracis-28377 240 37 renato renato PROPN bracis-28377 240 38 ribeiro ribeiro PROPN bracis-28377 240 39 manesco manesco PROPN bracis-28377 240 40 view view PROPN bracis-28377 240 41 author author NOUN bracis-28377 240 42 profile profile NOUN bracis-28377 240 43 search search NOUN bracis-28377 240 44 search search NOUN bracis-28377 240 45 by by ADP bracis-28377 240 46 keyword keyword NOUN bracis-28377 240 47 or or CCONJ bracis-28377 240 48 author author NOUN bracis-28377 240 49 search search NOUN bracis-28377 240 50 navigation navigation NOUN bracis-28377 240 51 find find VERB bracis-28377 240 52 a a DET bracis-28377 240 53 journal journal NOUN bracis-28377 240 54 publish publish VERB bracis-28377 240 55 with with ADP bracis-28377 240 56 us we PRON bracis-28377 240 57 track track VERB bracis-28377 240 58 your your PRON bracis-28377 240 59 research research NOUN bracis-28377 240 60 discover discover VERB bracis-28377 240 61 content content NOUN bracis-28377 240 62 journals journal NOUN bracis-28377 240 63 a a DET bracis-28377 240 64 - - PUNCT bracis-28377 240 65 z z NOUN bracis-28377 240 66 books book NOUN bracis-28377 240 67 a a DET bracis-28377 240 68 - - PUNCT bracis-28377 240 69 z z NOUN bracis-28377 240 70 publish publish NOUN bracis-28377 240 71 with with ADP bracis-28377 240 72 us us PROPN bracis-28377 240 73 journal journal PROPN bracis-28377 240 74 finder finder PROPN bracis-28377 240 75 publish publish VERB bracis-28377 240 76 your your PRON bracis-28377 240 77 research research NOUN bracis-28377 240 78 language language NOUN bracis-28377 240 79 editing edit VERB bracis-28377 240 80 open open ADJ bracis-28377 240 81 access access NOUN bracis-28377 240 82 publishing publishing NOUN bracis-28377 240 83 products product NOUN bracis-28377 240 84 and and CCONJ bracis-28377 240 85 services service NOUN bracis-28377 240 86 our our PRON bracis-28377 240 87 products product NOUN bracis-28377 240 88 librarians librarian VERB bracis-28377 240 89 societies society NOUN bracis-28377 240 90 partners partner NOUN bracis-28377 240 91 and and CCONJ bracis-28377 240 92 advertisers advertiser NOUN bracis-28377 240 93 our our PRON bracis-28377 240 94 brands brand NOUN bracis-28377 240 95 springer springer NOUN bracis-28377 240 96 nature nature PROPN bracis-28377 240 97 portfolio portfolio PROPN bracis-28377 240 98 bmc bmc PROPN bracis-28377 240 99 palgrave palgrave PROPN bracis-28377 240 100 macmillan macmillan PROPN bracis-28377 240 101 apress apress PROPN bracis-28377 240 102 discover discover VERB bracis-28377 240 103 your your PRON bracis-28377 240 104 privacy privacy NOUN bracis-28377 240 105 choices choice NOUN bracis-28377 240 106 / / SYM bracis-28377 240 107 manage manage NOUN bracis-28377 240 108 cookies cookie NOUN bracis-28377 240 109 your your PRON bracis-28377 240 110 us us PROPN bracis-28377 241 1 state state NOUN bracis-28377 241 2 privacy privacy NOUN bracis-28377 241 3 rights right NOUN bracis-28377 241 4 accessibility accessibility NOUN bracis-28377 241 5 statement statement NOUN bracis-28377 241 6 terms term NOUN bracis-28377 241 7 and and CCONJ bracis-28377 241 8 conditions condition NOUN bracis-28377 241 9 privacy privacy NOUN bracis-28377 241 10 policy policy NOUN bracis-28377 241 11 help help NOUN bracis-28377 241 12 and and CCONJ bracis-28377 241 13 support support VERB bracis-28377 241 14 legal legal ADJ bracis-28377 241 15 notice notice NOUN bracis-28377 241 16 cancel cancel VERB bracis-28377 241 17 contracts contract NOUN bracis-28377 241 18 here here ADV bracis-28377 241 19 not not PART bracis-28377 241 20 affiliated affiliate VERB bracis-28377 241 21 © © PROPN bracis-28377 241 22 2025 2025 NUM bracis-28377 241 23 springer springer NOUN bracis-28377 241 24 nature nature NOUN