id sid tid token lemma pos ajst-12714 1 1 academic academic ADJ ajst-12714 1 2 journal journal NOUN ajst-12714 1 3 of of ADP ajst-12714 1 4 science science NOUN ajst-12714 1 5 and and CCONJ ajst-12714 1 6 technology technology NOUN ajst-12714 1 7 issn issn NOUN ajst-12714 1 8 : : PUNCT ajst-12714 1 9 2771 2771 NUM ajst-12714 1 10 - - SYM ajst-12714 1 11 3032 3032 NUM ajst-12714 1 12 | | NOUN ajst-12714 1 13 vol vol NOUN ajst-12714 1 14 . . PROPN ajst-12714 2 1 7 7 NUM ajst-12714 2 2 , , PUNCT ajst-12714 2 3 no no INTJ ajst-12714 2 4 . . NOUN ajst-12714 2 5 3 3 NUM ajst-12714 2 6 , , PUNCT ajst-12714 2 7 2023 2023 NUM ajst-12714 2 8 34 34 NUM ajst-12714 2 9 research research NOUN ajst-12714 2 10 on on ADP ajst-12714 2 11 geometric geometric ADJ ajst-12714 2 12 parameter parameter NOUN ajst-12714 2 13 prediction prediction NOUN ajst-12714 2 14 algorithm algorithm NOUN ajst-12714 2 15 for for ADP ajst-12714 2 16 oil oil NOUN ajst-12714 2 17 and and CCONJ ajst-12714 2 18 gas gas NOUN ajst-12714 2 19 pipeline pipeline NOUN ajst-12714 2 20 defects defect NOUN ajst-12714 2 21 yijun yijun PROPN ajst-12714 2 22 liu liu PROPN ajst-12714 2 23 , , PUNCT ajst-12714 2 24 wei wei PROPN ajst-12714 2 25 wu wu PROPN ajst-12714 2 26 , , PUNCT ajst-12714 2 27 xiaolin xiaolin PROPN ajst-12714 2 28 ren ren PROPN ajst-12714 2 29 , , PUNCT ajst-12714 2 30 le le X ajst-12714 2 31 qin qin PROPN ajst-12714 2 32 , , PUNCT ajst-12714 2 33 yukun yukun PROPN ajst-12714 2 34 wang wang PROPN ajst-12714 2 35 school school PROPN ajst-12714 2 36 of of ADP ajst-12714 2 37 mechanical mechanical ADJ ajst-12714 2 38 engineering engineering NOUN ajst-12714 2 39 , , PUNCT ajst-12714 2 40 xi'an xi'an PROPN ajst-12714 2 41 shiyou shiyou PROPN ajst-12714 2 42 university university PROPN ajst-12714 2 43 , , PUNCT ajst-12714 2 44 shaanxi shaanxi PROPN ajst-12714 2 45 xi’an xi’an PROPN ajst-12714 2 46 710065 710065 NUM ajst-12714 2 47 , , PUNCT ajst-12714 2 48 china china PROPN ajst-12714 2 49 abstract abstract NOUN ajst-12714 2 50 : : PUNCT ajst-12714 2 51 aiming aim VERB ajst-12714 2 52 at at ADP ajst-12714 2 53 the the DET ajst-12714 2 54 problem problem NOUN ajst-12714 2 55 of of ADP ajst-12714 2 56 insufficient insufficient ADJ ajst-12714 2 57 feature feature NOUN ajst-12714 2 58 extraction extraction NOUN ajst-12714 2 59 of of ADP ajst-12714 2 60 magnetic magnetic ADJ ajst-12714 2 61 leakage leakage NOUN ajst-12714 2 62 signals signal NOUN ajst-12714 2 63 by by ADP ajst-12714 2 64 traditional traditional ADJ ajst-12714 2 65 neural neural ADJ ajst-12714 2 66 networks network NOUN ajst-12714 2 67 , , PUNCT ajst-12714 2 68 this this DET ajst-12714 2 69 paper paper NOUN ajst-12714 2 70 proposes propose VERB ajst-12714 2 71 an an DET ajst-12714 2 72 attention attention NOUN ajst-12714 2 73 depth depth NOUN ajst-12714 2 74 convolutional convolutional ADJ ajst-12714 2 75 neural neural ADJ ajst-12714 2 76 network network NOUN ajst-12714 2 77 model model NOUN ajst-12714 2 78 ( ( PUNCT ajst-12714 2 79 eca eca NOUN ajst-12714 2 80 - - PUNCT ajst-12714 2 81 vgg16 vgg16 NOUN ajst-12714 2 82 ) ) PUNCT ajst-12714 2 83 , , PUNCT ajst-12714 2 84 which which PRON ajst-12714 2 85 adds add VERB ajst-12714 2 86 an an DET ajst-12714 2 87 attention attention NOUN ajst-12714 2 88 mechanism mechanism NOUN ajst-12714 2 89 combined combine VERB ajst-12714 2 90 with with ADP ajst-12714 2 91 convolutional convolutional ADJ ajst-12714 2 92 neural neural ADJ ajst-12714 2 93 network network NOUN ajst-12714 2 94 on on ADP ajst-12714 2 95 the the DET ajst-12714 2 96 basis basis NOUN ajst-12714 2 97 of of ADP ajst-12714 2 98 deep deep ADJ ajst-12714 2 99 convolutional convolutional ADJ ajst-12714 2 100 neural neural ADJ ajst-12714 2 101 network network NOUN ajst-12714 2 102 vgg16 vgg16 NOUN ajst-12714 2 103 , , PUNCT ajst-12714 2 104 so so SCONJ ajst-12714 2 105 that that SCONJ ajst-12714 2 106 the the DET ajst-12714 2 107 neural neural ADJ ajst-12714 2 108 network network NOUN ajst-12714 2 109 can can AUX ajst-12714 2 110 focus focus VERB ajst-12714 2 111 on on ADP ajst-12714 2 112 the the DET ajst-12714 2 113 key key ADJ ajst-12714 2 114 information information NOUN ajst-12714 2 115 of of ADP ajst-12714 2 116 the the DET ajst-12714 2 117 input input NOUN ajst-12714 2 118 data datum NOUN ajst-12714 2 119 , , PUNCT ajst-12714 2 120 further far ADV ajst-12714 2 121 improve improve VERB ajst-12714 2 122 the the DET ajst-12714 2 123 ability ability NOUN ajst-12714 2 124 of of ADP ajst-12714 2 125 the the DET ajst-12714 2 126 grid grid NOUN ajst-12714 2 127 to to PART ajst-12714 2 128 extract extract VERB ajst-12714 2 129 image image NOUN ajst-12714 2 130 data datum NOUN ajst-12714 2 131 features feature NOUN ajst-12714 2 132 , , PUNCT ajst-12714 2 133 and and CCONJ ajst-12714 2 134 realize realize VERB ajst-12714 2 135 the the DET ajst-12714 2 136 accurate accurate ADJ ajst-12714 2 137 expression expression NOUN ajst-12714 2 138 of of ADP ajst-12714 2 139 defect defect NOUN ajst-12714 2 140 features feature NOUN ajst-12714 2 141 . . PUNCT ajst-12714 3 1 1 1 X ajst-12714 3 2 . . X ajst-12714 3 3 introduction introduction NOUN ajst-12714 3 4 oil oil NOUN ajst-12714 3 5 and and CCONJ ajst-12714 3 6 gas gas NOUN ajst-12714 3 7 is be AUX ajst-12714 3 8 a a DET ajst-12714 3 9 very very ADV ajst-12714 3 10 important important ADJ ajst-12714 3 11 national national ADJ ajst-12714 3 12 energy energy NOUN ajst-12714 3 13 , , PUNCT ajst-12714 3 14 pipelines pipeline NOUN ajst-12714 3 15 in in ADP ajst-12714 3 16 oil oil NOUN ajst-12714 3 17 and and CCONJ ajst-12714 3 18 gas gas NOUN ajst-12714 3 19 transportation transportation NOUN ajst-12714 3 20 has have VERB ajst-12714 3 21 the the DET ajst-12714 3 22 advantages advantage NOUN ajst-12714 3 23 of of ADP ajst-12714 3 24 long long ADJ ajst-12714 3 25 transportation transportation NOUN ajst-12714 3 26 distance distance NOUN ajst-12714 3 27 , , PUNCT ajst-12714 3 28 high high ADJ ajst-12714 3 29 efficiency efficiency NOUN ajst-12714 3 30 , , PUNCT ajst-12714 3 31 low low ADJ ajst-12714 3 32 cost cost NOUN ajst-12714 3 33 and and CCONJ ajst-12714 3 34 safety safety NOUN ajst-12714 3 35 and and CCONJ ajst-12714 3 36 reliability reliability NOUN ajst-12714 3 37 , , PUNCT ajst-12714 3 38 and and CCONJ ajst-12714 3 39 transportation transportation NOUN ajst-12714 3 40 pipelines pipeline NOUN ajst-12714 3 41 are be AUX ajst-12714 3 42 often often ADV ajst-12714 3 43 underground underground ADJ ajst-12714 3 44 , , PUNCT ajst-12714 3 45 after after ADP ajst-12714 3 46 a a DET ajst-12714 3 47 period period NOUN ajst-12714 3 48 of of ADP ajst-12714 3 49 use use NOUN ajst-12714 3 50 , , PUNCT ajst-12714 3 51 the the DET ajst-12714 3 52 pipeline pipeline NOUN ajst-12714 3 53 will will AUX ajst-12714 3 54 appear appear VERB ajst-12714 3 55 deformation deformation NOUN ajst-12714 3 56 and and CCONJ ajst-12714 3 57 defects defect NOUN ajst-12714 3 58 , , PUNCT ajst-12714 3 59 easy easy ADJ ajst-12714 3 60 to to PART ajst-12714 3 61 cause cause VERB ajst-12714 3 62 leakage leakage NOUN ajst-12714 3 63 accidents[1 accidents[1 PROPN ajst-12714 3 64 ] ] PUNCT ajst-12714 3 65 . . PUNCT ajst-12714 4 1 in in ADP ajst-12714 4 2 response response NOUN ajst-12714 4 3 to to ADP ajst-12714 4 4 these these DET ajst-12714 4 5 problems problem NOUN ajst-12714 4 6 , , PUNCT ajst-12714 4 7 the the DET ajst-12714 4 8 pipe pipe NOUN ajst-12714 4 9 must must AUX ajst-12714 4 10 be be AUX ajst-12714 4 11 inspected inspect VERB ajst-12714 4 12 and and CCONJ ajst-12714 4 13 the the DET ajst-12714 4 14 specific specific ADJ ajst-12714 4 15 size size NOUN ajst-12714 4 16 of of ADP ajst-12714 4 17 the the DET ajst-12714 4 18 defect defect NOUN ajst-12714 4 19 must must AUX ajst-12714 4 20 be be AUX ajst-12714 4 21 predicted predict VERB ajst-12714 4 22 . . PUNCT ajst-12714 5 1 at at ADP ajst-12714 5 2 present present ADJ ajst-12714 5 3 , , PUNCT ajst-12714 5 4 the the DET ajst-12714 5 5 detection detection NOUN ajst-12714 5 6 of of ADP ajst-12714 5 7 oil oil NOUN ajst-12714 5 8 and and CCONJ ajst-12714 5 9 gas gas NOUN ajst-12714 5 10 pipeline pipeline NOUN ajst-12714 5 11 defects defect NOUN ajst-12714 5 12 mainly mainly ADV ajst-12714 5 13 adopts adopt VERB ajst-12714 5 14 magnetic magnetic ADJ ajst-12714 5 15 leakage leakage NOUN ajst-12714 5 16 detection detection NOUN ajst-12714 5 17 technology technology NOUN ajst-12714 5 18 . . PUNCT ajst-12714 6 1 the the DET ajst-12714 6 2 identification identification NOUN ajst-12714 6 3 of of ADP ajst-12714 6 4 pipeline pipeline NOUN ajst-12714 6 5 defect defect NOUN ajst-12714 6 6 size size NOUN ajst-12714 6 7 mainly mainly ADV ajst-12714 6 8 relies rely VERB ajst-12714 6 9 on on ADP ajst-12714 6 10 manual manual ADJ ajst-12714 6 11 interpretation interpretation NOUN ajst-12714 6 12 , , PUNCT ajst-12714 6 13 because because SCONJ ajst-12714 6 14 the the DET ajst-12714 6 15 staff staff NOUN ajst-12714 6 16 diagnoses diagnose VERB ajst-12714 6 17 the the DET ajst-12714 6 18 defect defect NOUN ajst-12714 6 19 by by ADP ajst-12714 6 20 observing observe VERB ajst-12714 6 21 the the DET ajst-12714 6 22 collected collect VERB ajst-12714 6 23 pipeline pipeline NOUN ajst-12714 6 24 leakage leakage NOUN ajst-12714 6 25 signal signal NOUN ajst-12714 6 26 . . PUNCT ajst-12714 7 1 the the DET ajst-12714 7 2 method method NOUN ajst-12714 7 3 of of ADP ajst-12714 7 4 manual manual ADJ ajst-12714 7 5 interpretation interpretation NOUN ajst-12714 7 6 has have VERB ajst-12714 7 7 certain certain ADJ ajst-12714 7 8 limitations limitation NOUN ajst-12714 7 9 , , PUNCT ajst-12714 7 10 and and CCONJ ajst-12714 7 11 the the DET ajst-12714 7 12 subjective subjective ADJ ajst-12714 7 13 consciousness consciousness NOUN ajst-12714 7 14 is be AUX ajst-12714 7 15 too too ADV ajst-12714 7 16 strong[2 strong[2 PRON ajst-12714 7 17 ] ] PUNCT ajst-12714 7 18 . . PUNCT ajst-12714 8 1 with with ADP ajst-12714 8 2 the the DET ajst-12714 8 3 deepening deepening NOUN ajst-12714 8 4 of of ADP ajst-12714 8 5 the the DET ajst-12714 8 6 research research NOUN ajst-12714 8 7 of of ADP ajst-12714 8 8 pipeline pipeline NOUN ajst-12714 8 9 magnetic magnetic ADJ ajst-12714 8 10 leakage leakage NOUN ajst-12714 8 11 detection detection NOUN ajst-12714 8 12 technology technology NOUN ajst-12714 8 13 , , PUNCT ajst-12714 8 14 image image NOUN ajst-12714 8 15 recognition recognition NOUN ajst-12714 8 16 technology technology NOUN ajst-12714 8 17 based base VERB ajst-12714 8 18 on on ADP ajst-12714 8 19 deep deep ADJ ajst-12714 8 20 learning learning NOUN ajst-12714 8 21 has have AUX ajst-12714 8 22 gradually gradually ADV ajst-12714 8 23 matured mature VERB ajst-12714 8 24 in in ADP ajst-12714 8 25 the the DET ajst-12714 8 26 field field NOUN ajst-12714 8 27 of of ADP ajst-12714 8 28 oil oil NOUN ajst-12714 8 29 and and CCONJ ajst-12714 8 30 gas gas NOUN ajst-12714 8 31 pipeline pipeline NOUN ajst-12714 8 32 defect defect NOUN ajst-12714 8 33 recognition recognition NOUN ajst-12714 8 34 . . PUNCT ajst-12714 9 1 in in ADP ajst-12714 9 2 this this DET ajst-12714 9 3 paper paper NOUN ajst-12714 9 4 , , PUNCT ajst-12714 9 5 a a DET ajst-12714 9 6 recursive recursive ADJ ajst-12714 9 7 graph graph NOUN ajst-12714 9 8 method method NOUN ajst-12714 9 9 is be AUX ajst-12714 9 10 proposed propose VERB ajst-12714 9 11 to to PART ajst-12714 9 12 convert convert VERB ajst-12714 9 13 the the DET ajst-12714 9 14 one one NUM ajst-12714 9 15 - - PUNCT ajst-12714 9 16 dimensional dimensional ADJ ajst-12714 9 17 magnetic magnetic ADJ ajst-12714 9 18 leakage leakage NOUN ajst-12714 9 19 signal signal NOUN ajst-12714 9 20 into into ADP ajst-12714 9 21 a a DET ajst-12714 9 22 two two NUM ajst-12714 9 23 - - PUNCT ajst-12714 9 24 dimensional dimensional ADJ ajst-12714 9 25 image image NOUN ajst-12714 9 26 , , PUNCT ajst-12714 9 27 and and CCONJ ajst-12714 9 28 then then ADV ajst-12714 9 29 use use VERB ajst-12714 9 30 the the DET ajst-12714 9 31 laplace laplace NOUN ajst-12714 9 32 pyramid pyramid NOUN ajst-12714 9 33 to to PART ajst-12714 9 34 regression regression VERB ajst-12714 9 35 prediction prediction NOUN ajst-12714 9 36 of of ADP ajst-12714 9 37 the the DET ajst-12714 9 38 image image NOUN ajst-12714 9 39 to to PART ajst-12714 9 40 improve improve VERB ajst-12714 9 41 the the DET ajst-12714 9 42 prediction prediction NOUN ajst-12714 9 43 accuracy accuracy NOUN ajst-12714 9 44 of of ADP ajst-12714 9 45 pipeline pipeline NOUN ajst-12714 9 46 defect defect NOUN ajst-12714 9 47 size size NOUN ajst-12714 9 48 . . PUNCT ajst-12714 10 1 2 2 X ajst-12714 10 2 . . X ajst-12714 10 3 finite finite PROPN ajst-12714 10 4 element element NOUN ajst-12714 10 5 analysis analysis NOUN ajst-12714 10 6 of of ADP ajst-12714 10 7 magnetic magnetic ADJ ajst-12714 10 8 flux flux NOUN ajst-12714 10 9 leakage leakage NOUN ajst-12714 10 10 detection detection NOUN ajst-12714 10 11 technology technology NOUN ajst-12714 10 12 for for ADP ajst-12714 10 13 pipeline pipeline NOUN ajst-12714 10 14 defects defect VERB ajst-12714 10 15 2.1 2.1 NUM ajst-12714 10 16 . . PUNCT ajst-12714 11 1 formation formation NOUN ajst-12714 11 2 of of ADP ajst-12714 11 3 leakage leakage NOUN ajst-12714 11 4 magnetic magnetic ADJ ajst-12714 11 5 field field NOUN ajst-12714 11 6 for for ADP ajst-12714 11 7 pipe pipe NOUN ajst-12714 11 8 defects defect NOUN ajst-12714 11 9 the the DET ajst-12714 11 10 pipeline pipeline NOUN ajst-12714 11 11 defect defect NOUN ajst-12714 11 12 flux flux NOUN ajst-12714 11 13 leakage leakage NOUN ajst-12714 11 14 detection detection NOUN ajst-12714 11 15 device device NOUN ajst-12714 11 16 is be AUX ajst-12714 11 17 composed compose VERB ajst-12714 11 18 of of ADP ajst-12714 11 19 an an DET ajst-12714 11 20 excitation excitation NOUN ajst-12714 11 21 source source NOUN ajst-12714 11 22 , , PUNCT ajst-12714 11 23 a a DET ajst-12714 11 24 pipeline pipeline NOUN ajst-12714 11 25 to to PART ajst-12714 11 26 be be AUX ajst-12714 11 27 measured measure VERB ajst-12714 11 28 and and CCONJ ajst-12714 11 29 a a DET ajst-12714 11 30 sensor sensor NOUN ajst-12714 11 31 , , PUNCT ajst-12714 11 32 and and CCONJ ajst-12714 11 33 its its PRON ajst-12714 11 34 working work VERB ajst-12714 11 35 principle principle NOUN ajst-12714 11 36 is be AUX ajst-12714 11 37 shown show VERB ajst-12714 11 38 in in ADP ajst-12714 11 39 figure figure NOUN ajst-12714 11 40 1 1 NUM ajst-12714 11 41 . . PUNCT ajst-12714 12 1 the the DET ajst-12714 12 2 specific specific ADJ ajst-12714 12 3 detection detection NOUN ajst-12714 12 4 process process NOUN ajst-12714 12 5 is be AUX ajst-12714 12 6 : : PUNCT ajst-12714 12 7 the the DET ajst-12714 12 8 excitation excitation NOUN ajst-12714 12 9 source source NOUN ajst-12714 12 10 first first ADV ajst-12714 12 11 locally locally ADV ajst-12714 12 12 magnetizes magnetize VERB ajst-12714 12 13 the the DET ajst-12714 12 14 pipe pipe NOUN ajst-12714 12 15 wall wall NOUN ajst-12714 12 16 to to ADP ajst-12714 12 17 a a DET ajst-12714 12 18 saturated saturated ADJ ajst-12714 12 19 state[3 state[3 NOUN ajst-12714 12 20 ] ] PUNCT ajst-12714 12 21 . . PUNCT ajst-12714 13 1 if if SCONJ ajst-12714 13 2 there there PRON ajst-12714 13 3 is be VERB ajst-12714 13 4 no no DET ajst-12714 13 5 defect defect NOUN ajst-12714 13 6 on on ADP ajst-12714 13 7 the the DET ajst-12714 13 8 surface surface NOUN ajst-12714 13 9 of of ADP ajst-12714 13 10 the the DET ajst-12714 13 11 pipe pipe NOUN ajst-12714 13 12 to to PART ajst-12714 13 13 be be AUX ajst-12714 13 14 measured measure VERB ajst-12714 13 15 , , PUNCT ajst-12714 13 16 the the DET ajst-12714 13 17 magnetic magnetic ADJ ajst-12714 13 18 induction induction NOUN ajst-12714 13 19 lines line NOUN ajst-12714 13 20 in in ADP ajst-12714 13 21 the the DET ajst-12714 13 22 pipe pipe NOUN ajst-12714 13 23 are be AUX ajst-12714 13 24 confined confine VERB ajst-12714 13 25 inside inside ADP ajst-12714 13 26 the the DET ajst-12714 13 27 pipe pipe NOUN ajst-12714 13 28 wall wall NOUN ajst-12714 13 29 , , PUNCT ajst-12714 13 30 as as SCONJ ajst-12714 13 31 shown show VERB ajst-12714 13 32 in in ADP ajst-12714 13 33 figure figure NOUN ajst-12714 13 34 1 1 NUM ajst-12714 13 35 ( ( PUNCT ajst-12714 13 36 a a NOUN ajst-12714 13 37 ) ) PUNCT ajst-12714 13 38 . . PUNCT ajst-12714 14 1 if if SCONJ ajst-12714 14 2 the the DET ajst-12714 14 3 test test NOUN ajst-12714 14 4 pipeline pipeline NOUN ajst-12714 14 5 has have VERB ajst-12714 14 6 a a DET ajst-12714 14 7 defect defect NOUN ajst-12714 14 8 , , PUNCT ajst-12714 14 9 the the DET ajst-12714 14 10 magnetic magnetic ADJ ajst-12714 14 11 permeability permeability NOUN ajst-12714 14 12 of of ADP ajst-12714 14 13 the the DET ajst-12714 14 14 air air NOUN ajst-12714 14 15 at at ADP ajst-12714 14 16 the the DET ajst-12714 14 17 defect defect NOUN ajst-12714 14 18 is be AUX ajst-12714 14 19 smaller small ADJ ajst-12714 14 20 than than ADP ajst-12714 14 21 that that PRON ajst-12714 14 22 of of ADP ajst-12714 14 23 the the DET ajst-12714 14 24 pipeline pipeline NOUN ajst-12714 14 25 , , PUNCT ajst-12714 14 26 and and CCONJ ajst-12714 14 27 the the DET ajst-12714 14 28 magnetic magnetic ADJ ajst-12714 14 29 induction induction NOUN ajst-12714 14 30 line line NOUN ajst-12714 14 31 at at ADP ajst-12714 14 32 the the DET ajst-12714 14 33 defect defect NOUN ajst-12714 14 34 will will AUX ajst-12714 14 35 be be AUX ajst-12714 14 36 refracted refract VERB ajst-12714 14 37 upwards upwards ADV ajst-12714 14 38 at at ADP ajst-12714 14 39 the the DET ajst-12714 14 40 defect defect NOUN ajst-12714 14 41 , , PUNCT ajst-12714 14 42 and and CCONJ ajst-12714 14 43 then then ADV ajst-12714 14 44 form form VERB ajst-12714 14 45 the the DET ajst-12714 14 46 leakage leakage NOUN ajst-12714 14 47 magnetic magnetic ADJ ajst-12714 14 48 field field NOUN ajst-12714 14 49 of of ADP ajst-12714 14 50 the the DET ajst-12714 14 51 defect defect NOUN ajst-12714 14 52 , , PUNCT ajst-12714 14 53 as as SCONJ ajst-12714 14 54 shown show VERB ajst-12714 14 55 in in ADP ajst-12714 14 56 figure figure NOUN ajst-12714 14 57 1 1 NUM ajst-12714 14 58 ( ( PUNCT ajst-12714 14 59 b b NOUN ajst-12714 14 60 ) ) PUNCT ajst-12714 14 61 . . PUNCT ajst-12714 15 1 ( ( PUNCT ajst-12714 15 2 a a X ajst-12714 15 3 ) ) PUNCT ajst-12714 15 4 is be AUX ajst-12714 15 5 free free ADJ ajst-12714 15 6 of of ADP ajst-12714 15 7 defects defect NOUN ajst-12714 15 8 ( ( PUNCT ajst-12714 15 9 b b NOUN ajst-12714 15 10 ) ) PUNCT ajst-12714 15 11 defective defective ADJ ajst-12714 15 12 figure figure NOUN ajst-12714 15 13 1 1 NUM ajst-12714 15 14 . . PUNCT ajst-12714 15 15 working work VERB ajst-12714 15 16 principle principle NOUN ajst-12714 15 17 of of ADP ajst-12714 15 18 pipeline pipeline NOUN ajst-12714 15 19 defect defect VERB ajst-12714 15 20 magnetic magnetic ADJ ajst-12714 15 21 leakage leakage NOUN ajst-12714 15 22 detector detector NOUN ajst-12714 15 23 2.2 2.2 NUM ajst-12714 15 24 . . PUNCT ajst-12714 16 1 establishment establishment NOUN ajst-12714 16 2 of of ADP ajst-12714 16 3 the the DET ajst-12714 16 4 ansys ansys PROPN ajst-12714 16 5 maxwell maxwell PROPN ajst-12714 16 6 finite finite PROPN ajst-12714 16 7 element element PROPN ajst-12714 16 8 model model NOUN ajst-12714 16 9 of of ADP ajst-12714 16 10 the the DET ajst-12714 16 11 leakage leakage NOUN ajst-12714 16 12 magnetic magnetic ADJ ajst-12714 16 13 field field NOUN ajst-12714 16 14 of of ADP ajst-12714 16 15 pipeline pipeline NOUN ajst-12714 16 16 defects defect VERB ajst-12714 16 17 the the DET ajst-12714 16 18 finite finite PROPN ajst-12714 16 19 element element NOUN ajst-12714 16 20 model model NOUN ajst-12714 16 21 of of ADP ajst-12714 16 22 ansys ansys PROPN ajst-12714 16 23 maxwell maxwell PROPN ajst-12714 16 24 constructs construct VERB ajst-12714 16 25 magnetic magnetic PROPN ajst-12714 16 26 flux flux PROPN ajst-12714 16 27 leakage leakage NOUN ajst-12714 16 28 detection detection NOUN ajst-12714 16 29 for for ADP ajst-12714 16 30 pipeline pipeline NOUN ajst-12714 16 31 defects defect NOUN ajst-12714 16 32 , , PUNCT ajst-12714 16 33 as as SCONJ ajst-12714 16 34 shown show VERB ajst-12714 16 35 in in ADP ajst-12714 16 36 figure figure NOUN ajst-12714 16 37 2 2 NUM ajst-12714 16 38 , , PUNCT ajst-12714 16 39 and and CCONJ ajst-12714 16 40 the the DET ajst-12714 16 41 designed design VERB ajst-12714 16 42 model model NOUN ajst-12714 16 43 consists consist VERB ajst-12714 16 44 of of ADP ajst-12714 16 45 iron iron NOUN ajst-12714 16 46 cores core NOUN ajst-12714 16 47 , , PUNCT ajst-12714 16 48 energized energized ADJ ajst-12714 16 49 coils coil NOUN ajst-12714 16 50 , , PUNCT ajst-12714 16 51 defective defective ADJ ajst-12714 16 52 pipes pipe NOUN ajst-12714 16 53 , , PUNCT ajst-12714 16 54 etc etc X ajst-12714 16 55 . . X ajst-12714 16 56 ansys ansys PROPN ajst-12714 16 57 maxwell maxwell PROPN ajst-12714 16 58 was be AUX ajst-12714 16 59 used use VERB ajst-12714 16 60 to to PART ajst-12714 16 61 build build VERB ajst-12714 16 62 a a DET ajst-12714 16 63 3d 3d NUM ajst-12714 16 64 pipeline pipeline NOUN ajst-12714 16 65 defect defect NOUN ajst-12714 16 66 leakage leakage NOUN ajst-12714 16 67 detection detection NOUN ajst-12714 16 68 model model NOUN ajst-12714 16 69 to to PART ajst-12714 16 70 capture capture VERB ajst-12714 16 71 the the DET ajst-12714 16 72 axial axial ADJ ajst-12714 16 73 and and CCONJ ajst-12714 16 74 radial radial ADJ ajst-12714 16 75 components component NOUN ajst-12714 16 76 of of ADP ajst-12714 16 77 the the DET ajst-12714 16 78 flux flux PROPN ajst-12714 16 79 leakage leakage NOUN ajst-12714 16 80 signal signal NOUN ajst-12714 16 81 . . PUNCT ajst-12714 17 1 figure figure NOUN ajst-12714 17 2 2 2 NUM ajst-12714 17 3 . . PUNCT ajst-12714 17 4 ansys ansys PROPN ajst-12714 17 5 model model PROPN ajst-12714 17 6 diagram diagram PROPN ajst-12714 17 7 of of ADP ajst-12714 17 8 magnetic magnetic ADJ ajst-12714 17 9 leakage leakage NOUN ajst-12714 17 10 detection detection NOUN ajst-12714 17 11 for for ADP ajst-12714 17 12 pipeline pipeline NOUN ajst-12714 17 13 defects defect NOUN ajst-12714 17 14 according accord VERB ajst-12714 17 15 to to ADP ajst-12714 17 16 the the DET ajst-12714 17 17 design design NOUN ajst-12714 17 18 of of ADP ajst-12714 17 19 the the DET ajst-12714 17 20 magnetic magnetic ADJ ajst-12714 17 21 flux flux PROPN ajst-12714 17 22 leakage leakage PROPN ajst-12714 17 23 simulation simulation NOUN ajst-12714 17 24 model model NOUN ajst-12714 17 25 , , PUNCT ajst-12714 17 26 the the DET ajst-12714 17 27 magnetic magnetic PROPN ajst-12714 17 28 flux flux PROPN ajst-12714 17 29 leakage leakage PROPN ajst-12714 17 30 signal signal NOUN ajst-12714 17 31 can can AUX ajst-12714 17 32 be be AUX ajst-12714 17 33 divided divide VERB ajst-12714 17 34 into into ADP ajst-12714 17 35 two two NUM ajst-12714 17 36 types type NOUN ajst-12714 17 37 : : PUNCT ajst-12714 17 38 radial radial ADJ ajst-12714 17 39 component component NOUN ajst-12714 17 40 and and CCONJ ajst-12714 17 41 axial axial ADJ ajst-12714 17 42 component component NOUN ajst-12714 17 43 . . PUNCT ajst-12714 18 1 as as SCONJ ajst-12714 18 2 shown show VERB ajst-12714 18 3 in in ADP ajst-12714 18 4 figure figure NOUN ajst-12714 18 5 3 3 NUM ajst-12714 18 6 , , PUNCT ajst-12714 18 7 the the DET ajst-12714 18 8 radial radial ADJ ajst-12714 18 9 magnetic magnetic ADJ ajst-12714 18 10 leakage leakage NOUN ajst-12714 18 11 component component NOUN ajst-12714 18 12 and and CCONJ ajst-12714 18 13 axial axial ADJ ajst-12714 18 14 magnetic magnetic ADJ ajst-12714 18 15 leakage leakage NOUN ajst-12714 18 16 component component NOUN ajst-12714 18 17 with with ADP ajst-12714 18 18 a a DET ajst-12714 18 19 depth depth NOUN ajst-12714 18 20 of of ADP ajst-12714 18 21 3 3 NUM ajst-12714 18 22 mm mm NOUN ajst-12714 18 23 , , PUNCT ajst-12714 18 24 a a DET ajst-12714 18 25 length length NOUN ajst-12714 18 26 of of ADP ajst-12714 18 27 4 4 NUM ajst-12714 18 28 mm mm NOUN ajst-12714 18 29 and and CCONJ ajst-12714 18 30 a a DET ajst-12714 18 31 width width NOUN ajst-12714 18 32 of of ADP ajst-12714 18 33 1~9 1~9 NUM ajst-12714 18 34 mm mm NOUN ajst-12714 18 35 are be AUX ajst-12714 18 36 depicted depict VERB ajst-12714 18 37 . . PUNCT ajst-12714 19 1 it it PRON ajst-12714 19 2 can can AUX ajst-12714 19 3 be be AUX ajst-12714 19 4 seen see VERB ajst-12714 19 5 from from ADP ajst-12714 19 6 the the DET ajst-12714 19 7 figure figure NOUN ajst-12714 19 8 that that SCONJ ajst-12714 19 9 the the DET ajst-12714 19 10 radial radial ADJ ajst-12714 19 11 leakage leakage NOUN ajst-12714 19 12 curve curve NOUN ajst-12714 19 13 has have VERB ajst-12714 19 14 a a DET ajst-12714 19 15 distinct distinct ADJ ajst-12714 19 16 peak peak NOUN ajst-12714 19 17 signal signal NOUN ajst-12714 19 18 and and CCONJ ajst-12714 19 19 a a DET ajst-12714 19 20 distinct distinct ADJ ajst-12714 19 21 valley valley NOUN ajst-12714 19 22 signal[4 signal[4 PROPN ajst-12714 19 23 ] ] PUNCT ajst-12714 19 24 . . PUNCT ajst-12714 20 1 the the DET ajst-12714 20 2 axial axial ADJ ajst-12714 20 3 flux flux NOUN ajst-12714 20 4 leakage leakage NOUN ajst-12714 20 5 curve curve NOUN ajst-12714 20 6 has have VERB ajst-12714 20 7 two two NUM ajst-12714 20 8 distinct distinct ADJ ajst-12714 20 9 peak peak NOUN ajst-12714 20 10 signals signal NOUN ajst-12714 20 11 and and CCONJ ajst-12714 20 12 35 35 NUM ajst-12714 20 13 two two NUM ajst-12714 20 14 valley valley NOUN ajst-12714 20 15 signals signal NOUN ajst-12714 20 16 , , PUNCT ajst-12714 20 17 and and CCONJ ajst-12714 20 18 a a DET ajst-12714 20 19 local local ADJ ajst-12714 20 20 valley valley NOUN ajst-12714 20 21 signal signal NOUN ajst-12714 20 22 between between ADP ajst-12714 20 23 the the DET ajst-12714 20 24 two two NUM ajst-12714 20 25 peaks peak NOUN ajst-12714 20 26 . . PUNCT ajst-12714 21 1 it it PRON ajst-12714 21 2 can can AUX ajst-12714 21 3 be be AUX ajst-12714 21 4 seen see VERB ajst-12714 21 5 from from ADP ajst-12714 21 6 the the DET ajst-12714 21 7 figure figure NOUN ajst-12714 21 8 that that SCONJ ajst-12714 21 9 when when SCONJ ajst-12714 21 10 the the DET ajst-12714 21 11 depth depth NOUN ajst-12714 21 12 and and CCONJ ajst-12714 21 13 length length NOUN ajst-12714 21 14 of of ADP ajst-12714 21 15 the the DET ajst-12714 21 16 defect defect NOUN ajst-12714 21 17 remain remain VERB ajst-12714 21 18 unchanged unchanged ADJ ajst-12714 21 19 , , PUNCT ajst-12714 21 20 the the PRON ajst-12714 21 21 larger large ADJ ajst-12714 21 22 the the DET ajst-12714 21 23 width width NOUN ajst-12714 21 24 of of ADP ajst-12714 21 25 the the DET ajst-12714 21 26 defect defect NOUN ajst-12714 21 27 , , PUNCT ajst-12714 21 28 the the PRON ajst-12714 21 29 larger large ADJ ajst-12714 21 30 the the DET ajst-12714 21 31 spacing spacing NOUN ajst-12714 21 32 between between ADP ajst-12714 21 33 positive positive ADJ ajst-12714 21 34 and and CCONJ ajst-12714 21 35 negative negative ADJ ajst-12714 21 36 extremums extremum NOUN ajst-12714 21 37 , , PUNCT ajst-12714 21 38 and and CCONJ ajst-12714 21 39 the the PRON ajst-12714 21 40 larger large ADJ ajst-12714 21 41 the the DET ajst-12714 21 42 peak peak NOUN ajst-12714 21 43 spacing spacing NOUN ajst-12714 21 44 and and CCONJ ajst-12714 21 45 valley valley NOUN ajst-12714 21 46 spacing spacing NOUN ajst-12714 21 47 . . PUNCT ajst-12714 22 1 ( ( PUNCT ajst-12714 22 2 a a X ajst-12714 22 3 ) ) PUNCT ajst-12714 22 4 radial radial ADJ ajst-12714 22 5 component component NOUN ajst-12714 22 6 of of ADP ajst-12714 22 7 defect defect ADJ ajst-12714 22 8 width width NOUN ajst-12714 22 9 change change NOUN ajst-12714 22 10 ( ( PUNCT ajst-12714 22 11 b b NOUN ajst-12714 22 12 ) ) PUNCT ajst-12714 22 13 axial axial ADJ ajst-12714 22 14 component component NOUN ajst-12714 22 15 of of ADP ajst-12714 22 16 defect defect ADJ ajst-12714 22 17 width width NOUN ajst-12714 22 18 change change NOUN ajst-12714 22 19 figure figure NOUN ajst-12714 22 20 3 3 NUM ajst-12714 22 21 . . PUNCT ajst-12714 22 22 field field NOUN ajst-12714 22 23 leakage leakage NOUN ajst-12714 22 24 curve curve NOUN ajst-12714 22 25 of of ADP ajst-12714 22 26 defect defect ADJ ajst-12714 22 27 width width NOUN ajst-12714 22 28 change change NOUN ajst-12714 22 29 as as SCONJ ajst-12714 22 30 shown show VERB ajst-12714 22 31 in in ADP ajst-12714 22 32 figure figure NOUN ajst-12714 22 33 4 4 NUM ajst-12714 22 34 , , PUNCT ajst-12714 22 35 the the DET ajst-12714 22 36 radial radial ADJ ajst-12714 22 37 leakage leakage NOUN ajst-12714 22 38 component component NOUN ajst-12714 22 39 and and CCONJ ajst-12714 22 40 axial axial ADJ ajst-12714 22 41 leakage leakage NOUN ajst-12714 22 42 component component NOUN ajst-12714 22 43 with with ADP ajst-12714 22 44 a a DET ajst-12714 22 45 depth depth NOUN ajst-12714 22 46 of of ADP ajst-12714 22 47 4 4 NUM ajst-12714 22 48 mm mm NOUN ajst-12714 22 49 , , PUNCT ajst-12714 22 50 a a DET ajst-12714 22 51 width width NOUN ajst-12714 22 52 of of ADP ajst-12714 22 53 4 4 NUM ajst-12714 22 54 mm mm NOUN ajst-12714 22 55 , , PUNCT ajst-12714 22 56 and and CCONJ ajst-12714 22 57 a a DET ajst-12714 22 58 depth depth NOUN ajst-12714 22 59 of of ADP ajst-12714 22 60 1~9 1~9 NUM ajst-12714 22 61 mm mm NOUN ajst-12714 22 62 are be AUX ajst-12714 22 63 depicted depict VERB ajst-12714 22 64 . . PUNCT ajst-12714 23 1 it it PRON ajst-12714 23 2 can can AUX ajst-12714 23 3 be be AUX ajst-12714 23 4 seen see VERB ajst-12714 23 5 from from ADP ajst-12714 23 6 the the DET ajst-12714 23 7 figure figure NOUN ajst-12714 23 8 that that SCONJ ajst-12714 23 9 when when SCONJ ajst-12714 23 10 the the DET ajst-12714 23 11 length length NOUN ajst-12714 23 12 and and CCONJ ajst-12714 23 13 width width NOUN ajst-12714 23 14 of of ADP ajst-12714 23 15 the the DET ajst-12714 23 16 defect defect NOUN ajst-12714 23 17 remain remain VERB ajst-12714 23 18 unchanged unchanged ADJ ajst-12714 23 19 , , PUNCT ajst-12714 23 20 the the PRON ajst-12714 23 21 deeper deep ADJ ajst-12714 23 22 the the DET ajst-12714 23 23 defect defect NOUN ajst-12714 23 24 depth depth NOUN ajst-12714 23 25 , , PUNCT ajst-12714 23 26 the the PRON ajst-12714 23 27 larger large ADJ ajst-12714 23 28 the the DET ajst-12714 23 29 signal signal ADJ ajst-12714 23 30 peak peak NOUN ajst-12714 23 31 . . PUNCT ajst-12714 24 1 ( ( PUNCT ajst-12714 24 2 a a X ajst-12714 24 3 ) ) PUNCT ajst-12714 24 4 radial radial ADJ ajst-12714 24 5 component component NOUN ajst-12714 24 6 of of ADP ajst-12714 24 7 defect defect ADJ ajst-12714 24 8 depth depth NOUN ajst-12714 24 9 change change NOUN ajst-12714 24 10 ( ( PUNCT ajst-12714 24 11 b b NOUN ajst-12714 24 12 ) ) PUNCT ajst-12714 24 13 axial axial ADJ ajst-12714 24 14 component component NOUN ajst-12714 24 15 of of ADP ajst-12714 24 16 defect defect ADJ ajst-12714 24 17 depth depth NOUN ajst-12714 24 18 change change NOUN ajst-12714 24 19 figure figure NOUN ajst-12714 24 20 4 4 NUM ajst-12714 24 21 . . PUNCT ajst-12714 24 22 field field NOUN ajst-12714 24 23 leakage leakage NOUN ajst-12714 24 24 curve curve NOUN ajst-12714 24 25 of of ADP ajst-12714 24 26 defect defect ADJ ajst-12714 24 27 depth depth NOUN ajst-12714 24 28 change change NOUN ajst-12714 24 29 2.3 2.3 NUM ajst-12714 24 30 . . PUNCT ajst-12714 25 1 eca eca NOUN ajst-12714 25 2 - - PUNCT ajst-12714 25 3 vgg16 vgg16 NOUN ajst-12714 25 4 neural neural ADJ ajst-12714 25 5 network network NOUN ajst-12714 25 6 structure structure NOUN ajst-12714 25 7 research research PROPN ajst-12714 25 8 vgg16 vgg16 PROPN ajst-12714 25 9 neural neural ADJ ajst-12714 25 10 network network NOUN ajst-12714 25 11 structure structure NOUN ajst-12714 25 12 the the DET ajst-12714 25 13 deep deep ADJ ajst-12714 25 14 learning learning NOUN ajst-12714 25 15 network network NOUN ajst-12714 25 16 model model NOUN ajst-12714 25 17 designed design VERB ajst-12714 25 18 for for ADP ajst-12714 25 19 intelligent intelligent ADJ ajst-12714 25 20 diagnosis diagnosis NOUN ajst-12714 25 21 of of ADP ajst-12714 25 22 pipeline pipeline NOUN ajst-12714 25 23 magnetic magnetic PROPN ajst-12714 25 24 flux flux PROPN ajst-12714 25 25 leakage leakage NOUN ajst-12714 25 26 detection detection NOUN ajst-12714 25 27 defects defect NOUN ajst-12714 25 28 is be AUX ajst-12714 25 29 based base VERB ajst-12714 25 30 on on ADP ajst-12714 25 31 vgg16 vgg16 PROPN ajst-12714 25 32 and and CCONJ ajst-12714 25 33 composed compose VERB ajst-12714 25 34 of of ADP ajst-12714 25 35 laplace laplace NOUN ajst-12714 25 36 pyramid[5 pyramid[5 PROPN ajst-12714 25 37 ] ] PUNCT ajst-12714 25 38 . . PUNCT ajst-12714 26 1 the the DET ajst-12714 26 2 overall overall ADJ ajst-12714 26 3 structure structure NOUN ajst-12714 26 4 of of ADP ajst-12714 26 5 the the DET ajst-12714 26 6 improved improved ADJ ajst-12714 26 7 vgg16 vgg16 NOUN ajst-12714 26 8 is be AUX ajst-12714 26 9 shown show VERB ajst-12714 26 10 in in ADP ajst-12714 26 11 figure figure NOUN ajst-12714 26 12 5 5 NUM ajst-12714 26 13 . . PUNCT ajst-12714 27 1 7 7 NUM ajst-12714 27 2 * * SYM ajst-12714 27 3 7 7 NUM ajst-12714 27 4 * * SYM ajst-12714 27 5 512 512 NUM ajst-12714 27 6 1 1 NUM ajst-12714 27 7 * * NUM ajst-12714 27 8 1 1 NUM ajst-12714 27 9 * * SYM ajst-12714 27 10 4096 4096 NUM ajst-12714 27 11 1 1 NUM ajst-12714 27 12 * * SYM ajst-12714 27 13 1 1 NUM ajst-12714 27 14 * * NUM ajst-12714 27 15 1000 1000 NUM ajst-12714 27 16 14 14 NUM ajst-12714 27 17 * * SYM ajst-12714 27 18 14 14 NUM ajst-12714 27 19 * * SYM ajst-12714 27 20 512 512 NUM ajst-12714 27 21 28 28 NUM ajst-12714 27 22 * * SYM ajst-12714 27 23 28 28 NUM ajst-12714 27 24 * * SYM ajst-12714 27 25 512 512 NUM ajst-12714 27 26 56 56 NUM ajst-12714 27 27 * * SYM ajst-12714 27 28 56 56 NUM ajst-12714 27 29 * * SYM ajst-12714 27 30 256 256 NUM ajst-12714 27 31 112 112 NUM ajst-12714 27 32 * * NUM ajst-12714 27 33 112 112 NUM ajst-12714 27 34 * * SYM ajst-12714 27 35 128 128 NUM ajst-12714 27 36 224 224 NUM ajst-12714 27 37 * * NUM ajst-12714 27 38 224 224 NUM ajst-12714 27 39 * * SYM ajst-12714 27 40 64224 64224 NUM ajst-12714 27 41 * * SYM ajst-12714 27 42 224 224 NUM ajst-12714 27 43 * * SYM ajst-12714 27 44 3224 3224 NUM ajst-12714 27 45 * * SYM ajst-12714 27 46 224 224 NUM ajst-12714 27 47 * * SYM ajst-12714 27 48 3 3 NUM ajst-12714 27 49 figure figure NOUN ajst-12714 27 50 5 5 NUM ajst-12714 27 51 . . PUNCT ajst-12714 27 52 overall overall ADJ ajst-12714 27 53 structure structure NOUN ajst-12714 27 54 diagram diagram NOUN ajst-12714 27 55 of of ADP ajst-12714 27 56 vgg16 vgg16 PROPN ajst-12714 27 57 vgg16 vgg16 PROPN ajst-12714 27 58 is be AUX ajst-12714 27 59 a a DET ajst-12714 27 60 convolutional convolutional ADJ ajst-12714 27 61 neural neural ADJ ajst-12714 27 62 network network NOUN ajst-12714 27 63 model model NOUN ajst-12714 27 64 proposed propose VERB ajst-12714 27 65 by by ADP ajst-12714 27 66 simonyan simonyan ADJ ajst-12714 27 67 and and CCONJ ajst-12714 27 68 zisserman zisserman NOUN ajst-12714 27 69 in in ADP ajst-12714 27 70 the the DET ajst-12714 27 71 visual visual ADJ ajst-12714 27 72 geometry geometry NOUN ajst-12714 27 73 group group NOUN ajst-12714 27 74 group group NOUN ajst-12714 27 75 , , PUNCT ajst-12714 27 76 which which PRON ajst-12714 27 77 is be AUX ajst-12714 27 78 exclusive exclusive ADJ ajst-12714 27 79 to to ADP ajst-12714 27 80 the the DET ajst-12714 27 81 convolutional convolutional ADJ ajst-12714 27 82 layer layer NOUN ajst-12714 27 83 in in ADP ajst-12714 27 84 the the DET ajst-12714 27 85 vgg16 vgg16 NOUN ajst-12714 27 86 network13 network13 ADJ ajst-12714 27 87 layers layer NOUN ajst-12714 27 88 , , PUNCT ajst-12714 27 89 the the DET ajst-12714 27 90 remaining remain VERB ajst-12714 27 91 3 3 NUM ajst-12714 27 92 layers layer NOUN ajst-12714 27 93 are be AUX ajst-12714 27 94 fully fully ADV ajst-12714 27 95 connected connected ADJ ajst-12714 27 96 layers layer NOUN ajst-12714 27 97 , , PUNCT ajst-12714 27 98 the the DET ajst-12714 27 99 model model NOUN ajst-12714 27 100 according accord VERB ajst-12714 27 101 to to ADP ajst-12714 27 102 the the DET ajst-12714 27 103 convolution convolution NOUN ajst-12714 27 104 kernel kernel NOUN ajst-12714 27 105 size size NOUN ajst-12714 27 106 and and CCONJ ajst-12714 27 107 the the DET ajst-12714 27 108 number number NOUN ajst-12714 27 109 of of ADP ajst-12714 27 110 convolutional convolutional ADJ ajst-12714 27 111 layers layer NOUN ajst-12714 27 112 , , PUNCT ajst-12714 27 113 vgg vgg PROPN ajst-12714 27 114 can can AUX ajst-12714 27 115 be be AUX ajst-12714 27 116 divided divide VERB ajst-12714 27 117 into into ADP ajst-12714 27 118 6 6 NUM ajst-12714 27 119 seed seed NOUN ajst-12714 27 120 models model NOUN ajst-12714 27 121 , , PUNCT ajst-12714 27 122 respectivelya respectivelya NOUN ajst-12714 27 123 , , PUNCT ajst-12714 27 124 a a DET ajst-12714 27 125 - - PUNCT ajst-12714 27 126 lrn lrn PROPN ajst-12714 27 127 , , PUNCT ajst-12714 27 128 b b NOUN ajst-12714 27 129 , , PUNCT ajst-12714 27 130 c c NOUN ajst-12714 27 131 , , PUNCT ajst-12714 27 132 d d NOUN ajst-12714 27 133 , , PUNCT ajst-12714 27 134 e e NOUN ajst-12714 27 135 , , PUNCT ajst-12714 27 136 the the DET ajst-12714 27 137 models model NOUN ajst-12714 27 138 we we PRON ajst-12714 27 139 often often ADV ajst-12714 27 140 see see VERB ajst-12714 27 141 and and CCONJ ajst-12714 27 142 use use VERB ajst-12714 27 143 are be AUX ajst-12714 27 144 basicallyd basicallyd ADJ ajst-12714 27 145 and and CCONJ ajst-12714 27 146 e e NOUN ajst-12714 27 147 models model NOUN ajst-12714 27 148 . . PUNCT ajst-12714 28 1 the the DET ajst-12714 28 2 image image NOUN ajst-12714 28 3 with with ADP ajst-12714 28 4 the the DET ajst-12714 28 5 original original ADJ ajst-12714 28 6 input input NOUN ajst-12714 28 7 size size NOUN ajst-12714 28 8 of of ADP ajst-12714 28 9 the the DET ajst-12714 28 10 network network NOUN ajst-12714 28 11 is be AUX ajst-12714 28 12 2 2 NUM ajst-12714 28 13 24 24 NUM ajst-12714 28 14 * * NUM ajst-12714 28 15 224 224 NUM ajst-12714 28 16 first first ADV ajst-12714 28 17 through through ADP ajst-12714 28 18 two two NUM ajst-12714 28 19 times time NOUN ajst-12714 28 20 of of ADP ajst-12714 28 21 64 64 NUM ajst-12714 28 22 convolution convolution NOUN ajst-12714 28 23 kernels kernel NOUN ajst-12714 28 24 of of ADP ajst-12714 28 25 the the DET ajst-12714 28 26 same same ADJ ajst-12714 28 27 size size NOUN ajst-12714 28 28 for for ADP ajst-12714 28 29 convolution convolution NOUN ajst-12714 28 30 operation operation NOUN ajst-12714 28 31 , , PUNCT ajst-12714 28 32 and and CCONJ ajst-12714 28 33 one one NUM ajst-12714 28 34 pooling pool VERB ajst-12714 28 35 operation operation NOUN ajst-12714 28 36 . . PUNCT ajst-12714 29 1 the the DET ajst-12714 29 2 resulting result VERB ajst-12714 29 3 feature feature NOUN ajst-12714 29 4 map map NOUN ajst-12714 29 5 is be AUX ajst-12714 29 6 1/2 1/2 NUM ajst-12714 29 7 of of ADP ajst-12714 29 8 the the DET ajst-12714 29 9 original original ADJ ajst-12714 29 10 map map NOUN ajst-12714 29 11 ; ; PUNCT ajst-12714 29 12 after after ADP ajst-12714 29 13 two two NUM ajst-12714 29 14 convolution convolution NOUN ajst-12714 29 15 operations operation NOUN ajst-12714 29 16 of of ADP ajst-12714 29 17 1 1 NUM ajst-12714 29 18 28 28 NUM ajst-12714 29 19 sizes size NOUN ajst-12714 29 20 and and CCONJ ajst-12714 29 21 one one NUM ajst-12714 29 22 pooling pool VERB ajst-12714 29 23 operation operation NOUN ajst-12714 29 24 , , PUNCT ajst-12714 29 25 the the DET ajst-12714 29 26 feature feature NOUN ajst-12714 29 27 map map NOUN ajst-12714 29 28 obtained obtain VERB ajst-12714 29 29 is be AUX ajst-12714 29 30 1/4 1/4 NUM ajst-12714 29 31 of of ADP ajst-12714 29 32 the the DET ajst-12714 29 33 original original ADJ ajst-12714 29 34 figure figure NOUN ajst-12714 29 35 ; ; PUNCT ajst-12714 29 36 after after ADP ajst-12714 29 37 three three NUM ajst-12714 29 38 times time NOUN ajst-12714 29 39 of of ADP ajst-12714 29 40 256 256 NUM ajst-12714 29 41 convolution convolution NOUN ajst-12714 29 42 operations operation NOUN ajst-12714 29 43 of of ADP ajst-12714 29 44 the the DET ajst-12714 29 45 same same ADJ ajst-12714 29 46 size size NOUN ajst-12714 29 47 , , PUNCT ajst-12714 29 48 and and CCONJ ajst-12714 29 49 one one NUM ajst-12714 29 50 pooling pool VERB ajst-12714 29 51 operation operation NOUN ajst-12714 29 52 , , PUNCT ajst-12714 29 53 the the DET ajst-12714 29 54 obtained obtain VERB ajst-12714 29 55 feature feature NOUN ajst-12714 29 56 map map NOUN ajst-12714 29 57 is be AUX ajst-12714 29 58 1/8 1/8 NUM ajst-12714 29 59 of of ADP ajst-12714 29 60 the the DET ajst-12714 29 61 original original NOUN ajst-12714 29 62 ; ; PUNCT ajst-12714 29 63 after after ADP ajst-12714 29 64 three three NUM ajst-12714 29 65 more more ADJ ajst-12714 29 66 512 512 NUM ajst-12714 29 67 convolution convolution NOUN ajst-12714 29 68 operations operation NOUN ajst-12714 29 69 of of ADP ajst-12714 29 70 the the DET ajst-12714 29 71 same same ADJ ajst-12714 29 72 size size NOUN ajst-12714 29 73 and and CCONJ ajst-12714 29 74 one one NUM ajst-12714 29 75 pooling pool VERB ajst-12714 29 76 36 36 NUM ajst-12714 29 77 operation operation NOUN ajst-12714 29 78 , , PUNCT ajst-12714 29 79 the the DET ajst-12714 29 80 feature feature NOUN ajst-12714 29 81 map map NOUN ajst-12714 29 82 obtained obtain VERB ajst-12714 29 83 is be AUX ajst-12714 29 84 1/32 1/32 NUM ajst-12714 29 85 of of ADP ajst-12714 29 86 the the DET ajst-12714 29 87 original original NOUN ajst-12714 29 88 ; ; PUNCT ajst-12714 29 89 finally finally ADV ajst-12714 29 90 , , PUNCT ajst-12714 29 91 after after ADP ajst-12714 29 92 three three NUM ajst-12714 29 93 fully fully ADV ajst-12714 29 94 connected connect VERB ajst-12714 29 95 layer layer NOUN ajst-12714 29 96 operations operation NOUN ajst-12714 29 97 , , PUNCT ajst-12714 29 98 the the DET ajst-12714 29 99 obtained obtain VERB ajst-12714 29 100 results result NOUN ajst-12714 29 101 are be AUX ajst-12714 29 102 predicted predict VERB ajst-12714 29 103 by by ADP ajst-12714 29 104 the the DET ajst-12714 29 105 soft soft ADJ ajst-12714 29 106 - - PUNCT ajst-12714 29 107 max max NOUN ajst-12714 29 108 prediction prediction NOUN ajst-12714 29 109 layer[6 layer[6 PROPN ajst-12714 29 110 ] ] PUNCT ajst-12714 29 111 . . PUNCT ajst-12714 30 1 the the DET ajst-12714 30 2 block block NOUN ajst-12714 30 3 diagram diagram NOUN ajst-12714 30 4 is be AUX ajst-12714 30 5 shown show VERB ajst-12714 30 6 in in ADP ajst-12714 30 7 figure figure NOUN ajst-12714 30 8 6 6 NUM ajst-12714 30 9 . . PUNCT ajst-12714 31 1 figure figure VERB ajst-12714 31 2 6 6 NUM ajst-12714 31 3 . . PUNCT ajst-12714 31 4 block block NOUN ajst-12714 31 5 diagram diagram NOUN ajst-12714 31 6 of of ADP ajst-12714 31 7 vgg16 vgg16 NOUN ajst-12714 31 8 system system NOUN ajst-12714 31 9 2.3.1 2.3.1 NUM ajst-12714 31 10 . . PUNCT ajst-12714 32 1 convolutional convolutional ADJ ajst-12714 32 2 layers layer NOUN ajst-12714 32 3 the the DET ajst-12714 32 4 convolutional convolutional ADJ ajst-12714 32 5 layer layer NOUN ajst-12714 32 6 of of ADP ajst-12714 32 7 vgg16 vgg16 NOUN ajst-12714 32 8 is be AUX ajst-12714 32 9 the the DET ajst-12714 32 10 core core ADJ ajst-12714 32 11 part part NOUN ajst-12714 32 12 of of ADP ajst-12714 32 13 the the DET ajst-12714 32 14 network network NOUN ajst-12714 32 15 and and CCONJ ajst-12714 32 16 is be AUX ajst-12714 32 17 responsible responsible ADJ ajst-12714 32 18 for for ADP ajst-12714 32 19 extracting extract VERB ajst-12714 32 20 the the DET ajst-12714 32 21 features feature NOUN ajst-12714 32 22 of of ADP ajst-12714 32 23 the the DET ajst-12714 32 24 input input NOUN ajst-12714 32 25 image image NOUN ajst-12714 32 26 . . PUNCT ajst-12714 33 1 vgg16 vgg16 PROPN ajst-12714 33 2 has have VERB ajst-12714 33 3 a a DET ajst-12714 33 4 total total NOUN ajst-12714 33 5 of of ADP ajst-12714 33 6 13 13 NUM ajst-12714 33 7 convolutional convolutional ADJ ajst-12714 33 8 layers layer NOUN ajst-12714 33 9 , , PUNCT ajst-12714 33 10 each each PRON ajst-12714 33 11 of of ADP ajst-12714 33 12 which which PRON ajst-12714 33 13 uses use VERB ajst-12714 33 14 a a DET ajst-12714 33 15 3x3 3x3 NUM ajst-12714 33 16 convolution convolution NOUN ajst-12714 33 17 kernel kernel NOUN ajst-12714 33 18 for for ADP ajst-12714 33 19 feature feature NOUN ajst-12714 33 20 extraction extraction NOUN ajst-12714 33 21 and and CCONJ ajst-12714 33 22 a a DET ajst-12714 33 23 nonlinear nonlinear ADJ ajst-12714 33 24 transformation transformation NOUN ajst-12714 33 25 using use VERB ajst-12714 33 26 the the DET ajst-12714 33 27 relu relu NOUN ajst-12714 33 28 activation activation NOUN ajst-12714 33 29 function function NOUN ajst-12714 33 30 . . PUNCT ajst-12714 34 1 the the DET ajst-12714 34 2 feature feature NOUN ajst-12714 34 3 map map NOUN ajst-12714 34 4 output output NOUN ajst-12714 34 5 by by ADP ajst-12714 34 6 the the DET ajst-12714 34 7 final final ADJ ajst-12714 34 8 convolutional convolutional ADJ ajst-12714 34 9 layer layer NOUN ajst-12714 34 10 will will AUX ajst-12714 34 11 be be AUX ajst-12714 34 12 classified classify VERB ajst-12714 34 13 as as ADP ajst-12714 34 14 input input NOUN ajst-12714 34 15 to to ADP ajst-12714 34 16 the the DET ajst-12714 34 17 fully fully ADV ajst-12714 34 18 connected connected ADJ ajst-12714 34 19 layer[7 layer[7 NOUN ajst-12714 34 20 ] ] PUNCT ajst-12714 34 21 . . PUNCT ajst-12714 35 1 by by ADP ajst-12714 35 2 stacking stack VERB ajst-12714 35 3 small small ADJ ajst-12714 35 4 - - PUNCT ajst-12714 35 5 sized sized ADJ ajst-12714 35 6 convolutional convolutional ADJ ajst-12714 35 7 kernels kernel NOUN ajst-12714 35 8 and and CCONJ ajst-12714 35 9 pooling pool VERB ajst-12714 35 10 layers layer NOUN ajst-12714 35 11 multiple multiple ADJ ajst-12714 35 12 times time NOUN ajst-12714 35 13 , , PUNCT ajst-12714 35 14 vgg16 vgg16 PROPN ajst-12714 35 15 can can AUX ajst-12714 35 16 effectively effectively ADV ajst-12714 35 17 increase increase VERB ajst-12714 35 18 the the DET ajst-12714 35 19 depth depth NOUN ajst-12714 35 20 of of ADP ajst-12714 35 21 the the DET ajst-12714 35 22 network network NOUN ajst-12714 35 23 and and CCONJ ajst-12714 35 24 improve improve VERB ajst-12714 35 25 the the DET ajst-12714 35 26 accuracy accuracy NOUN ajst-12714 35 27 of of ADP ajst-12714 35 28 image image NOUN ajst-12714 35 29 recognition recognition NOUN ajst-12714 35 30 . . PUNCT ajst-12714 36 1 2.3.2 2.3.2 X ajst-12714 36 2 . . X ajst-12714 36 3 pooling pool VERB ajst-12714 36 4 layer layer NOUN ajst-12714 36 5 the the DET ajst-12714 36 6 pooling pool VERB ajst-12714 36 7 layer layer NOUN ajst-12714 36 8 in in ADP ajst-12714 36 9 vgg16 vgg16 NOUN ajst-12714 36 10 uses use VERB ajst-12714 36 11 a a DET ajst-12714 36 12 2x2 2x2 NUM ajst-12714 36 13 sliding slide VERB ajst-12714 36 14 window window NOUN ajst-12714 36 15 by by ADP ajst-12714 36 16 using use VERB ajst-12714 36 17 the the DET ajst-12714 36 18 maximum maximum ADJ ajst-12714 36 19 value value NOUN ajst-12714 36 20 within within ADP ajst-12714 36 21 the the DET ajst-12714 36 22 window window NOUN ajst-12714 36 23 as as ADP ajst-12714 36 24 the the DET ajst-12714 36 25 pooled pooled ADJ ajst-12714 36 26 value value NOUN ajst-12714 36 27 . . PUNCT ajst-12714 37 1 doing do VERB ajst-12714 37 2 so so ADV ajst-12714 37 3 reduces reduce VERB ajst-12714 37 4 the the DET ajst-12714 37 5 size size NOUN ajst-12714 37 6 of of ADP ajst-12714 37 7 the the DET ajst-12714 37 8 feature feature NOUN ajst-12714 37 9 map map NOUN ajst-12714 37 10 while while SCONJ ajst-12714 37 11 keeping keep VERB ajst-12714 37 12 the the DET ajst-12714 37 13 spatial spatial ADJ ajst-12714 37 14 position position NOUN ajst-12714 37 15 of of ADP ajst-12714 37 16 the the DET ajst-12714 37 17 feature feature NOUN ajst-12714 37 18 unchanged unchanged ADJ ajst-12714 37 19 . . PUNCT ajst-12714 38 1 the the DET ajst-12714 38 2 pooling pool VERB ajst-12714 38 3 operation operation NOUN ajst-12714 38 4 can can AUX ajst-12714 38 5 effectively effectively ADV ajst-12714 38 6 reduce reduce VERB ajst-12714 38 7 the the DET ajst-12714 38 8 dimension dimension NOUN ajst-12714 38 9 of of ADP ajst-12714 38 10 the the DET ajst-12714 38 11 feature feature NOUN ajst-12714 38 12 map map NOUN ajst-12714 38 13 and and CCONJ ajst-12714 38 14 extract extract VERB ajst-12714 38 15 more more ADJ ajst-12714 38 16 abstract abstract ADJ ajst-12714 38 17 and and CCONJ ajst-12714 38 18 important important ADJ ajst-12714 38 19 features feature NOUN ajst-12714 38 20 . . PUNCT ajst-12714 39 1 the the DET ajst-12714 39 2 pooling pool VERB ajst-12714 39 3 layer layer NOUN ajst-12714 39 4 in in ADP ajst-12714 39 5 vgg16 vgg16 NOUN ajst-12714 39 6 is be AUX ajst-12714 39 7 usually usually ADV ajst-12714 39 8 immediately immediately ADV ajst-12714 39 9 followed follow VERB ajst-12714 39 10 by by ADP ajst-12714 39 11 the the DET ajst-12714 39 12 convolutional convolutional ADJ ajst-12714 39 13 layer layer NOUN ajst-12714 39 14 and and CCONJ ajst-12714 39 15 is be AUX ajst-12714 39 16 used use VERB ajst-12714 39 17 to to PART ajst-12714 39 18 gradually gradually ADV ajst-12714 39 19 reduce reduce VERB ajst-12714 39 20 the the DET ajst-12714 39 21 size size NOUN ajst-12714 39 22 of of ADP ajst-12714 39 23 the the DET ajst-12714 39 24 feature feature NOUN ajst-12714 39 25 map[8 map[8 NOUN ajst-12714 39 26 ] ] PUNCT ajst-12714 39 27 . . PUNCT ajst-12714 40 1 this this DET ajst-12714 40 2 design design NOUN ajst-12714 40 3 allows allow VERB ajst-12714 40 4 the the DET ajst-12714 40 5 network network NOUN ajst-12714 40 6 to to PART ajst-12714 40 7 learn learn VERB ajst-12714 40 8 more more ADV ajst-12714 40 9 abstract abstract ADJ ajst-12714 40 10 and and CCONJ ajst-12714 40 11 complex complex ADJ ajst-12714 40 12 features feature NOUN ajst-12714 40 13 at at ADP ajst-12714 40 14 a a DET ajst-12714 40 15 higher high ADJ ajst-12714 40 16 level level NOUN ajst-12714 40 17 , , PUNCT ajst-12714 40 18 thereby thereby ADV ajst-12714 40 19 improving improve VERB ajst-12714 40 20 the the DET ajst-12714 40 21 performance performance NOUN ajst-12714 40 22 of of ADP ajst-12714 40 23 the the DET ajst-12714 40 24 network[9 network[9 PROPN ajst-12714 40 25 ] ] PUNCT ajst-12714 40 26 . . PUNCT ajst-12714 41 1 the the DET ajst-12714 41 2 role role NOUN ajst-12714 41 3 of of ADP ajst-12714 41 4 pooling pool VERB ajst-12714 41 5 layers layer NOUN ajst-12714 41 6 is be AUX ajst-12714 41 7 to to PART ajst-12714 41 8 reduce reduce VERB ajst-12714 41 9 the the DET ajst-12714 41 10 computational computational ADJ ajst-12714 41 11 complexity complexity NOUN ajst-12714 41 12 of of ADP ajst-12714 41 13 subsequent subsequent ADJ ajst-12714 41 14 layers layer NOUN ajst-12714 41 15 by by ADP ajst-12714 41 16 reducing reduce VERB ajst-12714 41 17 the the DET ajst-12714 41 18 size size NOUN ajst-12714 41 19 of of ADP ajst-12714 41 20 the the DET ajst-12714 41 21 feature feature NOUN ajst-12714 41 22 map map NOUN ajst-12714 41 23 , , PUNCT ajst-12714 41 24 and and CCONJ ajst-12714 41 25 it it PRON ajst-12714 41 26 helps help VERB ajst-12714 41 27 to to PART ajst-12714 41 28 extract extract VERB ajst-12714 41 29 more more ADJ ajst-12714 41 30 robust robust ADJ ajst-12714 41 31 features feature NOUN ajst-12714 41 32 . . PUNCT ajst-12714 42 1 in in ADP ajst-12714 42 2 vgg16 vgg16 PROPN ajst-12714 42 3 , , PUNCT ajst-12714 42 4 each each DET ajst-12714 42 5 pooling pool VERB ajst-12714 42 6 layer layer NOUN ajst-12714 42 7 uses use VERB ajst-12714 42 8 a a DET ajst-12714 42 9 2x2 2x2 NUM ajst-12714 42 10 pooling pool VERB ajst-12714 42 11 window window NOUN ajst-12714 42 12 in in ADP ajst-12714 42 13 steps step NOUN ajst-12714 42 14 of of ADP ajst-12714 42 15 2 2 NUM ajst-12714 42 16 , , PUNCT ajst-12714 42 17 so so SCONJ ajst-12714 42 18 that that SCONJ ajst-12714 42 19 the the DET ajst-12714 42 20 size size NOUN ajst-12714 42 21 of of ADP ajst-12714 42 22 the the DET ajst-12714 42 23 pooled pool VERB ajst-12714 42 24 feature feature NOUN ajst-12714 42 25 map map NOUN ajst-12714 42 26 is be AUX ajst-12714 42 27 reduced reduce VERB ajst-12714 42 28 by by ADP ajst-12714 42 29 half half NOUN ajst-12714 42 30 . . PUNCT ajst-12714 43 1 at at ADP ajst-12714 43 2 the the DET ajst-12714 43 3 same same ADJ ajst-12714 43 4 time time NOUN ajst-12714 43 5 , , PUNCT ajst-12714 43 6 the the DET ajst-12714 43 7 max max PROPN ajst-12714 43 8 pooling pooling NOUN ajst-12714 43 9 operation operation NOUN ajst-12714 43 10 is be AUX ajst-12714 43 11 used use VERB ajst-12714 43 12 to to PART ajst-12714 43 13 retain retain VERB ajst-12714 43 14 the the DET ajst-12714 43 15 maximum maximum ADJ ajst-12714 43 16 value value NOUN ajst-12714 43 17 in in ADP ajst-12714 43 18 each each DET ajst-12714 43 19 pooling pool VERB ajst-12714 43 20 window window NOUN ajst-12714 43 21 to to PART ajst-12714 43 22 ensure ensure VERB ajst-12714 43 23 that that SCONJ ajst-12714 43 24 important important ADJ ajst-12714 43 25 characteristic characteristic ADJ ajst-12714 43 26 information information NOUN ajst-12714 43 27 is be AUX ajst-12714 43 28 retained retain VERB ajst-12714 43 29 . . PUNCT ajst-12714 44 1 by by ADP ajst-12714 44 2 stacking stack VERB ajst-12714 44 3 convolutional convolutional ADJ ajst-12714 44 4 layers layer NOUN ajst-12714 44 5 and and CCONJ ajst-12714 44 6 pooling pool VERB ajst-12714 44 7 layers layer NOUN ajst-12714 44 8 multiple multiple ADJ ajst-12714 44 9 times time NOUN ajst-12714 44 10 , , PUNCT ajst-12714 44 11 the the DET ajst-12714 44 12 vgg16 vgg16 NOUN ajst-12714 44 13 network network NOUN ajst-12714 44 14 can can AUX ajst-12714 44 15 gradually gradually ADV ajst-12714 44 16 reduce reduce VERB ajst-12714 44 17 the the DET ajst-12714 44 18 size size NOUN ajst-12714 44 19 of of ADP ajst-12714 44 20 the the DET ajst-12714 44 21 feature feature NOUN ajst-12714 44 22 map map NOUN ajst-12714 44 23 and and CCONJ ajst-12714 44 24 extract extract VERB ajst-12714 44 25 higher high ADJ ajst-12714 44 26 - - PUNCT ajst-12714 44 27 level level NOUN ajst-12714 44 28 abstract abstract ADJ ajst-12714 44 29 features feature NOUN ajst-12714 44 30 at at ADP ajst-12714 44 31 the the DET ajst-12714 44 32 same same ADJ ajst-12714 44 33 time time NOUN ajst-12714 44 34 , , PUNCT ajst-12714 44 35 so so SCONJ ajst-12714 44 36 that that SCONJ ajst-12714 44 37 the the DET ajst-12714 44 38 network network NOUN ajst-12714 44 39 has have VERB ajst-12714 44 40 better well ADJ ajst-12714 44 41 receptive receptive ADJ ajst-12714 44 42 fields field NOUN ajst-12714 44 43 and and CCONJ ajst-12714 44 44 stronger strong ADJ ajst-12714 44 45 expression expression NOUN ajst-12714 44 46 ability ability NOUN ajst-12714 44 47 . . PUNCT ajst-12714 45 1 this this DET ajst-12714 45 2 structural structural ADJ ajst-12714 45 3 design design NOUN ajst-12714 45 4 enables enables AUX ajst-12714 45 5 vgg16 vgg16 VERB ajst-12714 45 6 to to PART ajst-12714 45 7 achieve achieve VERB ajst-12714 45 8 good good ADJ ajst-12714 45 9 performance performance NOUN ajst-12714 45 10 in in ADP ajst-12714 45 11 computer computer NOUN ajst-12714 45 12 vision vision NOUN ajst-12714 45 13 tasks task NOUN ajst-12714 45 14 such such ADJ ajst-12714 45 15 as as ADP ajst-12714 45 16 image image NOUN ajst-12714 45 17 classification classification NOUN ajst-12714 45 18 . . PUNCT ajst-12714 46 1 2.3.3 2.3.3 X ajst-12714 46 2 . . X ajst-12714 46 3 normalization normalization NOUN ajst-12714 46 4 instead instead ADV ajst-12714 46 5 of of ADP ajst-12714 46 6 an an DET ajst-12714 46 7 explicit explicit ADJ ajst-12714 46 8 normalization normalization NOUN ajst-12714 46 9 layer layer NOUN ajst-12714 46 10 , , PUNCT ajst-12714 46 11 the the DET ajst-12714 46 12 vgg16 vgg16 NOUN ajst-12714 46 13 model model NOUN ajst-12714 46 14 uses use VERB ajst-12714 46 15 a a DET ajst-12714 46 16 local local ADJ ajst-12714 46 17 response response NOUN ajst-12714 46 18 normalization normalization NOUN ajst-12714 46 19 method method NOUN ajst-12714 46 20 called call VERB ajst-12714 46 21 " " PUNCT ajst-12714 46 22 local local ADJ ajst-12714 46 23 response response NOUN ajst-12714 46 24 normalization normalization NOUN ajst-12714 46 25 " " PUNCT ajst-12714 46 26 ( ( PUNCT ajst-12714 46 27 lrn lrn PROPN ajst-12714 46 28 ) ) PUNCT ajst-12714 46 29 . . PUNCT ajst-12714 47 1 in in ADP ajst-12714 47 2 the the DET ajst-12714 47 3 design design NOUN ajst-12714 47 4 of of ADP ajst-12714 47 5 vgg16 vgg16 PROPN ajst-12714 47 6 , , PUNCT ajst-12714 47 7 lrn lrn PROPN ajst-12714 47 8 operations operation NOUN ajst-12714 47 9 are be AUX ajst-12714 47 10 used use VERB ajst-12714 47 11 to to PART ajst-12714 47 12 enhance enhance VERB ajst-12714 47 13 the the DET ajst-12714 47 14 generalization generalization NOUN ajst-12714 47 15 ability ability NOUN ajst-12714 47 16 and and CCONJ ajst-12714 47 17 adversarial adversarial ADJ ajst-12714 47 18 performance performance NOUN ajst-12714 47 19 of of ADP ajst-12714 47 20 the the DET ajst-12714 47 21 model model NOUN ajst-12714 47 22 . . PUNCT ajst-12714 48 1 specifically specifically ADV ajst-12714 48 2 , , PUNCT ajst-12714 48 3 the the DET ajst-12714 48 4 lrn lrn PROPN ajst-12714 48 5 operation operation NOUN ajst-12714 48 6 is be AUX ajst-12714 48 7 applied apply VERB ajst-12714 48 8 after after ADP ajst-12714 48 9 the the DET ajst-12714 48 10 convolutional convolutional ADJ ajst-12714 48 11 layer layer NOUN ajst-12714 48 12 , , PUNCT ajst-12714 48 13 which which PRON ajst-12714 48 14 normalizes normalize VERB ajst-12714 48 15 the the DET ajst-12714 48 16 output output NOUN ajst-12714 48 17 of of ADP ajst-12714 48 18 each each DET ajst-12714 48 19 convolution convolution NOUN ajst-12714 48 20 kernel[10 kernel[10 PROPN ajst-12714 48 21 ] ] PUNCT ajst-12714 48 22 . . PUNCT ajst-12714 49 1 the the DET ajst-12714 49 2 purpose purpose NOUN ajst-12714 49 3 of of ADP ajst-12714 49 4 lrn lrn PROPN ajst-12714 49 5 manipulation manipulation NOUN ajst-12714 49 6 is be AUX ajst-12714 49 7 to to PART ajst-12714 49 8 enhance enhance VERB ajst-12714 49 9 local local ADJ ajst-12714 49 10 response response NOUN ajst-12714 49 11 patterns pattern NOUN ajst-12714 49 12 and and CCONJ ajst-12714 49 13 inhibit inhibit VERB ajst-12714 49 14 neurons neuron NOUN ajst-12714 49 15 with with ADP ajst-12714 49 16 larger large ADJ ajst-12714 49 17 responses response NOUN ajst-12714 49 18 , , PUNCT ajst-12714 49 19 thereby thereby ADV ajst-12714 49 20 improving improve VERB ajst-12714 49 21 the the DET ajst-12714 49 22 robustness robustness NOUN ajst-12714 49 23 of of ADP ajst-12714 49 24 the the DET ajst-12714 49 25 model model NOUN ajst-12714 49 26 . . PUNCT ajst-12714 50 1 the the DET ajst-12714 50 2 lrn lrn PROPN ajst-12714 50 3 operation operation NOUN ajst-12714 50 4 is be AUX ajst-12714 50 5 calculated calculate VERB ajst-12714 50 6 as as SCONJ ajst-12714 50 7 follows follow VERB ajst-12714 50 8 : : PUNCT ajst-12714 50 9 𝑏 𝑏 PROPN ajst-12714 50 10 𝑖 𝑖 SYM ajst-12714 50 11 , , PUNCT ajst-12714 50 12 𝑗 𝑗 PROPN ajst-12714 50 13 , , PUNCT ajst-12714 50 14 π‘˜ π‘˜ PRON ajst-12714 50 15 π‘Ž π‘Ž NOUN ajst-12714 50 16 𝑖 𝑖 SYM ajst-12714 50 17 , , PUNCT ajst-12714 50 18 𝑗 𝑗 INTJ ajst-12714 50 19 , , PUNCT ajst-12714 50 20 π‘˜ π‘˜ PROPN ajst-12714 50 21 / / SYM ajst-12714 50 22 π‘˜ π‘˜ PROPN ajst-12714 50 23 π‘Ž π‘Ž NOUN ajst-12714 50 24 βˆ— βˆ— NOUN ajst-12714 50 25 π‘Ž π‘Ž NOUN ajst-12714 50 26 𝑖 𝑖 NOUN ajst-12714 50 27 , , PUNCT ajst-12714 50 28 , , PUNCT ajst-12714 50 29 𝑗 𝑗 INTJ ajst-12714 50 30 , , PUNCT ajst-12714 50 31 , , PUNCT ajst-12714 50 32 π‘˜ π‘˜ PROPN ajst-12714 50 33 , , PUNCT ajst-12714 50 34 among among ADP ajst-12714 50 35 them they PRON ajst-12714 50 36 , , PUNCT ajst-12714 50 37 is be AUX ajst-12714 50 38 the the DET ajst-12714 50 39 𝑏 𝑏 PROPN ajst-12714 50 40 𝑖 𝑖 SYM ajst-12714 50 41 , , PUNCT ajst-12714 50 42 𝑗 𝑗 PROPN ajst-12714 50 43 , , PUNCT ajst-12714 50 44 π‘˜ π‘˜ DET ajst-12714 50 45 normalized normalize VERB ajst-12714 50 46 output output NOUN ajst-12714 50 47 , , PUNCT ajst-12714 50 48 is be AUX ajst-12714 50 49 the the DET ajst-12714 50 50 π‘Ž π‘Ž PROPN ajst-12714 50 51 𝑖 𝑖 SYM ajst-12714 50 52 , , PUNCT ajst-12714 50 53 𝑗 𝑗 INTJ ajst-12714 50 54 , , PUNCT ajst-12714 50 55 π‘˜ π‘˜ DET ajst-12714 50 56 original original ADJ ajst-12714 50 57 value value NOUN ajst-12714 50 58 of of ADP ajst-12714 50 59 the the DET ajst-12714 50 60 convolutional convolutional ADJ ajst-12714 50 61 layer layer NOUN ajst-12714 50 62 output output NOUN ajst-12714 50 63 , , PUNCT ajst-12714 50 64 is be AUX ajst-12714 50 65 a a DET ajst-12714 50 66 hyperparameter hyperparameter NOUN ajst-12714 50 67 for for ADP ajst-12714 50 68 controlling control VERB ajst-12714 50 69 the the DET ajst-12714 50 70 π‘˜ π‘˜ PROPN ajst-12714 50 71 bias bias NOUN ajst-12714 50 72 term term NOUN ajst-12714 50 73 , , PUNCT ajst-12714 50 74 is be AUX ajst-12714 50 75 a a DET ajst-12714 50 76 hyperparameter hyperparameter NOUN ajst-12714 50 77 for for ADP ajst-12714 50 78 controlling control VERB ajst-12714 50 79 the the DET ajst-12714 50 80 degree degree NOUN ajst-12714 50 81 of of ADP ajst-12714 50 82 𝛼normalization 𝛼normalization NOUN ajst-12714 50 83 , , PUNCT ajst-12714 50 84 π‘–γ€π‘—γ€π‘˜respectively π‘–γ€π‘—γ€π‘˜respectively ADV ajst-12714 50 85 represents represent VERB ajst-12714 50 86 the the DET ajst-12714 50 87 position position NOUN ajst-12714 50 88 coordinates coordinate NOUN ajst-12714 50 89 of of ADP ajst-12714 50 90 the the DET ajst-12714 50 91 feature feature NOUN ajst-12714 50 92 map[11 map[11 VERB ajst-12714 50 93 ] ] PUNCT ajst-12714 50 94 . . PUNCT ajst-12714 51 1 it it PRON ajst-12714 51 2 is be AUX ajst-12714 51 3 important important ADJ ajst-12714 51 4 to to PART ajst-12714 51 5 note note VERB ajst-12714 51 6 that that SCONJ ajst-12714 51 7 lrn lrn PROPN ajst-12714 51 8 operations operation NOUN ajst-12714 51 9 are be AUX ajst-12714 51 10 less less ADV ajst-12714 51 11 used use VERB ajst-12714 51 12 in in ADP ajst-12714 51 13 modern modern ADJ ajst-12714 51 14 deep deep ADJ ajst-12714 51 15 learning learning NOUN ajst-12714 51 16 models model NOUN ajst-12714 51 17 . . PUNCT ajst-12714 52 1 compared compare VERB ajst-12714 52 2 to to ADP ajst-12714 52 3 lrn lrn PROPN ajst-12714 52 4 , , PUNCT ajst-12714 52 5 batch batch NOUN ajst-12714 52 6 normalization normalization NOUN ajst-12714 52 7 has have AUX ajst-12714 52 8 become become VERB ajst-12714 52 9 a a DET ajst-12714 52 10 more more ADV ajst-12714 52 11 common common ADJ ajst-12714 52 12 and and CCONJ ajst-12714 52 13 effective effective ADJ ajst-12714 52 14 normalization normalization NOUN ajst-12714 52 15 method method NOUN ajst-12714 52 16 . . PUNCT ajst-12714 53 1 in in ADP ajst-12714 53 2 subsequent subsequent ADJ ajst-12714 53 3 network network NOUN ajst-12714 53 4 architectures architecture NOUN ajst-12714 53 5 , , PUNCT ajst-12714 53 6 such such ADJ ajst-12714 53 7 as as ADP ajst-12714 53 8 resnet resnet NOUN ajst-12714 53 9 and and CCONJ ajst-12714 53 10 inception inception NOUN ajst-12714 53 11 , , PUNCT ajst-12714 53 12 batch batch NOUN ajst-12714 53 13 normalization normalization NOUN ajst-12714 53 14 is be AUX ajst-12714 53 15 often often ADV ajst-12714 53 16 used use VERB ajst-12714 53 17 for for ADP ajst-12714 53 18 normalization normalization NOUN ajst-12714 53 19 . . PUNCT ajst-12714 54 1 2.3.4 2.3.4 X ajst-12714 54 2 . . PUNCT ajst-12714 54 3 fully fully ADV ajst-12714 54 4 connected connect VERB ajst-12714 54 5 layer layer NOUN ajst-12714 54 6 the the DET ajst-12714 54 7 last last ADJ ajst-12714 54 8 three three NUM ajst-12714 54 9 layers layer NOUN ajst-12714 54 10 of of ADP ajst-12714 54 11 the the DET ajst-12714 54 12 vgg16 vgg16 NOUN ajst-12714 54 13 model model NOUN ajst-12714 54 14 are be AUX ajst-12714 54 15 fully fully ADV ajst-12714 54 16 connected connected ADJ ajst-12714 54 17 layers layer NOUN ajst-12714 54 18 , , PUNCT ajst-12714 54 19 which which PRON ajst-12714 54 20 are be AUX ajst-12714 54 21 used use VERB ajst-12714 54 22 to to PART ajst-12714 54 23 transform transform VERB ajst-12714 54 24 the the DET ajst-12714 54 25 feature feature NOUN ajst-12714 54 26 map map NOUN ajst-12714 54 27 extracted extract VERB ajst-12714 54 28 by by ADP ajst-12714 54 29 the the DET ajst-12714 54 30 convolutional convolutional ADJ ajst-12714 54 31 layer layer NOUN ajst-12714 54 32 into into ADP ajst-12714 54 33 the the DET ajst-12714 54 34 final final ADJ ajst-12714 54 35 classification classification NOUN ajst-12714 54 36 result result NOUN ajst-12714 54 37 . . PUNCT ajst-12714 55 1 in in ADP ajst-12714 55 2 vgg16 vgg16 PROPN ajst-12714 55 3 , , PUNCT ajst-12714 55 4 the the DET ajst-12714 55 5 fully fully ADV ajst-12714 55 6 connected connected ADJ ajst-12714 55 7 layer layer NOUN ajst-12714 55 8 is be AUX ajst-12714 55 9 composed compose VERB ajst-12714 55 10 of of ADP ajst-12714 55 11 two two NUM ajst-12714 55 12 hidden hidden ADJ ajst-12714 55 13 layers layer NOUN ajst-12714 55 14 with with ADP ajst-12714 55 15 4096 4096 NUM ajst-12714 55 16 neurons neuron NOUN ajst-12714 55 17 and and CCONJ ajst-12714 55 18 an an DET ajst-12714 55 19 output output NOUN ajst-12714 55 20 layer layer NOUN ajst-12714 55 21 with with ADP ajst-12714 55 22 1000 1000 NUM ajst-12714 55 23 neurons[12 neurons[12 NOUN ajst-12714 55 24 ] ] PUNCT ajst-12714 55 25 . . PUNCT ajst-12714 56 1 these these DET ajst-12714 56 2 fully fully ADV ajst-12714 56 3 connected connected ADJ ajst-12714 56 4 layers layer NOUN ajst-12714 56 5 receive receive VERB ajst-12714 56 6 flattened flatten VERB ajst-12714 56 7 features feature NOUN ajst-12714 56 8 from from ADP ajst-12714 56 9 the the DET ajst-12714 56 10 convolutional convolutional ADJ ajst-12714 56 11 layer layer NOUN ajst-12714 56 12 as as ADP ajst-12714 56 13 input input NOUN ajst-12714 56 14 , , PUNCT ajst-12714 56 15 perform perform VERB ajst-12714 56 16 linear linear ADJ ajst-12714 56 17 transformations transformation NOUN ajst-12714 56 18 and and CCONJ ajst-12714 56 19 nonlinear nonlinear ADJ ajst-12714 56 20 activation activation NOUN ajst-12714 56 21 , , PUNCT ajst-12714 56 22 and and CCONJ ajst-12714 56 23 finally finally ADV ajst-12714 56 24 output output VERB ajst-12714 56 25 a a DET ajst-12714 56 26 1000 1000 NUM ajst-12714 56 27 - - PUNCT ajst-12714 56 28 dimensional dimensional ADJ ajst-12714 56 29 vector vector NOUN ajst-12714 56 30 representing represent VERB ajst-12714 56 31 probability probability NOUN ajst-12714 56 32 distributions distribution NOUN ajst-12714 56 33 of of ADP ajst-12714 56 34 different different ADJ ajst-12714 56 35 classes class NOUN ajst-12714 56 36 . . PUNCT ajst-12714 57 1 the the DET ajst-12714 57 2 function function NOUN ajst-12714 57 3 of of ADP ajst-12714 57 4 the the DET ajst-12714 57 5 fully fully ADV ajst-12714 57 6 connected connect VERB ajst-12714 57 7 layer layer NOUN ajst-12714 57 8 is be AUX ajst-12714 57 9 to to PART ajst-12714 57 10 map map VERB ajst-12714 57 11 the the DET ajst-12714 57 12 high high ADJ ajst-12714 57 13 - - PUNCT ajst-12714 57 14 level level NOUN ajst-12714 57 15 feature feature NOUN ajst-12714 57 16 map map NOUN ajst-12714 57 17 extracted extract VERB ajst-12714 57 18 by by ADP ajst-12714 57 19 the the DET ajst-12714 57 20 convolutional convolutional ADJ ajst-12714 57 21 layer layer NOUN ajst-12714 57 22 onto onto ADP ajst-12714 57 23 the the DET ajst-12714 57 24 classification classification NOUN ajst-12714 57 25 label label NOUN ajst-12714 57 26 . . PUNCT ajst-12714 58 1 through through ADP ajst-12714 58 2 the the DET ajst-12714 58 3 nonlinear nonlinear ADJ ajst-12714 58 4 transformation transformation NOUN ajst-12714 58 5 of of ADP ajst-12714 58 6 multiple multiple ADJ ajst-12714 58 7 hidden hide VERB ajst-12714 58 8 layers layer NOUN ajst-12714 58 9 , , PUNCT ajst-12714 58 10 the the DET ajst-12714 58 11 fully fully ADV ajst-12714 58 12 connected connect VERB ajst-12714 58 13 layer layer NOUN ajst-12714 58 14 can can AUX ajst-12714 58 15 learn learn VERB ajst-12714 58 16 more more ADV ajst-12714 58 17 abstract abstract ADJ ajst-12714 58 18 and and CCONJ ajst-12714 58 19 complex complex ADJ ajst-12714 58 20 feature feature NOUN ajst-12714 58 21 representations representation NOUN ajst-12714 58 22 , , PUNCT ajst-12714 58 23 thereby thereby ADV ajst-12714 58 24 improving improve VERB ajst-12714 58 25 the the DET ajst-12714 58 26 classification classification NOUN ajst-12714 58 27 accuracy[13 accuracy[13 NOUN ajst-12714 58 28 ] ] PUNCT ajst-12714 58 29 . . PUNCT ajst-12714 59 1 it it PRON ajst-12714 59 2 should should AUX ajst-12714 59 3 be be AUX ajst-12714 59 4 noted note VERB ajst-12714 59 5 that that SCONJ ajst-12714 59 6 the the DET ajst-12714 59 7 final final ADJ ajst-12714 59 8 fully fully ADV ajst-12714 59 9 connected connected ADJ ajst-12714 59 10 layer layer NOUN ajst-12714 59 11 of of ADP ajst-12714 59 12 the the DET ajst-12714 59 13 vgg16 vgg16 NOUN ajst-12714 59 14 model model NOUN ajst-12714 59 15 outputs output VERB ajst-12714 59 16 the the DET ajst-12714 59 17 probability probability NOUN ajst-12714 59 18 of of ADP ajst-12714 59 19 1000 1000 NUM ajst-12714 59 20 classes class NOUN ajst-12714 59 21 , , PUNCT ajst-12714 59 22 which which PRON ajst-12714 59 23 is be AUX ajst-12714 59 24 suitable suitable ADJ ajst-12714 59 25 for for ADP ajst-12714 59 26 classification classification NOUN ajst-12714 59 27 tasks task NOUN ajst-12714 59 28 on on ADP ajst-12714 59 29 the the DET ajst-12714 59 30 imagenet imagenet NOUN ajst-12714 59 31 dataset dataset NOUN ajst-12714 59 32 . . PUNCT ajst-12714 60 1 if if SCONJ ajst-12714 60 2 you you PRON ajst-12714 60 3 want want VERB ajst-12714 60 4 to to PART ajst-12714 60 5 classify classify VERB ajst-12714 60 6 on on ADP ajst-12714 60 7 other other ADJ ajst-12714 60 8 datasets dataset NOUN ajst-12714 60 9 , , PUNCT ajst-12714 60 10 you you PRON ajst-12714 60 11 need need VERB ajst-12714 60 12 to to PART ajst-12714 60 13 adjust adjust VERB ajst-12714 60 14 accordingly accordingly ADV ajst-12714 60 15 . . PUNCT ajst-12714 61 1 the the DET ajst-12714 61 2 obtained obtain VERB ajst-12714 61 3 results result NOUN ajst-12714 61 4 are be AUX ajst-12714 61 5 dimensionally dimensionally ADV ajst-12714 61 6 predicted predict VERB ajst-12714 61 7 using use VERB ajst-12714 61 8 thesoftmax thesoftmax NOUN ajst-12714 61 9 prediction prediction NOUN ajst-12714 61 10 layer layer NOUN ajst-12714 61 11 . . PUNCT ajst-12714 62 1 3 3 X ajst-12714 62 2 . . X ajst-12714 62 3 attention attention NOUN ajst-12714 62 4 mechanism mechanism NOUN ajst-12714 62 5 the the DET ajst-12714 62 6 attention attention NOUN ajst-12714 62 7 mechanism mechanism NOUN ajst-12714 62 8 refers refer VERB ajst-12714 62 9 to to ADP ajst-12714 62 10 reading read VERB ajst-12714 62 11 data datum NOUN ajst-12714 62 12 , , PUNCT ajst-12714 62 13 focusing focus VERB ajst-12714 62 14 only only ADV ajst-12714 62 15 on on ADP ajst-12714 62 16 important important ADJ ajst-12714 62 17 data datum NOUN ajst-12714 62 18 and and CCONJ ajst-12714 62 19 ignoring ignore VERB ajst-12714 62 20 unimportant unimportant ADJ ajst-12714 62 21 data datum NOUN ajst-12714 62 22 . . PUNCT ajst-12714 63 1 in in ADP ajst-12714 63 2 the the DET ajst-12714 63 3 field field NOUN ajst-12714 63 4 of of ADP ajst-12714 63 5 image image NOUN ajst-12714 63 6 recognition recognition NOUN ajst-12714 63 7 , , PUNCT ajst-12714 63 8 if if SCONJ ajst-12714 63 9 the the DET ajst-12714 63 10 key key ADJ ajst-12714 63 11 data data NOUN ajst-12714 63 12 is be AUX ajst-12714 63 13 confused confuse VERB ajst-12714 63 14 with with ADP ajst-12714 63 15 the the DET ajst-12714 63 16 non non ADJ ajst-12714 63 17 - - ADJ ajst-12714 63 18 key key ADJ ajst-12714 63 19 data datum NOUN ajst-12714 63 20 in in ADP ajst-12714 63 21 the the DET ajst-12714 63 22 process process NOUN ajst-12714 63 23 of of ADP ajst-12714 63 24 image image NOUN ajst-12714 63 25 feature feature NOUN ajst-12714 63 26 extraction extraction NOUN ajst-12714 63 27 , , PUNCT ajst-12714 63 28 it it PRON ajst-12714 63 29 will will AUX ajst-12714 63 30 increase increase VERB ajst-12714 63 31 the the DET ajst-12714 63 32 workload workload NOUN ajst-12714 63 33 , , PUNCT ajst-12714 63 34 lead lead VERB ajst-12714 63 35 to to ADP ajst-12714 63 36 waste waste NOUN ajst-12714 63 37 of of ADP ajst-12714 63 38 resources resource NOUN ajst-12714 63 39 , , PUNCT ajst-12714 63 40 and and CCONJ ajst-12714 63 41 also also ADV ajst-12714 63 42 affect affect VERB ajst-12714 63 43 the the DET ajst-12714 63 44 experimental experimental ADJ ajst-12714 63 45 effect effect NOUN ajst-12714 63 46 of of ADP ajst-12714 63 47 the the DET ajst-12714 63 48 model model NOUN ajst-12714 63 49 . . PUNCT ajst-12714 64 1 therefore therefore ADV ajst-12714 64 2 , , PUNCT ajst-12714 64 3 it it PRON ajst-12714 64 4 was be AUX ajst-12714 64 5 chosen choose VERB ajst-12714 64 6 to to PART ajst-12714 64 7 improve improve VERB ajst-12714 64 8 the the DET ajst-12714 64 9 neural neural ADJ ajst-12714 64 10 network network NOUN ajst-12714 64 11 by by ADP ajst-12714 64 12 introducing introduce VERB ajst-12714 64 13 a a DET ajst-12714 64 14 channel channel NOUN ajst-12714 64 15 attention attention NOUN ajst-12714 64 16 mechanism mechanism NOUN ajst-12714 64 17 to to PART ajst-12714 64 18 guide guide VERB ajst-12714 64 19 the the DET ajst-12714 64 20 optimization optimization NOUN ajst-12714 64 21 settings setting NOUN ajst-12714 64 22 of of ADP ajst-12714 64 23 the the DET ajst-12714 64 24 network[14 network[14 PROPN ajst-12714 64 25 ] ] PUNCT ajst-12714 64 26 . . PUNCT ajst-12714 65 1 aiming aim VERB ajst-12714 65 2 at at ADP ajst-12714 65 3 the the DET ajst-12714 65 4 goal goal NOUN ajst-12714 65 5 of of ADP ajst-12714 65 6 attention attention NOUN ajst-12714 65 7 weight weight NOUN ajst-12714 65 8 attention attention NOUN ajst-12714 65 9 , , PUNCT ajst-12714 65 10 this this DET ajst-12714 65 11 paper paper NOUN ajst-12714 65 12 uses use VERB ajst-12714 65 13 four four NUM ajst-12714 65 14 attention attention NOUN ajst-12714 65 15 mechanisms mechanism NOUN ajst-12714 65 16 37 37 NUM ajst-12714 65 17 and and CCONJ ajst-12714 65 18 integrates integrate VERB ajst-12714 65 19 them they PRON ajst-12714 65 20 into into ADP ajst-12714 65 21 the the DET ajst-12714 65 22 vgg16 vgg16 NOUN ajst-12714 65 23 neural neural ADJ ajst-12714 65 24 network network NOUN ajst-12714 65 25 , , PUNCT ajst-12714 65 26 and and CCONJ ajst-12714 65 27 finally finally ADV ajst-12714 65 28 selects select VERB ajst-12714 65 29 the the DET ajst-12714 65 30 attention attention NOUN ajst-12714 65 31 mechanism mechanism NOUN ajst-12714 65 32 with with ADP ajst-12714 65 33 the the DET ajst-12714 65 34 best good ADJ ajst-12714 65 35 prediction prediction NOUN ajst-12714 65 36 effect effect NOUN ajst-12714 65 37 of of ADP ajst-12714 65 38 pipe pipe NOUN ajst-12714 65 39 defect defect NOUN ajst-12714 65 40 size size NOUN ajst-12714 65 41 as as ADP ajst-12714 65 42 the the DET ajst-12714 65 43 final final ADJ ajst-12714 65 44 neural neural ADJ ajst-12714 65 45 network network NOUN ajst-12714 65 46 mechanism mechanism NOUN ajst-12714 65 47 through through ADP ajst-12714 65 48 ablation ablation NOUN ajst-12714 65 49 experiment experiment NOUN ajst-12714 65 50 comparison comparison NOUN ajst-12714 65 51 . . PUNCT ajst-12714 66 1 3.1 3.1 NUM ajst-12714 66 2 . . PUNCT ajst-12714 66 3 se se PROPN ajst-12714 66 4 attention attention NOUN ajst-12714 66 5 mechanism mechanism NOUN ajst-12714 66 6 the the DET ajst-12714 66 7 se se PROPN ajst-12714 66 8 attention attention NOUN ajst-12714 66 9 module module NOUN ajst-12714 66 10 is be AUX ajst-12714 66 11 a a DET ajst-12714 66 12 commonly commonly ADV ajst-12714 66 13 used use VERB ajst-12714 66 14 channel channel NOUN ajst-12714 66 15 attention attention NOUN ajst-12714 66 16 mechanism mechanism NOUN ajst-12714 66 17 for for ADP ajst-12714 66 18 weighting weight VERB ajst-12714 66 19 input input NOUN ajst-12714 66 20 in in ADP ajst-12714 66 21 natural natural ADJ ajst-12714 66 22 language language NOUN ajst-12714 66 23 processing processing NOUN ajst-12714 66 24 tasks task NOUN ajst-12714 66 25 . . PUNCT ajst-12714 67 1 se se PROPN ajst-12714 67 2 stands stand VERB ajst-12714 67 3 for for ADP ajst-12714 67 4 " " PUNCT ajst-12714 67 5 selective selective ADJ ajst-12714 67 6 interpolation interpolation NOUN ajst-12714 67 7 " " PUNCT ajst-12714 67 8 and and CCONJ ajst-12714 67 9 this this DET ajst-12714 67 10 module module NOUN ajst-12714 67 11 does do VERB ajst-12714 67 12 this this PRON ajst-12714 67 13 by by ADP ajst-12714 67 14 learning learn VERB ajst-12714 67 15 two two NUM ajst-12714 67 16 different different ADJ ajst-12714 67 17 distribution distribution NOUN ajst-12714 67 18 of of ADP ajst-12714 67 19 attention attention NOUN ajst-12714 67 20 weights weight NOUN ajst-12714 67 21 . . PUNCT ajst-12714 68 1 specifically specifically ADV ajst-12714 68 2 , , PUNCT ajst-12714 68 3 the the DET ajst-12714 68 4 se se PROPN ajst-12714 68 5 attention attention NOUN ajst-12714 68 6 module module NOUN ajst-12714 68 7 performs perform VERB ajst-12714 68 8 two two NUM ajst-12714 68 9 linear linear ADJ ajst-12714 68 10 transformations transformation NOUN ajst-12714 68 11 of of ADP ajst-12714 68 12 the the DET ajst-12714 68 13 input input NOUN ajst-12714 68 14 features feature NOUN ajst-12714 68 15 and and CCONJ ajst-12714 68 16 then then ADV ajst-12714 68 17 activates activate VERB ajst-12714 68 18 them they PRON ajst-12714 68 19 separately separately ADV ajst-12714 68 20 through through ADP ajst-12714 68 21 activation activation NOUN ajst-12714 68 22 functions[15 functions[15 NOUN ajst-12714 68 23 ] ] PUNCT ajst-12714 68 24 . . PUNCT ajst-12714 69 1 next next ADV ajst-12714 69 2 , , PUNCT ajst-12714 69 3 the the DET ajst-12714 69 4 softmax softmax NOUN ajst-12714 69 5 function function NOUN ajst-12714 69 6 is be AUX ajst-12714 69 7 used use VERB ajst-12714 69 8 to to PART ajst-12714 69 9 convert convert VERB ajst-12714 69 10 the the DET ajst-12714 69 11 two two NUM ajst-12714 69 12 transformed transform VERB ajst-12714 69 13 vectors vector NOUN ajst-12714 69 14 into into ADP ajst-12714 69 15 attention attention NOUN ajst-12714 69 16 weight weight NOUN ajst-12714 69 17 distributions distribution NOUN ajst-12714 69 18 . . PUNCT ajst-12714 70 1 finally finally ADV ajst-12714 70 2 , , PUNCT ajst-12714 70 3 the the DET ajst-12714 70 4 input input NOUN ajst-12714 70 5 features feature NOUN ajst-12714 70 6 are be AUX ajst-12714 70 7 multiplied multiply VERB ajst-12714 70 8 and and CCONJ ajst-12714 70 9 added add VERB ajst-12714 70 10 to to ADP ajst-12714 70 11 the the DET ajst-12714 70 12 attention attention NOUN ajst-12714 70 13 weights weight NOUN ajst-12714 70 14 to to PART ajst-12714 70 15 obtain obtain VERB ajst-12714 70 16 the the DET ajst-12714 70 17 weighted weight VERB ajst-12714 70 18 feature feature NOUN ajst-12714 70 19 representation representation NOUN ajst-12714 70 20 . . PUNCT ajst-12714 71 1 the the DET ajst-12714 71 2 se se PROPN ajst-12714 71 3 attention attention NOUN ajst-12714 71 4 module module NOUN ajst-12714 71 5 adaptively adaptively ADV ajst-12714 71 6 selects select VERB ajst-12714 71 7 different different ADJ ajst-12714 71 8 features feature NOUN ajst-12714 71 9 for for ADP ajst-12714 71 10 weighting weighting NOUN ajst-12714 71 11 , , PUNCT ajst-12714 71 12 thereby thereby ADV ajst-12714 71 13 improving improve VERB ajst-12714 71 14 the the DET ajst-12714 71 15 expressive expressive ADJ ajst-12714 71 16 ability ability NOUN ajst-12714 71 17 of of ADP ajst-12714 71 18 the the DET ajst-12714 71 19 model model NOUN ajst-12714 71 20 when when SCONJ ajst-12714 71 21 processing process VERB ajst-12714 71 22 the the DET ajst-12714 71 23 input input NOUN ajst-12714 71 24 sequence sequence NOUN ajst-12714 71 25 . . PUNCT ajst-12714 72 1 the the DET ajst-12714 72 2 advantage advantage NOUN ajst-12714 72 3 of of ADP ajst-12714 72 4 the the DET ajst-12714 72 5 s s PROPN ajst-12714 72 6 e e NOUN ajst-12714 72 7 attention attention NOUN ajst-12714 72 8 module module NOUN ajst-12714 72 9 is be AUX ajst-12714 72 10 that that SCONJ ajst-12714 72 11 it it PRON ajst-12714 72 12 does do AUX ajst-12714 72 13 not not PART ajst-12714 72 14 change change VERB ajst-12714 72 15 the the DET ajst-12714 72 16 dimensions dimension NOUN ajst-12714 72 17 of of ADP ajst-12714 72 18 input input NOUN ajst-12714 72 19 and and CCONJ ajst-12714 72 20 output output NOUN ajst-12714 72 21 , , PUNCT ajst-12714 72 22 and and CCONJ ajst-12714 72 23 can can AUX ajst-12714 72 24 be be AUX ajst-12714 72 25 easily easily ADV ajst-12714 72 26 combined combine VERB ajst-12714 72 27 with with ADP ajst-12714 72 28 other other ADJ ajst-12714 72 29 networks network NOUN ajst-12714 72 30 . . PUNCT ajst-12714 73 1 the the DET ajst-12714 73 2 algorithm algorithm NOUN ajst-12714 73 3 flow flow NOUN ajst-12714 73 4 diagram diagram NOUN ajst-12714 73 5 of of ADP ajst-12714 73 6 the the DET ajst-12714 73 7 s s PROPN ajst-12714 73 8 e e NOUN ajst-12714 73 9 module module NOUN ajst-12714 73 10 is be AUX ajst-12714 73 11 shown show VERB ajst-12714 73 12 in in ADP ajst-12714 73 13 figure figure NOUN ajst-12714 73 14 7 7 NUM ajst-12714 73 15 x x SYM ajst-12714 73 16 h h NOUN ajst-12714 74 1 ο‚’ ο‚’ ADV ajst-12714 75 1 w w PROPN ajst-12714 75 2 ο‚’ ο‚’ NUM ajst-12714 75 3 cο‚’ cο‚’ PROPN ajst-12714 75 4 trf trf PROPN ajst-12714 75 5 u u PROPN ajst-12714 75 6 h h NOUN ajst-12714 75 7 w w PROPN ajst-12714 75 8 c c PROPN ajst-12714 75 9 ( ( PUNCT ajst-12714 75 10 * * NOUN ajst-12714 75 11 ) ) PUNCT ajst-12714 75 12 sqf sqf NOUN ajst-12714 75 13 ( ( PUNCT ajst-12714 75 14 * * NOUN ajst-12714 75 15 , , PUNCT ajst-12714 75 16 ) ) PUNCT ajst-12714 76 1 ex ex PRON ajst-12714 76 2 wf wf PROPN ajst-12714 76 3 ( ( PUNCT ajst-12714 76 4 * * NOUN ajst-12714 76 5 , , PUNCT ajst-12714 76 6 * * NOUN ajst-12714 76 7 ) ) PUNCT ajst-12714 76 8 scalef scalef NOUN ajst-12714 76 9 x x X ajst-12714 76 10 οΏ½ οΏ½ PROPN ajst-12714 76 11 h h NOUN ajst-12714 76 12 w w PROPN ajst-12714 76 13 c c PROPN ajst-12714 76 14 figure figure NOUN ajst-12714 76 15 7 7 NUM ajst-12714 76 16 . . PUNCT ajst-12714 76 17 flow flow NOUN ajst-12714 76 18 chart chart NOUN ajst-12714 76 19 of of ADP ajst-12714 76 20 se se X ajst-12714 76 21 attention attention NOUN ajst-12714 76 22 mechanism mechanism NOUN ajst-12714 76 23 3.2 3.2 NUM ajst-12714 76 24 . . PUNCT ajst-12714 77 1 stn stn PROPN ajst-12714 77 2 attention attention NOUN ajst-12714 77 3 mechanism mechanism NOUN ajst-12714 77 4 stn stn PROPN ajst-12714 77 5 ( ( PUNCT ajst-12714 77 6 spatial spatial ADJ ajst-12714 77 7 transformer transformer NOUN ajst-12714 77 8 network network NOUN ajst-12714 77 9 ) ) PUNCT ajst-12714 77 10 is be AUX ajst-12714 77 11 a a DET ajst-12714 77 12 spatial spatial ADJ ajst-12714 77 13 attention attention NOUN ajst-12714 77 14 module module NOUN ajst-12714 77 15 used use VERB ajst-12714 77 16 to to PART ajst-12714 77 17 learn learn VERB ajst-12714 77 18 to to PART ajst-12714 77 19 perform perform VERB ajst-12714 77 20 geometric geometric ADJ ajst-12714 77 21 transformations transformation NOUN ajst-12714 77 22 on on ADP ajst-12714 77 23 input input NOUN ajst-12714 77 24 data datum NOUN ajst-12714 77 25 in in ADP ajst-12714 77 26 neural neural ADJ ajst-12714 77 27 networks network NOUN ajst-12714 77 28 . . PUNCT ajst-12714 78 1 it it PRON ajst-12714 78 2 can can AUX ajst-12714 78 3 rotate rotate VERB ajst-12714 78 4 , , PUNCT ajst-12714 78 5 translate translate ADJ ajst-12714 78 6 , , PUNCT ajst-12714 78 7 scale scale NOUN ajst-12714 78 8 and and CCONJ ajst-12714 78 9 other other ADJ ajst-12714 78 10 operations operation NOUN ajst-12714 78 11 on on ADP ajst-12714 78 12 the the DET ajst-12714 78 13 input input NOUN ajst-12714 78 14 data datum NOUN ajst-12714 78 15 through through ADP ajst-12714 78 16 the the DET ajst-12714 78 17 learned learn VERB ajst-12714 78 18 transformation transformation NOUN ajst-12714 78 19 parameters parameter NOUN ajst-12714 78 20 , , PUNCT ajst-12714 78 21 so so SCONJ ajst-12714 78 22 that that SCONJ ajst-12714 78 23 the the DET ajst-12714 78 24 network network NOUN ajst-12714 78 25 can can AUX ajst-12714 78 26 better well ADV ajst-12714 78 27 adapt adapt VERB ajst-12714 78 28 to to ADP ajst-12714 78 29 the the DET ajst-12714 78 30 input input NOUN ajst-12714 78 31 of of ADP ajst-12714 78 32 different different ADJ ajst-12714 78 33 scales scale NOUN ajst-12714 78 34 , , PUNCT ajst-12714 78 35 angles angle NOUN ajst-12714 78 36 , , PUNCT ajst-12714 78 37 and and CCONJ ajst-12714 78 38 positions position NOUN ajst-12714 78 39 . . PUNCT ajst-12714 79 1 stn stn PROPN ajst-12714 79 2 modules module NOUN ajst-12714 79 3 can can AUX ajst-12714 79 4 be be AUX ajst-12714 79 5 embedded embed VERB ajst-12714 79 6 anywhere anywhere ADV ajst-12714 79 7 in in ADP ajst-12714 79 8 a a DET ajst-12714 79 9 deep deep ADJ ajst-12714 79 10 learning learning NOUN ajst-12714 79 11 network network NOUN ajst-12714 79 12 and and CCONJ ajst-12714 79 13 are be AUX ajst-12714 79 14 often often ADV ajst-12714 79 15 used use VERB ajst-12714 79 16 to to PART ajst-12714 79 17 improve improve VERB ajst-12714 79 18 the the DET ajst-12714 79 19 transformation transformation NOUN ajst-12714 79 20 invariance invariance NOUN ajst-12714 79 21 of of ADP ajst-12714 79 22 the the DET ajst-12714 79 23 network network NOUN ajst-12714 79 24 's 's PART ajst-12714 79 25 input input NOUN ajst-12714 79 26 data datum NOUN ajst-12714 79 27 . . PUNCT ajst-12714 80 1 it it PRON ajst-12714 80 2 consists consist VERB ajst-12714 80 3 of of ADP ajst-12714 80 4 three three NUM ajst-12714 80 5 main main ADJ ajst-12714 80 6 components component NOUN ajst-12714 80 7 : : PUNCT ajst-12714 80 8 a a DET ajst-12714 80 9 localization localization NOUN ajst-12714 80 10 network network NOUN ajst-12714 80 11 , , PUNCT ajst-12714 80 12 a a DET ajst-12714 80 13 grid grid NOUN ajst-12714 80 14 generator generator NOUN ajst-12714 80 15 , , PUNCT ajst-12714 80 16 and and CCONJ ajst-12714 80 17 a a DET ajst-12714 80 18 sampler sampler NOUN ajst-12714 80 19 ) ) PUNCT ajst-12714 81 1 [ [ X ajst-12714 81 2 16 16 NUM ajst-12714 81 3 ] ] PUNCT ajst-12714 81 4 . . PUNCT ajst-12714 82 1 the the DET ajst-12714 82 2 localized localize VERB ajst-12714 82 3 coordinate coordinate NOUN ajst-12714 82 4 regression regression NOUN ajst-12714 82 5 network network NOUN ajst-12714 82 6 is be AUX ajst-12714 82 7 responsible responsible ADJ ajst-12714 82 8 for for ADP ajst-12714 82 9 learning learn VERB ajst-12714 82 10 from from ADP ajst-12714 82 11 the the DET ajst-12714 82 12 input input NOUN ajst-12714 82 13 data data VERB ajst-12714 82 14 the the DET ajst-12714 82 15 transformation transformation NOUN ajst-12714 82 16 parameters parameter NOUN ajst-12714 82 17 , , PUNCT ajst-12714 82 18 which which PRON ajst-12714 82 19 describe describe VERB ajst-12714 82 20 the the DET ajst-12714 82 21 geometric geometric ADJ ajst-12714 82 22 transformations transformation NOUN ajst-12714 82 23 required require VERB ajst-12714 82 24 to to PART ajst-12714 82 25 map map VERB ajst-12714 82 26 the the DET ajst-12714 82 27 input input NOUN ajst-12714 82 28 data datum NOUN ajst-12714 82 29 to to ADP ajst-12714 82 30 the the DET ajst-12714 82 31 output output NOUN ajst-12714 82 32 space space NOUN ajst-12714 82 33 . . PUNCT ajst-12714 83 1 the the DET ajst-12714 83 2 grid grid NOUN ajst-12714 83 3 builder builder NOUN ajst-12714 83 4 uses use VERB ajst-12714 83 5 these these DET ajst-12714 83 6 parameters parameter NOUN ajst-12714 83 7 to to PART ajst-12714 83 8 generate generate VERB ajst-12714 83 9 a a DET ajst-12714 83 10 sampling sample VERB ajst-12714 83 11 grid grid NOUN ajst-12714 83 12 that that PRON ajst-12714 83 13 defines define VERB ajst-12714 83 14 how how SCONJ ajst-12714 83 15 the the DET ajst-12714 83 16 input input NOUN ajst-12714 83 17 data data NOUN ajst-12714 83 18 is be AUX ajst-12714 83 19 sampled sample VERB ajst-12714 83 20 . . PUNCT ajst-12714 84 1 finally finally ADV ajst-12714 84 2 , , PUNCT ajst-12714 84 3 the the DET ajst-12714 84 4 sampler sampler NOUN ajst-12714 84 5 samples sample VERB ajst-12714 84 6 the the DET ajst-12714 84 7 input input NOUN ajst-12714 84 8 data datum NOUN ajst-12714 84 9 using use VERB ajst-12714 84 10 a a DET ajst-12714 84 11 sampling sample VERB ajst-12714 84 12 mesh mesh NOUN ajst-12714 84 13 to to PART ajst-12714 84 14 produce produce VERB ajst-12714 84 15 a a DET ajst-12714 84 16 geometrically geometrically ADV ajst-12714 84 17 transformed transform VERB ajst-12714 84 18 output[17 output[17 PROPN ajst-12714 84 19 ] ] PUNCT ajst-12714 84 20 . . PUNCT ajst-12714 85 1 by by ADP ajst-12714 85 2 introducing introduce VERB ajst-12714 85 3 the the DET ajst-12714 85 4 stn stn PROPN ajst-12714 85 5 module module NOUN ajst-12714 85 6 , , PUNCT ajst-12714 85 7 the the DET ajst-12714 85 8 neural neural ADJ ajst-12714 85 9 network network NOUN ajst-12714 85 10 can can AUX ajst-12714 85 11 automatically automatically ADV ajst-12714 85 12 learn learn VERB ajst-12714 85 13 how how SCONJ ajst-12714 85 14 to to PART ajst-12714 85 15 perform perform VERB ajst-12714 85 16 geometric geometric ADJ ajst-12714 85 17 transformations transformation NOUN ajst-12714 85 18 on on ADP ajst-12714 85 19 the the DET ajst-12714 85 20 input input NOUN ajst-12714 85 21 data datum NOUN ajst-12714 85 22 , , PUNCT ajst-12714 85 23 thereby thereby ADV ajst-12714 85 24 improving improve VERB ajst-12714 85 25 the the DET ajst-12714 85 26 robustness robustness NOUN ajst-12714 85 27 and and CCONJ ajst-12714 85 28 generalization generalization NOUN ajst-12714 85 29 ability ability NOUN ajst-12714 85 30 of of ADP ajst-12714 85 31 the the DET ajst-12714 85 32 network network NOUN ajst-12714 85 33 to to ADP ajst-12714 85 34 the the DET ajst-12714 85 35 input input NOUN ajst-12714 85 36 data datum NOUN ajst-12714 85 37 . . PUNCT ajst-12714 86 1 stn stn PROPN ajst-12714 86 2 modules module NOUN ajst-12714 86 3 are be AUX ajst-12714 86 4 widely widely ADV ajst-12714 86 5 used use VERB ajst-12714 86 6 in in ADP ajst-12714 86 7 many many ADJ ajst-12714 86 8 computer computer NOUN ajst-12714 86 9 vision vision NOUN ajst-12714 86 10 tasks task NOUN ajst-12714 86 11 , , PUNCT ajst-12714 86 12 such such ADJ ajst-12714 86 13 as as ADP ajst-12714 86 14 image image NOUN ajst-12714 86 15 classification classification NOUN ajst-12714 86 16 , , PUNCT ajst-12714 86 17 object object NOUN ajst-12714 86 18 detection detection NOUN ajst-12714 86 19 , , PUNCT ajst-12714 86 20 and and CCONJ ajst-12714 86 21 image image NOUN ajst-12714 86 22 generation generation NOUN ajst-12714 86 23 . . PUNCT ajst-12714 87 1 the the DET ajst-12714 87 2 flowchart flowchart NOUN ajst-12714 87 3 of of ADP ajst-12714 87 4 the the DET ajst-12714 87 5 attention attention NOUN ajst-12714 87 6 mechanism mechanism NOUN ajst-12714 87 7 of of ADP ajst-12714 87 8 tn tn NOUN ajst-12714 87 9 is be AUX ajst-12714 87 10 shown show VERB ajst-12714 87 11 in in ADP ajst-12714 87 12 figure figure NOUN ajst-12714 87 13 8 8 NUM ajst-12714 87 14   NOUN ajst-12714 87 15 ( ( PUNCT ajst-12714 87 16 ) ) PUNCT ajst-12714 87 17 gt gt PROPN ajst-12714 87 18   PROPN ajst-12714 87 19 u u PROPN ajst-12714 87 20 v v ADP ajst-12714 87 21 local local ADJ ajst-12714 87 22 Β  Β  SPACE ajst-12714 87 23 network network NOUN ajst-12714 87 24 mesh mesh NOUN ajst-12714 87 25 Β  Β  SPACE ajst-12714 87 26 generator generator NOUN ajst-12714 87 27 sampler sampler ADJ ajst-12714 87 28 figure figure NOUN ajst-12714 87 29 8 8 NUM ajst-12714 87 30 . . PUNCT ajst-12714 88 1 flow flow NOUN ajst-12714 88 2 chart chart NOUN ajst-12714 88 3 of of ADP ajst-12714 88 4 stn stn PROPN ajst-12714 88 5 attention attention NOUN ajst-12714 88 6 mechanism mechanism NOUN ajst-12714 88 7 3.3 3.3 NUM ajst-12714 88 8 . . PUNCT ajst-12714 89 1 cbam cbam NOUN ajst-12714 89 2 attention attention NOUN ajst-12714 89 3 mechanism mechanism NOUN ajst-12714 89 4 cbam cbam NOUN ajst-12714 89 5 ( ( PUNCT ajst-12714 89 6 convolutional convolutional ADJ ajst-12714 89 7 block block NOUN ajst-12714 89 8 attention attention NOUN ajst-12714 89 9 module module NOUN ajst-12714 89 10 ) ) PUNCT ajst-12714 89 11 is be AUX ajst-12714 89 12 a a DET ajst-12714 89 13 hybrid hybrid ADJ ajst-12714 89 14 attention attention NOUN ajst-12714 89 15 mechanism mechanism NOUN ajst-12714 89 16 commonly commonly ADV ajst-12714 89 17 used use VERB ajst-12714 89 18 in in ADP ajst-12714 89 19 image image NOUN ajst-12714 89 20 recognition recognition NOUN ajst-12714 89 21 tasks task NOUN ajst-12714 89 22 . . PUNCT ajst-12714 90 1 it it PRON ajst-12714 90 2 aims aim VERB ajst-12714 90 3 to to PART ajst-12714 90 4 improve improve VERB ajst-12714 90 5 the the DET ajst-12714 90 6 feature feature NOUN ajst-12714 90 7 representation representation NOUN ajst-12714 90 8 capabilities capability NOUN ajst-12714 90 9 of of ADP ajst-12714 90 10 convolutional convolutional ADJ ajst-12714 90 11 neural neural ADJ ajst-12714 90 12 networks network NOUN ajst-12714 90 13 ( ( PUNCT ajst-12714 90 14 cnns cnns PROPN ajst-12714 90 15 ) ) PUNCT ajst-12714 90 16 , , PUNCT ajst-12714 90 17 allowing allow VERB ajst-12714 90 18 them they PRON ajst-12714 90 19 to to PART ajst-12714 90 20 better well ADV ajst-12714 90 21 capture capture VERB ajst-12714 90 22 important important ADJ ajst-12714 90 23 information information NOUN ajst-12714 90 24 in in ADP ajst-12714 90 25 images image NOUN ajst-12714 90 26 . . PUNCT ajst-12714 91 1 the the DET ajst-12714 91 2 cbam cbam NOUN ajst-12714 91 3 module module NOUN ajst-12714 91 4 consists consist VERB ajst-12714 91 5 of of ADP ajst-12714 91 6 two two NUM ajst-12714 91 7 sub sub NOUN ajst-12714 91 8 - - NOUN ajst-12714 91 9 modules module NOUN ajst-12714 91 10 : : PUNCT ajst-12714 91 11 the the DET ajst-12714 91 12 channel channel NOUN ajst-12714 91 13 attention attention NOUN ajst-12714 91 14 module module NOUN ajst-12714 91 15 and and CCONJ ajst-12714 91 16 the the DET ajst-12714 91 17 spatial spatial ADJ ajst-12714 91 18 attention attention NOUN ajst-12714 91 19 module module NOUN ajst-12714 91 20 ) ) PUNCT ajst-12714 91 21 . . PUNCT ajst-12714 92 1 the the DET ajst-12714 92 2 channel channel NOUN ajst-12714 92 3 attention attention NOUN ajst-12714 92 4 module module NOUN ajst-12714 92 5 adaptively adaptively ADV ajst-12714 92 6 weighted weight VERB ajst-12714 92 7 averages average NOUN ajst-12714 92 8 the the DET ajst-12714 92 9 channel channel NOUN ajst-12714 92 10 dimensions dimension NOUN ajst-12714 92 11 by by ADP ajst-12714 92 12 learning learn VERB ajst-12714 92 13 the the DET ajst-12714 92 14 importance importance NOUN ajst-12714 92 15 weights weight NOUN ajst-12714 92 16 of of ADP ajst-12714 92 17 each each DET ajst-12714 92 18 channel channel NOUN ajst-12714 92 19 . . PUNCT ajst-12714 93 1 this this PRON ajst-12714 93 2 allows allow VERB ajst-12714 93 3 the the DET ajst-12714 93 4 network network NOUN ajst-12714 93 5 to to PART ajst-12714 93 6 focus focus VERB ajst-12714 93 7 on on ADP ajst-12714 93 8 the the DET ajst-12714 93 9 most most ADV ajst-12714 93 10 relevant relevant ADJ ajst-12714 93 11 features feature NOUN ajst-12714 93 12 based base VERB ajst-12714 93 13 on on ADP ajst-12714 93 14 different different ADJ ajst-12714 93 15 channel channel NOUN ajst-12714 93 16 contributions contribution NOUN ajst-12714 93 17 , , PUNCT ajst-12714 93 18 resulting result VERB ajst-12714 93 19 in in ADP ajst-12714 93 20 a a DET ajst-12714 93 21 more more ADV ajst-12714 93 22 representative representative ADJ ajst-12714 93 23 feature feature NOUN ajst-12714 93 24 . . PUNCT ajst-12714 94 1 the the DET ajst-12714 94 2 spatial spatial ADJ ajst-12714 94 3 attention attention NOUN ajst-12714 94 4 module module NOUN ajst-12714 94 5 adaptively adaptively ADV ajst-12714 94 6 weights weight VERB ajst-12714 94 7 the the DET ajst-12714 94 8 spatial spatial ADJ ajst-12714 94 9 dimension dimension NOUN ajst-12714 94 10 of of ADP ajst-12714 94 11 the the DET ajst-12714 94 12 feature feature NOUN ajst-12714 94 13 map map NOUN ajst-12714 94 14 by by ADP ajst-12714 94 15 learning learn VERB ajst-12714 94 16 the the DET ajst-12714 94 17 importance importance NOUN ajst-12714 94 18 weight weight NOUN ajst-12714 94 19 of of ADP ajst-12714 94 20 each each DET ajst-12714 94 21 spatial spatial ADJ ajst-12714 94 22 location location NOUN ajst-12714 94 23 . . PUNCT ajst-12714 95 1 this this PRON ajst-12714 95 2 directs direct VERB ajst-12714 95 3 the the DET ajst-12714 95 4 network network NOUN ajst-12714 95 5 to to PART ajst-12714 95 6 focus focus VERB ajst-12714 95 7 on on ADP ajst-12714 95 8 different different ADJ ajst-12714 95 9 areas area NOUN ajst-12714 95 10 in in ADP ajst-12714 95 11 the the DET ajst-12714 95 12 image image NOUN ajst-12714 95 13 , , PUNCT ajst-12714 95 14 allowing allow VERB ajst-12714 95 15 it it PRON ajst-12714 95 16 to to PART ajst-12714 95 17 better well ADV ajst-12714 95 18 understand understand VERB ajst-12714 95 19 the the DET ajst-12714 95 20 spatial spatial ADJ ajst-12714 95 21 structure structure NOUN ajst-12714 95 22 of of ADP ajst-12714 95 23 the the DET ajst-12714 95 24 target target NOUN ajst-12714 95 25 object object NOUN ajst-12714 95 26 or or CCONJ ajst-12714 95 27 scene[17 scene[17 NOUN ajst-12714 95 28 ] ] PUNCT ajst-12714 95 29 . . PUNCT ajst-12714 96 1 by by ADP ajst-12714 96 2 combining combine VERB ajst-12714 96 3 channel channel NOUN ajst-12714 96 4 attention attention NOUN ajst-12714 96 5 and and CCONJ ajst-12714 96 6 spatial spatial ADJ ajst-12714 96 7 attention attention NOUN ajst-12714 96 8 , , PUNCT ajst-12714 96 9 the the DET ajst-12714 96 10 cbam cbam NOUN ajst-12714 96 11 module module NOUN ajst-12714 96 12 is be AUX ajst-12714 96 13 able able ADJ ajst-12714 96 14 to to PART ajst-12714 96 15 capture capture VERB ajst-12714 96 16 important important ADJ ajst-12714 96 17 information information NOUN ajst-12714 96 18 in in ADP ajst-12714 96 19 images image NOUN ajst-12714 96 20 more more ADV ajst-12714 96 21 comprehensively comprehensively ADV ajst-12714 96 22 and and CCONJ ajst-12714 96 23 improve improve VERB ajst-12714 96 24 the the DET ajst-12714 96 25 performance performance NOUN ajst-12714 96 26 of of ADP ajst-12714 96 27 convolutional convolutional ADJ ajst-12714 96 28 neural neural ADJ ajst-12714 96 29 networks network NOUN ajst-12714 96 30 in in ADP ajst-12714 96 31 various various ADJ ajst-12714 96 32 image image NOUN ajst-12714 96 33 recognition recognition NOUN ajst-12714 96 34 tasks task NOUN ajst-12714 96 35 . . PUNCT ajst-12714 97 1 the the DET ajst-12714 97 2 flowchart flowchart NOUN ajst-12714 97 3 of of ADP ajst-12714 97 4 the the DET ajst-12714 97 5 attention attention NOUN ajst-12714 97 6 mechanism mechanism NOUN ajst-12714 97 7 of of ADP ajst-12714 97 8 cbam cbam NOUN ajst-12714 97 9 is be AUX ajst-12714 97 10 shown show VERB ajst-12714 97 11 in in ADP ajst-12714 97 12 figure figure NOUN ajst-12714 97 13 9 9 NUM ajst-12714 97 14 38 38 NUM ajst-12714 97 15 Β  Β  SPACE ajst-12714 97 16 input input NOUN ajst-12714 97 17 Β  Β  SPACE ajst-12714 97 18 feature feature NOUN ajst-12714 97 19 Β  Β  SPACE ajst-12714 97 20 mapf mapf NOUN ajst-12714 97 21 optimized optimize VERB ajst-12714 97 22 Β  Β  SPACE ajst-12714 97 23 feature feature NOUN ajst-12714 97 24 Β  Β  SPACE ajst-12714 97 25 map map NOUN ajst-12714 97 26 f f PROPN ajst-12714 97 27 ο‚’ο‚’ ο‚’ο‚’ PROPN ajst-12714 97 28 f f PROPN ajst-12714 97 29 ο‚’ ο‚’ PROPN ajst-12714 97 30 c c PROPN ajst-12714 97 31 w w PROPN ajst-12714 97 32 h h PROPN ajst-12714 98 1 c c PROPN ajst-12714 98 2 w w PROPN ajst-12714 98 3 h h PROPN ajst-12714 98 4 maximum maximum ADJ ajst-12714 98 5 Β  Β  SPACE ajst-12714 98 6 pooling pool VERB ajst-12714 98 7 Β  Β  SPACE ajst-12714 98 8 layer layer NOUN ajst-12714 98 9 average average NOUN ajst-12714 98 10 Β  Β  SPACE ajst-12714 98 11 pooling pool VERB ajst-12714 98 12 Β  Β  SPACE ajst-12714 98 13 layer layer NOUN ajst-12714 98 14 multilayer multilayer PROPN ajst-12714 98 15 Β  Β  SPACE ajst-12714 98 16 perceptron perceptron PROPN ajst-12714 98 17 channel channel PROPN ajst-12714 98 18 Β  Β  SPACE ajst-12714 98 19 attention attention NOUN ajst-12714 98 20 Β  Β  SPACE ajst-12714 98 21 module module NOUN ajst-12714 98 22 channel channel NOUN ajst-12714 98 23 Β  Β  SPACE ajst-12714 98 24 attention attention NOUN ajst-12714 98 25 ( ( PUNCT ajst-12714 98 26 ) ) PUNCT ajst-12714 98 27 c c NOUN ajst-12714 98 28 fm fm PROPN ajst-12714 98 29   X ajst-12714 98 30   X ajst-12714 98 31 maximum maximum ADJ ajst-12714 98 32 Β  Β  SPACE ajst-12714 98 33 pooling pool VERB ajst-12714 98 34 Β  Β  SPACE ajst-12714 98 35 layer layer NOUN ajst-12714 98 36 average average NOUN ajst-12714 98 37 Β  Β  SPACE ajst-12714 98 38 pooling pool VERB ajst-12714 98 39 Β  Β  SPACE ajst-12714 98 40 layer layer NOUN ajst-12714 98 41 convo convo PROPN ajst-12714 98 42 lution lution PROPN ajst-12714 98 43 spatial spatial ADJ ajst-12714 98 44 Β  Β  SPACE ajst-12714 98 45 attention attention NOUN ajst-12714 98 46 Β  Β  SPACE ajst-12714 98 47 module module NOUN ajst-12714 98 48 spatial spatial ADJ ajst-12714 98 49 Β  Β  SPACE ajst-12714 98 50 attention attention NOUN ajst-12714 98 51 s s PART ajst-12714 98 52 ( ( PUNCT ajst-12714 98 53 ) ) PUNCT ajst-12714 98 54 fm fm PROPN ajst-12714 98 55 figure figure NOUN ajst-12714 98 56 9 9 NUM ajst-12714 98 57 . . PUNCT ajst-12714 99 1 flow flow VERB ajst-12714 99 2 chart chart NOUN ajst-12714 99 3 of of ADP ajst-12714 99 4 cbam cbam NOUN ajst-12714 99 5 attention attention NOUN ajst-12714 99 6 mechanism mechanism NOUN ajst-12714 99 7 3.4 3.4 NUM ajst-12714 99 8 . . PUNCT ajst-12714 100 1 eca eca NOUN ajst-12714 100 2 attention attention NOUN ajst-12714 100 3 mechanism mechanism NOUN ajst-12714 100 4 the the DET ajst-12714 100 5 eca eca NOUN ajst-12714 100 6 attention attention NOUN ajst-12714 100 7 mechanism mechanism NOUN ajst-12714 100 8 is be AUX ajst-12714 100 9 an an DET ajst-12714 100 10 attention attention NOUN ajst-12714 100 11 mechanism mechanism NOUN ajst-12714 100 12 based base VERB ajst-12714 100 13 on on ADP ajst-12714 100 14 convolutional convolutional ADJ ajst-12714 100 15 neural neural ADJ ajst-12714 100 16 networks network NOUN ajst-12714 100 17 , , PUNCT ajst-12714 100 18 which which PRON ajst-12714 100 19 is be AUX ajst-12714 100 20 used use VERB ajst-12714 100 21 in in ADP ajst-12714 100 22 image image NOUN ajst-12714 100 23 recognition recognition NOUN ajst-12714 100 24 tasks task NOUN ajst-12714 100 25 . . PUNCT ajst-12714 101 1 eca eca NOUN ajst-12714 101 2 stands stand VERB ajst-12714 101 3 for for ADP ajst-12714 101 4 " " PUNCT ajst-12714 101 5 efficient efficient ADJ ajst-12714 101 6 channel channel NOUN ajst-12714 101 7 attention attention NOUN ajst-12714 101 8 , , PUNCT ajst-12714 101 9 " " PUNCT ajst-12714 101 10 which which PRON ajst-12714 101 11 improves improve VERB ajst-12714 101 12 the the DET ajst-12714 101 13 performance performance NOUN ajst-12714 101 14 of of ADP ajst-12714 101 15 feature feature NOUN ajst-12714 101 16 representation representation NOUN ajst-12714 101 17 by by ADP ajst-12714 101 18 modeling model VERB ajst-12714 101 19 associations association NOUN ajst-12714 101 20 between between ADP ajst-12714 101 21 different different ADJ ajst-12714 101 22 channels channel NOUN ajst-12714 101 23 . . PUNCT ajst-12714 102 1 traditional traditional ADJ ajst-12714 102 2 self self NOUN ajst-12714 102 3 - - PUNCT ajst-12714 102 4 attention attention NOUN ajst-12714 102 5 mechanisms mechanism NOUN ajst-12714 102 6 , , PUNCT ajst-12714 102 7 such such ADJ ajst-12714 102 8 as as ADP ajst-12714 102 9 senet senet NOUN ajst-12714 102 10 and and CCONJ ajst-12714 102 11 cbam cbam NOUN ajst-12714 102 12 , , PUNCT ajst-12714 102 13 typically typically ADV ajst-12714 102 14 compute compute VERB ajst-12714 102 15 attention attention NOUN ajst-12714 102 16 weights weight NOUN ajst-12714 102 17 based base VERB ajst-12714 102 18 on on ADP ajst-12714 102 19 spatial spatial ADJ ajst-12714 102 20 dimensions dimension NOUN ajst-12714 102 21 , , PUNCT ajst-12714 102 22 while while SCONJ ajst-12714 102 23 eca eca NOUN ajst-12714 102 24 attention attention NOUN ajst-12714 102 25 mechanisms mechanism NOUN ajst-12714 102 26 focus focus VERB ajst-12714 102 27 on on ADP ajst-12714 102 28 channel channel NOUN ajst-12714 102 29 dimensions dimension NOUN ajst-12714 102 30 . . PUNCT ajst-12714 103 1 it it PRON ajst-12714 103 2 models model VERB ajst-12714 103 3 dependencies dependency NOUN ajst-12714 103 4 between between ADP ajst-12714 103 5 individual individual ADJ ajst-12714 103 6 channels channel NOUN ajst-12714 103 7 by by ADP ajst-12714 103 8 introducing introduce VERB ajst-12714 103 9 one one NUM ajst-12714 103 10 - - PUNCT ajst-12714 103 11 dimensional dimensional ADJ ajst-12714 103 12 convolution convolution NOUN ajst-12714 103 13 operations[18 operations[18 X ajst-12714 103 14 ] ] PUNCT ajst-12714 103 15 . . PUNCT ajst-12714 104 1 by by ADP ajst-12714 104 2 introducing introduce VERB ajst-12714 104 3 the the DET ajst-12714 104 4 eca eca NOUN ajst-12714 104 5 attention attention NOUN ajst-12714 104 6 mechanism mechanism NOUN ajst-12714 104 7 , , PUNCT ajst-12714 104 8 the the DET ajst-12714 104 9 performance performance NOUN ajst-12714 104 10 of of ADP ajst-12714 104 11 convolutional convolutional ADJ ajst-12714 104 12 neural neural ADJ ajst-12714 104 13 networks network NOUN ajst-12714 104 14 in in ADP ajst-12714 104 15 tasks task NOUN ajst-12714 104 16 such such ADJ ajst-12714 104 17 as as ADP ajst-12714 104 18 image image NOUN ajst-12714 104 19 classification classification NOUN ajst-12714 104 20 and and CCONJ ajst-12714 104 21 object object NOUN ajst-12714 104 22 detection detection NOUN ajst-12714 104 23 can can AUX ajst-12714 104 24 be be AUX ajst-12714 104 25 improved improve VERB ajst-12714 104 26 , , PUNCT ajst-12714 104 27 and and CCONJ ajst-12714 104 28 it it PRON ajst-12714 104 29 has have VERB ajst-12714 104 30 the the DET ajst-12714 104 31 advantages advantage NOUN ajst-12714 104 32 of of ADP ajst-12714 104 33 high high ADJ ajst-12714 104 34 computational computational ADJ ajst-12714 104 35 efficiency efficiency NOUN ajst-12714 104 36 and and CCONJ ajst-12714 104 37 low low ADJ ajst-12714 104 38 number number NOUN ajst-12714 104 39 of of ADP ajst-12714 104 40 parameters[19 parameters[19 NOUN ajst-12714 104 41 ] ] PUNCT ajst-12714 104 42 . . PUNCT ajst-12714 105 1 the the DET ajst-12714 105 2 flowchart flowchart NOUN ajst-12714 105 3 of of ADP ajst-12714 105 4 the the DET ajst-12714 105 5 attention attention NOUN ajst-12714 105 6 mechanism mechanism NOUN ajst-12714 105 7 of of ADP ajst-12714 105 8 cbam cbam NOUN ajst-12714 105 9 is be AUX ajst-12714 105 10 shown show VERB ajst-12714 105 11 in in ADP ajst-12714 105 12 figure figure NOUN ajst-12714 105 13 10 10 NUM ajst-12714 105 14 figure figure NOUN ajst-12714 105 15 10 10 NUM ajst-12714 105 16 . . PUNCT ajst-12714 106 1 flow flow VERB ajst-12714 106 2 chart chart NOUN ajst-12714 106 3 of of ADP ajst-12714 106 4 eca eca NOUN ajst-12714 106 5 attention attention NOUN ajst-12714 106 6 mechanism mechanism NOUN ajst-12714 106 7 in in ADP ajst-12714 106 8 this this DET ajst-12714 106 9 paper paper NOUN ajst-12714 106 10 , , PUNCT ajst-12714 106 11 the the DET ajst-12714 106 12 vgg16 vgg16 NOUN ajst-12714 106 13 network network NOUN ajst-12714 106 14 structure structure NOUN ajst-12714 106 15 was be AUX ajst-12714 106 16 used use VERB ajst-12714 106 17 as as ADP ajst-12714 106 18 the the DET ajst-12714 106 19 main main ADJ ajst-12714 106 20 stem stem NOUN ajst-12714 106 21 to to PART ajst-12714 106 22 perform perform VERB ajst-12714 106 23 ablation ablation NOUN ajst-12714 106 24 experiments experiment NOUN ajst-12714 106 25 on on ADP ajst-12714 106 26 four four NUM ajst-12714 106 27 attention attention NOUN ajst-12714 106 28 mechanisms[20 mechanisms[20 NOUN ajst-12714 106 29 ] ] PUNCT ajst-12714 106 30 . . PUNCT ajst-12714 107 1 while while SCONJ ajst-12714 107 2 extracting extract VERB ajst-12714 107 3 deep deep ADJ ajst-12714 107 4 - - PUNCT ajst-12714 107 5 level level NOUN ajst-12714 107 6 features feature NOUN ajst-12714 107 7 , , PUNCT ajst-12714 107 8 the the DET ajst-12714 107 9 v v NOUN ajst-12714 107 10 gg16 gg16 PROPN ajst-12714 107 11 network network NOUN ajst-12714 107 12 accesses access VERB ajst-12714 107 13 different different ADJ ajst-12714 107 14 attention attention NOUN ajst-12714 107 15 modules module NOUN ajst-12714 107 16 in in ADP ajst-12714 107 17 each each DET ajst-12714 107 18 convolutional convolutional ADJ ajst-12714 107 19 layer layer NOUN ajst-12714 107 20 , , PUNCT ajst-12714 107 21 so so SCONJ ajst-12714 107 22 as as SCONJ ajst-12714 107 23 to to PART ajst-12714 107 24 compare compare VERB ajst-12714 107 25 which which DET ajst-12714 107 26 module module NOUN ajst-12714 107 27 has have VERB ajst-12714 107 28 the the DET ajst-12714 107 29 best good ADJ ajst-12714 107 30 performance performance NOUN ajst-12714 107 31 for for ADP ajst-12714 107 32 the the DET ajst-12714 107 33 network network NOUN ajst-12714 107 34 . . PUNCT ajst-12714 108 1 based base VERB ajst-12714 108 2 on on ADP ajst-12714 108 3 different different ADJ ajst-12714 108 4 attention attention NOUN ajst-12714 108 5 mechanisms mechanism NOUN ajst-12714 108 6 , , PUNCT ajst-12714 108 7 four four NUM ajst-12714 108 8 attention attention NOUN ajst-12714 108 9 mechanisms mechanism NOUN ajst-12714 108 10 are be AUX ajst-12714 108 11 designed design VERB ajst-12714 108 12 for for ADP ajst-12714 108 13 comparison comparison NOUN ajst-12714 108 14 , , PUNCT ajst-12714 108 15 namely namely ADV ajst-12714 108 16 the the DET ajst-12714 108 17 v v NOUN ajst-12714 108 18 gg16 gg16 PROPN ajst-12714 108 19 network network NOUN ajst-12714 108 20 ( ( PUNCT ajst-12714 108 21 s s NOUN ajst-12714 108 22 e e NOUN ajst-12714 108 23 - - NOUN ajst-12714 108 24 vgg16 vgg16 NOUN ajst-12714 108 25 ) ) PUNCT ajst-12714 108 26 combined combine VERB ajst-12714 108 27 with with ADP ajst-12714 108 28 the the DET ajst-12714 108 29 channel channel NOUN ajst-12714 108 30 attention attention NOUN ajst-12714 108 31 mechanism mechanism NOUN ajst-12714 108 32 , , PUNCT ajst-12714 108 33 and and CCONJ ajst-12714 108 34 the the DET ajst-12714 108 35 vgg16 vgg16 NOUN ajst-12714 108 36 network network NOUN ajst-12714 108 37 combined combine VERB ajst-12714 108 38 with with ADP ajst-12714 108 39 the the DET ajst-12714 108 40 spatial spatial ADJ ajst-12714 108 41 attention attention NOUN ajst-12714 108 42 mechanism mechanism NOUN ajst-12714 108 43 ( ( PUNCT ajst-12714 108 44 stn stn PROPN ajst-12714 108 45 - - PUNCT ajst-12714 108 46 vgg16 vgg16 PROPN ajst-12714 108 47 ) ) PUNCT ajst-12714 108 48 , , PUNCT ajst-12714 108 49 v v ADP ajst-12714 108 50 gg16 gg16 PROPN ajst-12714 108 51 network network NOUN ajst-12714 108 52 combined combine VERB ajst-12714 108 53 with with ADP ajst-12714 108 54 mixed mixed ADJ ajst-12714 108 55 attention attention NOUN ajst-12714 108 56 mechanism mechanism NOUN ajst-12714 108 57 ( ( PUNCT ajst-12714 108 58 cbam cbam NOUN ajst-12714 108 59 - - PUNCT ajst-12714 108 60 vgg16 vgg16 NOUN ajst-12714 108 61 ) ) PUNCT ajst-12714 108 62 , , PUNCT ajst-12714 108 63 attention attention NOUN ajst-12714 108 64 mechanism mechanism NOUN ajst-12714 108 65 combined combine VERB ajst-12714 108 66 with with ADP ajst-12714 108 67 convolutional convolutional ADJ ajst-12714 108 68 neural neural ADJ ajst-12714 108 69 network network NOUN ajst-12714 108 70 ( ( PUNCT ajst-12714 108 71 eca eca NOUN ajst-12714 108 72 - - PUNCT ajst-12714 108 73 vgg16 vgg16 NOUN ajst-12714 108 74 ) ) PUNCT ajst-12714 108 75 . . PUNCT ajst-12714 109 1 experimental experimental ADJ ajst-12714 109 2 environment environment NOUN ajst-12714 109 3 introduction introduction NOUN ajst-12714 109 4 : : PUNCT ajst-12714 109 5 windows10 windows10 NOUN ajst-12714 109 6 64bit 64bit NOUN ajst-12714 109 7 operating operating NOUN ajst-12714 109 8 device device NOUN ajst-12714 109 9 , , PUNCT ajst-12714 109 10 ansys ansys PROPN ajst-12714 109 11 maxwell maxwell PROPN ajst-12714 109 12 2021 2021 NUM ajst-12714 109 13 r2 r2 PROPN ajst-12714 109 14 , , PUNCT ajst-12714 109 15 python python NOUN ajst-12714 109 16 10.0 10.0 NUM ajst-12714 109 17 ; ; PUNCT ajst-12714 109 18 hardware hardware NOUN ajst-12714 109 19 : : PUNCT ajst-12714 109 20 intel(r intel(r NOUN ajst-12714 109 21 ) ) PUNCT ajst-12714 109 22 core(tm core(tm NOUN ajst-12714 109 23 ) ) PUNCT ajst-12714 109 24 i7 i7 ADJ ajst-12714 109 25 - - PUNCT ajst-12714 109 26 7700hq 7700hq NOUN ajst-12714 109 27 cpu cpu NOUN ajst-12714 109 28 @ @ ADP ajst-12714 109 29 2.80ghz 2.80ghz NUM ajst-12714 109 30 , , PUNCT ajst-12714 109 31 8 8 NUM ajst-12714 109 32 gb gb NOUN ajst-12714 109 33 of of ADP ajst-12714 109 34 running run VERB ajst-12714 109 35 memory memory NOUN ajst-12714 109 36 . . PUNCT ajst-12714 110 1 the the DET ajst-12714 110 2 model model NOUN ajst-12714 110 3 performance performance NOUN ajst-12714 110 4 evaluation evaluation NOUN ajst-12714 110 5 is be AUX ajst-12714 110 6 shown show VERB ajst-12714 110 7 in in ADP ajst-12714 110 8 table table NOUN ajst-12714 110 9 2 2 NUM ajst-12714 110 10 - - SYM ajst-12714 110 11 1 1 NUM ajst-12714 110 12 , , PUNCT ajst-12714 110 13 where where SCONJ ajst-12714 110 14 " " PUNCT ajst-12714 110 15 √ √ PROPN ajst-12714 110 16 " " PUNCT ajst-12714 110 17 indicates indicate VERB ajst-12714 110 18 that that SCONJ ajst-12714 110 19 the the DET ajst-12714 110 20 module module NOUN ajst-12714 110 21 is be AUX ajst-12714 110 22 added add VERB ajst-12714 110 23 for for ADP ajst-12714 110 24 the the DET ajst-12714 110 25 vgg16 vgg16 NOUN ajst-12714 110 26 network network NOUN ajst-12714 110 27 , , PUNCT ajst-12714 110 28 and and CCONJ ajst-12714 110 29 no no DET ajst-12714 110 30 " " PUNCT ajst-12714 110 31 √ √ NOUN ajst-12714 110 32 " " PUNCT ajst-12714 110 33 indicates indicate VERB ajst-12714 110 34 that that SCONJ ajst-12714 110 35 the the DET ajst-12714 110 36 module module NOUN ajst-12714 110 37 is be AUX ajst-12714 110 38 not not PART ajst-12714 110 39 added add VERB ajst-12714 110 40 . . PUNCT ajst-12714 111 1 table table NOUN ajst-12714 111 2 1 1 NUM ajst-12714 111 3 . . PUNCT ajst-12714 112 1 attention attention NOUN ajst-12714 112 2 modules module NOUN ajst-12714 112 3 were be AUX ajst-12714 112 4 added add VERB ajst-12714 112 5 one one NUM ajst-12714 112 6 by by ADP ajst-12714 112 7 one one NUM ajst-12714 112 8 for for ADP ajst-12714 112 9 ablation ablation NOUN ajst-12714 112 10 analysis analysis NOUN ajst-12714 112 11 experiments experiment NOUN ajst-12714 112 12 grid grid NOUN ajst-12714 112 13 name name PROPN ajst-12714 112 14 se se PROPN ajst-12714 112 15 stn stn PROPN ajst-12714 112 16 cbam cbam PROPN ajst-12714 112 17 it'slike it'slike PROPN ajst-12714 112 18 accuracy accuracy NOUN ajst-12714 112 19 in in ADP ajst-12714 112 20 gg16 gg16 PROPN ajst-12714 112 21 90.33 90.33 NUM ajst-12714 112 22 % % NOUN ajst-12714 112 23 sevgg16 sevgg16 NOUN ajst-12714 112 24 √ √ NOUN ajst-12714 112 25 94.67 94.67 NUM ajst-12714 112 26 % % NOUN ajst-12714 112 27 stnvgg16 stnvgg16 NOUN ajst-12714 112 28 √ √ VERB ajst-12714 112 29 93.72 93.72 NUM ajst-12714 112 30 % % NOUN ajst-12714 112 31 cbamvgg16 cbamvgg16 NOUN ajst-12714 112 32 √ √ ADP ajst-12714 112 33 96.44 96.44 NUM ajst-12714 112 34 % % NOUN ajst-12714 112 35 ecavgg16 ecavgg16 NOUN ajst-12714 112 36 √ √ NOUN ajst-12714 112 37 98.93 98.93 NUM ajst-12714 112 38 % % NOUN ajst-12714 112 39 note note NOUN ajst-12714 112 40 : : PUNCT ajst-12714 112 41 the the DET ajst-12714 112 42 text text NOUN ajst-12714 112 43 in in ADP ajst-12714 112 44 bold bold ADJ ajst-12714 112 45 in in ADP ajst-12714 112 46 the the DET ajst-12714 112 47 table table NOUN ajst-12714 112 48 represents represent VERB ajst-12714 112 49 the the DET ajst-12714 112 50 optimal optimal ADJ ajst-12714 112 51 network network NOUN ajst-12714 112 52 as as SCONJ ajst-12714 112 53 shown show VERB ajst-12714 112 54 in in ADP ajst-12714 112 55 the the DET ajst-12714 112 56 table table NOUN ajst-12714 112 57 , , PUNCT ajst-12714 112 58 the the DET ajst-12714 112 59 eca eca NOUN ajst-12714 112 60 - - PUNCT ajst-12714 112 61 vgg vgg PROPN ajst-12714 112 62 6 6 NUM ajst-12714 112 63 network network NOUN ajst-12714 112 64 has have VERB ajst-12714 112 65 the the DET ajst-12714 112 66 highest high ADJ ajst-12714 112 67 accuracy accuracy NOUN ajst-12714 112 68 in in ADP ajst-12714 112 69 this this DET ajst-12714 112 70 paper paper NOUN ajst-12714 112 71 , , PUNCT ajst-12714 112 72 with with ADP ajst-12714 112 73 the the DET ajst-12714 112 74 same same ADJ ajst-12714 112 75 accuracy accuracy NOUN ajst-12714 112 76 as as ADP ajst-12714 112 77 v v ADP ajst-12714 112 78 gg16 gg16 PROPN ajst-12714 112 79 , , PUNCT ajst-12714 112 80 s s PART ajst-12714 112 81 e e NOUN ajst-12714 112 82 - - NOUN ajst-12714 112 83 vgg16 vgg16 ADJ ajst-12714 112 84 , , PUNCT ajst-12714 112 85 s s PART ajst-12714 112 86 tn tn NOUN ajst-12714 112 87 - - PUNCT ajst-12714 112 88 vgg16 vgg16 NOUN ajst-12714 112 89 , , PUNCT ajst-12714 112 90 c c NOUN ajst-12714 112 91 bam bam NOUN ajst-12714 112 92 - - PUNCT ajst-12714 112 93 vgg16 vgg16 NOUN ajst-12714 112 94 compared compare VERB ajst-12714 112 95 with with ADP ajst-12714 112 96 other other ADJ ajst-12714 112 97 networks network NOUN ajst-12714 112 98 , , PUNCT ajst-12714 112 99 the the DET ajst-12714 112 100 e e NOUN ajst-12714 112 101 ca can AUX ajst-12714 112 102 - - PUNCT ajst-12714 112 103 vgg16 vgg16 NOUN ajst-12714 112 104 neural neural ADJ ajst-12714 112 105 network network NOUN ajst-12714 112 106 model model NOUN ajst-12714 112 107 combined combine VERB ajst-12714 112 108 with with ADP ajst-12714 112 109 the the DET ajst-12714 112 110 attention attention NOUN ajst-12714 112 111 mechanism mechanism NOUN ajst-12714 112 112 of of ADP ajst-12714 112 113 convolutional convolutional ADJ ajst-12714 112 114 neural neural ADJ ajst-12714 112 115 networks network NOUN ajst-12714 112 116 proposed propose VERB ajst-12714 112 117 in in ADP ajst-12714 112 118 this this DET ajst-12714 112 119 paper paper NOUN ajst-12714 112 120 is be AUX ajst-12714 112 121 better well ADJ ajst-12714 112 122 . . PUNCT ajst-12714 113 1 in in ADP ajst-12714 113 2 this this DET ajst-12714 113 3 paper paper NOUN ajst-12714 113 4 , , PUNCT ajst-12714 113 5 an an DET ajst-12714 113 6 attention attention NOUN ajst-12714 113 7 mechanism mechanism NOUN ajst-12714 113 8 combining combine VERB ajst-12714 113 9 convolutional convolutional ADJ ajst-12714 113 10 neural neural ADJ ajst-12714 113 11 networks network NOUN ajst-12714 113 12 is be AUX ajst-12714 113 13 proposed propose VERB ajst-12714 113 14 to to PART ajst-12714 113 15 improve improve VERB ajst-12714 113 16 the the DET ajst-12714 113 17 vgg16 vgg16 NOUN ajst-12714 113 18 neural neural ADJ ajst-12714 113 19 network network NOUN ajst-12714 113 20 to to PART ajst-12714 113 21 obtain obtain VERB ajst-12714 113 22 the the DET ajst-12714 113 23 eca eca NOUN ajst-12714 113 24 - - PUNCT ajst-12714 113 25 vgg16 vgg16 NOUN ajst-12714 113 26 neural neural ADJ ajst-12714 113 27 network network NOUN ajst-12714 113 28 . . PUNCT ajst-12714 114 1 the the DET ajst-12714 114 2 basic basic ADJ ajst-12714 114 3 components component NOUN ajst-12714 114 4 of of ADP ajst-12714 114 5 the the DET ajst-12714 114 6 eca eca NOUN ajst-12714 114 7 - - PUNCT ajst-12714 114 8 vgg16 vgg16 NOUN ajst-12714 114 9 neural neural ADJ ajst-12714 114 10 network network NOUN ajst-12714 114 11 introduced introduce VERB ajst-12714 114 12 in in ADP ajst-12714 114 13 this this DET ajst-12714 114 14 paper paper NOUN ajst-12714 114 15 introduce introduce VERB ajst-12714 114 16 the the DET ajst-12714 114 17 convolutional convolutional ADJ ajst-12714 114 18 layer layer NOUN ajst-12714 114 19 , , PUNCT ajst-12714 114 20 pooling pool VERB ajst-12714 114 21 layer layer NOUN ajst-12714 114 22 , , PUNCT ajst-12714 114 23 normalization normalization NOUN ajst-12714 114 24 , , PUNCT ajst-12714 114 25 fully fully ADV ajst-12714 114 26 connected connected ADJ ajst-12714 114 27 layer layer NOUN ajst-12714 114 28 , , PUNCT ajst-12714 114 29 activation activation NOUN ajst-12714 114 30 function function NOUN ajst-12714 114 31 and and CCONJ ajst-12714 114 32 loss loss NOUN ajst-12714 114 33 function function NOUN ajst-12714 114 34 . . PUNCT ajst-12714 115 1 the the DET ajst-12714 115 2 basic basic ADJ ajst-12714 115 3 components component NOUN ajst-12714 115 4 of of ADP ajst-12714 115 5 the the DET ajst-12714 115 6 attention attention NOUN ajst-12714 115 7 mechanism mechanism NOUN ajst-12714 115 8 are be AUX ajst-12714 115 9 briefly briefly ADV ajst-12714 115 10 described describe VERB ajst-12714 115 11 , , PUNCT ajst-12714 115 12 and and CCONJ ajst-12714 115 13 the the DET ajst-12714 115 14 structure structure NOUN ajst-12714 115 15 and and CCONJ ajst-12714 115 16 operation operation NOUN ajst-12714 115 17 process process NOUN ajst-12714 115 18 of of ADP ajst-12714 115 19 the the DET ajst-12714 115 20 four four NUM ajst-12714 115 21 attention attention NOUN ajst-12714 115 22 mechanisms mechanism NOUN ajst-12714 115 23 are be AUX ajst-12714 115 24 introduced introduce VERB ajst-12714 115 25 from from ADP ajst-12714 115 26 four four NUM ajst-12714 115 27 aspects aspect NOUN ajst-12714 115 28 : : PUNCT ajst-12714 115 29 channel channel NOUN ajst-12714 115 30 attention attention NOUN ajst-12714 115 31 mechanism mechanism NOUN ajst-12714 115 32 , , PUNCT ajst-12714 115 33 spatial spatial ADJ ajst-12714 115 34 attention attention NOUN ajst-12714 115 35 mechanism mechanism NOUN ajst-12714 115 36 , , PUNCT ajst-12714 115 37 mixed mixed ADJ ajst-12714 115 38 attention attention NOUN ajst-12714 115 39 mechanism mechanism NOUN ajst-12714 115 40 , , PUNCT ajst-12714 115 41 and and CCONJ ajst-12714 115 42 attention attention NOUN ajst-12714 115 43 mechanism mechanism NOUN ajst-12714 115 44 combined combine VERB ajst-12714 115 45 with with ADP ajst-12714 115 46 convolutional convolutional ADJ ajst-12714 115 47 neural neural ADJ ajst-12714 115 48 network network NOUN ajst-12714 115 49 . . PUNCT ajst-12714 116 1 combined combine VERB ajst-12714 116 2 with with ADP ajst-12714 116 3 the the DET ajst-12714 116 4 concentration concentration NOUN ajst-12714 116 5 mechanism mechanism NOUN ajst-12714 116 6 , , PUNCT ajst-12714 116 7 the the DET ajst-12714 116 8 v v NOUN ajst-12714 116 9 gg16 gg16 PROPN ajst-12714 116 10 neural neural ADJ ajst-12714 116 11 network network NOUN ajst-12714 116 12 was be AUX ajst-12714 116 13 optimized optimize VERB ajst-12714 116 14 , , PUNCT ajst-12714 116 15 and and CCONJ ajst-12714 116 16 the the DET ajst-12714 116 17 s s X ajst-12714 116 18 e e ADJ ajst-12714 116 19 - - ADJ ajst-12714 116 20 vgg16 vgg16 ADJ ajst-12714 116 21 network network NOUN ajst-12714 116 22 , , PUNCT ajst-12714 116 23 s s PART ajst-12714 116 24 tn tn NOUN ajst-12714 116 25 - - PUNCT ajst-12714 116 26 vgg16 vgg16 NOUN ajst-12714 116 27 network network NOUN ajst-12714 116 28 , , PUNCT ajst-12714 116 29 c c NOUN ajst-12714 116 30 bam bam NOUN ajst-12714 116 31 - - PUNCT ajst-12714 116 32 vgg16 vgg16 NOUN ajst-12714 116 33 network network NOUN ajst-12714 116 34 and and CCONJ ajst-12714 116 35 eca eca NOUN ajst-12714 116 36 - - PUNCT ajst-12714 116 37 vgg16 vgg16 PROPN ajst-12714 116 38 were be AUX ajst-12714 116 39 formed form VERB ajst-12714 116 40 network network NOUN ajst-12714 116 41 , , PUNCT ajst-12714 116 42 then then ADV ajst-12714 116 43 perform perform VERB ajst-12714 116 44 ablation ablation NOUN ajst-12714 116 45 experiments experiment NOUN ajst-12714 116 46 on on ADP ajst-12714 116 47 the the DET ajst-12714 116 48 pipe pipe NOUN ajst-12714 116 49 flux flux NOUN ajst-12714 116 50 leakage leakage NOUN ajst-12714 116 51 data datum NOUN ajst-12714 116 52 . . PUNCT ajst-12714 117 1 the the DET ajst-12714 117 2 results result NOUN ajst-12714 117 3 show show VERB ajst-12714 117 4 that that SCONJ ajst-12714 117 5 compared compare VERB ajst-12714 117 6 with with ADP ajst-12714 117 7 v v ADP ajst-12714 117 8 gg16 gg16 PROPN ajst-12714 117 9 , , PUNCT ajst-12714 117 10 s s PART ajst-12714 117 11 e e NOUN ajst-12714 117 12 - - NOUN ajst-12714 117 13 vgg16 vgg16 ADJ ajst-12714 117 14 , , PUNCT ajst-12714 117 15 s s PART ajst-12714 117 16 tn tn NOUN ajst-12714 117 17 - - PUNCT ajst-12714 117 18 vgg16 vgg16 NOUN ajst-12714 117 19 , , PUNCT ajst-12714 117 20 cbam cbam NOUN ajst-12714 117 21 - - PUNCT ajst-12714 117 22 vgg16 vgg16 PROPN ajst-12714 117 23 and and CCONJ ajst-12714 117 24 other other ADJ ajst-12714 117 25 networks network NOUN ajst-12714 117 26 , , PUNCT ajst-12714 117 27 the the DET ajst-12714 117 28 proposed propose VERB ajst-12714 117 29 e e NOUN ajst-12714 117 30 is be AUX ajst-12714 117 31 proposed propose VERB ajst-12714 117 32 to to PART ajst-12714 117 33 add add VERB ajst-12714 117 34 the the DET ajst-12714 117 35 attention attention NOUN ajst-12714 117 36 mechanism mechanism NOUN ajst-12714 117 37 combined combine VERB ajst-12714 117 38 with with ADP ajst-12714 117 39 convolutional convolutional ADJ ajst-12714 117 40 neural neural ADJ ajst-12714 117 41 networksthe networksthe NOUN ajst-12714 117 42 accuracy accuracy NOUN ajst-12714 117 43 of of ADP ajst-12714 117 44 the the DET ajst-12714 117 45 ca ca NOUN ajst-12714 117 46 - - PUNCT ajst-12714 117 47 vgg16 vgg16 NOUN ajst-12714 117 48 neural neural ADJ ajst-12714 117 49 network network NOUN ajst-12714 117 50 model model NOUN ajst-12714 117 51 is be AUX ajst-12714 117 52 higher high ADJ ajst-12714 117 53 than than ADP ajst-12714 117 54 that that PRON ajst-12714 117 55 of of ADP ajst-12714 117 56 other other ADJ ajst-12714 117 57 networks network NOUN ajst-12714 117 58 . . PUNCT ajst-12714 118 1 eca eca NOUN ajst-12714 118 2 - - PUNCT ajst-12714 118 3 vgg16 vgg16 PROPN ajst-12714 118 4 has have VERB ajst-12714 118 5 the the DET ajst-12714 118 6 most most ADV ajst-12714 118 7 obvious obvious ADJ ajst-12714 118 8 performance performance NOUN ajst-12714 118 9 improvement improvement NOUN ajst-12714 118 10 , , PUNCT ajst-12714 118 11 and and CCONJ ajst-12714 118 12 the the DET ajst-12714 118 13 accuracy accuracy NOUN ajst-12714 118 14 of of ADP ajst-12714 118 15 pipe pipe NOUN ajst-12714 118 16 defect defect NOUN ajst-12714 118 17 size size NOUN ajst-12714 118 18 prediction prediction NOUN ajst-12714 118 19 is be AUX ajst-12714 118 20 higher high ADJ ajst-12714 118 21 . . PUNCT ajst-12714 119 1 39 39 NUM ajst-12714 119 2 4 4 NUM ajst-12714 119 3 . . X ajst-12714 119 4 concluding conclude VERB ajst-12714 119 5 remarks remark NOUN ajst-12714 119 6 aiming aim VERB ajst-12714 119 7 at at ADP ajst-12714 119 8 the the DET ajst-12714 119 9 low low ADJ ajst-12714 119 10 efficiency efficiency NOUN ajst-12714 119 11 of of ADP ajst-12714 119 12 manual manual ADJ ajst-12714 119 13 determination determination NOUN ajst-12714 119 14 of of ADP ajst-12714 119 15 pipeline pipeline NOUN ajst-12714 119 16 defects defect NOUN ajst-12714 119 17 and and CCONJ ajst-12714 119 18 the the DET ajst-12714 119 19 low low ADJ ajst-12714 119 20 accuracy accuracy NOUN ajst-12714 119 21 of of ADP ajst-12714 119 22 shallow shallow ADJ ajst-12714 119 23 network network NOUN ajst-12714 119 24 prediction prediction NOUN ajst-12714 119 25 of of ADP ajst-12714 119 26 pipeline pipeline NOUN ajst-12714 119 27 defect defect NOUN ajst-12714 119 28 size size NOUN ajst-12714 119 29 , , PUNCT ajst-12714 119 30 an an DET ajst-12714 119 31 e e NOUN ajst-12714 119 32 ca can AUX ajst-12714 119 33 - - PUNCT ajst-12714 119 34 vgg16 vgg16 NOUN ajst-12714 119 35 neural neural ADJ ajst-12714 119 36 network network NOUN ajst-12714 119 37 model model NOUN ajst-12714 119 38 combined combine VERB ajst-12714 119 39 with with ADP ajst-12714 119 40 the the DET ajst-12714 119 41 attention attention NOUN ajst-12714 119 42 mechanism mechanism NOUN ajst-12714 119 43 of of ADP ajst-12714 119 44 convolutional convolutional ADJ ajst-12714 119 45 neural neural ADJ ajst-12714 119 46 network network NOUN ajst-12714 119 47 is be AUX ajst-12714 119 48 proposed propose VERB ajst-12714 119 49 . . PUNCT ajst-12714 120 1 compared compare VERB ajst-12714 120 2 with with ADP ajst-12714 120 3 traditional traditional ADJ ajst-12714 120 4 convolutional convolutional ADJ ajst-12714 120 5 neural neural ADJ ajst-12714 120 6 networks network NOUN ajst-12714 120 7 , , PUNCT ajst-12714 120 8 the the DET ajst-12714 120 9 eca eca NOUN ajst-12714 120 10 - - PUNCT ajst-12714 120 11 vgg16 vgg16 NOUN ajst-12714 120 12 method method NOUN ajst-12714 120 13 has have VERB ajst-12714 120 14 more more ADV ajst-12714 120 15 accurate accurate ADJ ajst-12714 120 16 pipeline pipeline NOUN ajst-12714 120 17 defect defect NOUN ajst-12714 120 18 prediction prediction NOUN ajst-12714 120 19 ability ability NOUN ajst-12714 120 20 and and CCONJ ajst-12714 120 21 training training NOUN ajst-12714 120 22 efficiency efficiency NOUN ajst-12714 120 23 , , PUNCT ajst-12714 120 24 and and CCONJ ajst-12714 120 25 its its PRON ajst-12714 120 26 accuracy accuracy NOUN ajst-12714 120 27 rate rate NOUN ajst-12714 120 28 can can AUX ajst-12714 120 29 reach reach VERB ajst-12714 120 30 98.93 98.93 NUM ajst-12714 120 31 % % NOUN ajst-12714 120 32 . . PUNCT ajst-12714 121 1 the the DET ajst-12714 121 2 results result NOUN ajst-12714 121 3 show show VERB ajst-12714 121 4 that that SCONJ ajst-12714 121 5 the the DET ajst-12714 121 6 proposed propose VERB ajst-12714 121 7 method method NOUN ajst-12714 121 8 proposes propose VERB ajst-12714 121 9 a a DET ajst-12714 121 10 better well ADJ ajst-12714 121 11 idea idea NOUN ajst-12714 121 12 and and CCONJ ajst-12714 121 13 method method NOUN ajst-12714 121 14 for for ADP ajst-12714 121 15 predicting predict VERB ajst-12714 121 16 the the DET ajst-12714 121 17 defects defect NOUN ajst-12714 121 18 of of ADP ajst-12714 121 19 magnetic magnetic ADJ ajst-12714 121 20 flux flux NOUN ajst-12714 121 21 leakage leakage NOUN ajst-12714 121 22 detection detection NOUN ajst-12714 121 23 in in ADP ajst-12714 121 24 oil oil NOUN ajst-12714 121 25 and and CCONJ ajst-12714 121 26 gas gas NOUN ajst-12714 121 27 pipelines pipeline NOUN ajst-12714 121 28 . . PUNCT ajst-12714 122 1 references reference NOUN ajst-12714 122 2 [ [ X ajst-12714 122 3 1 1 NUM ajst-12714 122 4 ] ] X ajst-12714 122 5 geng geng PROPN ajst-12714 122 6 hao hao PROPN ajst-12714 122 7 , , PUNCT ajst-12714 122 8 xia xia PROPN ajst-12714 122 9 hao hao PROPN ajst-12714 122 10 , , PUNCT ajst-12714 122 11 wang wang PROPN ajst-12714 122 12 guoqing guoqing PROPN ajst-12714 122 13 . . PUNCT ajst-12714 123 1 research research NOUN ajst-12714 123 2 on on ADP ajst-12714 123 3 locating locate VERB ajst-12714 123 4 defects defect NOUN ajst-12714 123 5 of of ADP ajst-12714 123 6 inner inner ADJ ajst-12714 123 7 and and CCONJ ajst-12714 123 8 outer outer ADJ ajst-12714 123 9 wall wall NOUN ajst-12714 123 10 of of ADP ajst-12714 123 11 pipelines pipeline NOUN ajst-12714 123 12 during during ADP ajst-12714 123 13 highspeed highspeed NOUN ajst-12714 123 14 magnetic magnetic ADJ ajst-12714 123 15 leakage leakage NOUN ajst-12714 123 16 detection[j detection[j PROPN ajst-12714 123 17 ] ] PUNCT ajst-12714 123 18 . . PUNCT ajst-12714 124 1 chinese chinese ADJ ajst-12714 124 2 journal journal PROPN ajst-12714 124 3 of of ADP ajst-12714 124 4 scientific scientific ADJ ajst-12714 124 5 instrument instrument NOUN ajst-12714 124 6 , , PUNCT ajst-12714 124 7 2022 2022 NUM ajst-12714 124 8 , , PUNCT ajst-12714 124 9 43(04 43(04 NUM ajst-12714 124 10 ): ): PUNCT ajst-12714 124 11 70 70 NUM ajst-12714 124 12 - - SYM ajst-12714 124 13 78 78 NUM ajst-12714 124 14 . . PUNCT ajst-12714 125 1 [ [ X ajst-12714 125 2 2 2 NUM ajst-12714 125 3 ] ] PUNCT ajst-12714 125 4 li li PROPN ajst-12714 125 5 yunhui yunhui PROPN ajst-12714 125 6 . . PUNCT ajst-12714 126 1 research research NOUN ajst-12714 126 2 on on ADP ajst-12714 126 3 contour contour NOUN ajst-12714 126 4 reconstruction reconstruction NOUN ajst-12714 126 5 technology technology NOUN ajst-12714 126 6 of of ADP ajst-12714 126 7 magnetic magnetic ADJ ajst-12714 126 8 leakage leakage NOUN ajst-12714 126 9 detection detection NOUN ajst-12714 126 10 of of ADP ajst-12714 126 11 pipeline pipeline NOUN ajst-12714 126 12 defects[d defects[d PROPN ajst-12714 126 13 ] ] PUNCT ajst-12714 126 14 . . PUNCT ajst-12714 127 1 northeast northeast ADJ ajst-12714 127 2 petroleum petroleum PROPN ajst-12714 127 3 university university NOUN ajst-12714 127 4 , , PUNCT ajst-12714 127 5 2021 2021 NUM ajst-12714 127 6 . . PUNCT ajst-12714 128 1 [ [ X ajst-12714 128 2 3 3 NUM ajst-12714 128 3 ] ] X ajst-12714 128 4 su su PROPN ajst-12714 128 5 lin lin PROPN ajst-12714 128 6 , , PUNCT ajst-12714 128 7 cheng cheng PROPN ajst-12714 128 8 wenfeng wenfeng PROPN ajst-12714 128 9 , , PUNCT ajst-12714 128 10 xu xu PROPN ajst-12714 128 11 zhijun zhijun PROPN ajst-12714 128 12 , , PUNCT ajst-12714 128 13 et et PROPN ajst-12714 129 1 al al PROPN ajst-12714 129 2 . . PROPN ajst-12714 129 3 finite finite PROPN ajst-12714 129 4 element element PROPN ajst-12714 129 5 simulation simulation NOUN ajst-12714 129 6 of of ADP ajst-12714 129 7 magnetic magnetic ADJ ajst-12714 129 8 leakage leakage NOUN ajst-12714 129 9 detection detection NOUN ajst-12714 129 10 of of ADP ajst-12714 129 11 oil oil NOUN ajst-12714 129 12 and and CCONJ ajst-12714 129 13 gas gas NOUN ajst-12714 129 14 pipeline pipeline NOUN ajst-12714 129 15 defects[j defects[j PROPN ajst-12714 129 16 ] ] PUNCT ajst-12714 129 17 . . PUNCT ajst-12714 130 1 welded weld VERB ajst-12714 130 2 pipe pipe NOUN ajst-12714 130 3 , , PUNCT ajst-12714 130 4 2020 2020 NUM ajst-12714 130 5 , , PUNCT ajst-12714 130 6 43(04 43(04 NUM ajst-12714 130 7 ): ): PUNCT ajst-12714 130 8 8 8 NUM ajst-12714 130 9 - - SYM ajst-12714 130 10 13 13 NUM ajst-12714 130 11 + + NOUN ajst-12714 130 12 22 22 NUM ajst-12714 130 13 . . PUNCT ajst-12714 131 1 [ [ X ajst-12714 131 2 4 4 X ajst-12714 131 3 ] ] PUNCT ajst-12714 131 4 yang yang PROPN ajst-12714 131 5 yang yang PROPN ajst-12714 131 6 . . PUNCT ajst-12714 132 1 magnetic magnetic PROPN ajst-12714 132 2 flux flux PROPN ajst-12714 132 3 leakage leakage NOUN ajst-12714 132 4 detection detection NOUN ajst-12714 132 5 of of ADP ajst-12714 132 6 long long ADJ ajst-12714 132 7 - - PUNCT ajst-12714 132 8 distance distance NOUN ajst-12714 132 9 natural natural ADJ ajst-12714 132 10 gas gas NOUN ajst-12714 132 11 pipeline pipeline NOUN ajst-12714 132 12 defects[d defects[d PROPN ajst-12714 132 13 ] ] PUNCT ajst-12714 132 14 . . PUNCT ajst-12714 133 1 xihua xihua PROPN ajst-12714 133 2 university university PROPN ajst-12714 133 3 , , PUNCT ajst-12714 133 4 2020 2020 NUM ajst-12714 133 5 . . PUNCT ajst-12714 134 1 [ [ X ajst-12714 134 2 5 5 X ajst-12714 134 3 ] ] PUNCT ajst-12714 134 4 chen chen PROPN ajst-12714 134 5 junjie junjie PROPN ajst-12714 134 6 . . PUNCT ajst-12714 135 1 segmentation segmentation NOUN ajst-12714 135 2 identification identification NOUN ajst-12714 135 3 method method NOUN ajst-12714 135 4 of of ADP ajst-12714 135 5 defect defect ADJ ajst-12714 135 6 area area NOUN ajst-12714 135 7 of of ADP ajst-12714 135 8 magnetic magnetic ADJ ajst-12714 135 9 leakage leakage NOUN ajst-12714 135 10 detection detection NOUN ajst-12714 135 11 in in ADP ajst-12714 135 12 oil oil NOUN ajst-12714 135 13 and and CCONJ ajst-12714 135 14 gas gas NOUN ajst-12714 135 15 pipeline[j pipeline[j NOUN ajst-12714 135 16 ] ] PUNCT ajst-12714 135 17 . . PUNCT ajst-12714 136 1 china china PROPN ajst-12714 136 2 test test PROPN ajst-12714 136 3 , , PUNCT ajst-12714 136 4 2017 2017 NUM ajst-12714 136 5 , , PUNCT ajst-12714 136 6 43(11 43(11 NUM ajst-12714 136 7 ): ): PUNCT ajst-12714 136 8 1 1 NUM ajst-12714 136 9 - - SYM ajst-12714 136 10 7 7 NUM ajst-12714 136 11 . . PUNCT ajst-12714 137 1 [ [ X ajst-12714 137 2 6 6 NUM ajst-12714 137 3 ] ] PUNCT ajst-12714 137 4 chen chen PROPN ajst-12714 137 5 junjie junjie PROPN ajst-12714 137 6 , , PUNCT ajst-12714 137 7 huang huang PROPN ajst-12714 137 8 songling songling PROPN ajst-12714 137 9 , , PUNCT ajst-12714 137 10 zhao zhao PROPN ajst-12714 137 11 wei wei PROPN ajst-12714 137 12 . . PUNCT ajst-12714 138 1 research research NOUN ajst-12714 138 2 on on ADP ajst-12714 138 3 data datum NOUN ajst-12714 138 4 compression compression NOUN ajst-12714 138 5 algorithm algorithm NOUN ajst-12714 138 6 for for ADP ajst-12714 138 7 magnetic magnetic ADJ ajst-12714 138 8 leakage leakage NOUN ajst-12714 138 9 detection detection NOUN ajst-12714 138 10 of of ADP ajst-12714 138 11 oil oil NOUN ajst-12714 138 12 and and CCONJ ajst-12714 138 13 gas gas NOUN ajst-12714 138 14 pipeline pipeline NOUN ajst-12714 138 15 defects[j].electrical defects[j].electrical ADJ ajst-12714 138 16 measurement measurement NOUN ajst-12714 138 17 and and CCONJ ajst-12714 138 18 instrumentation instrumentation NOUN ajst-12714 138 19 , , PUNCT ajst-12714 138 20 2014 2014 NUM ajst-12714 138 21 , , PUNCT ajst-12714 138 22 51(15 51(15 NUM ajst-12714 138 23 ): ): PUNCT ajst-12714 138 24 100 100 NUM ajst-12714 138 25 - - SYM ajst-12714 138 26 104 104 NUM ajst-12714 138 27 . . PUNCT ajst-12714 139 1 [ [ X ajst-12714 139 2 7 7 X ajst-12714 139 3 ] ] SYM ajst-12714 139 4 min min PROPN ajst-12714 139 5 zhang zhang PROPN ajst-12714 139 6 , , PUNCT ajst-12714 139 7 yanbao yanbao PROPN ajst-12714 139 8 guo guo PROPN ajst-12714 139 9 , , PUNCT ajst-12714 139 10 qiuju qiuju PROPN ajst-12714 139 11 xie xie PROPN ajst-12714 139 12 , , PUNCT ajst-12714 139 13 yuansheng yuansheng PROPN ajst-12714 139 14 zhang zhang PROPN ajst-12714 139 15 , , PUNCT ajst-12714 139 16 et et PROPN ajst-12714 139 17 al al PROPN ajst-12714 139 18 . . PUNCT ajst-12714 139 19 defect defect VERB ajst-12714 139 20 identification identification NOUN ajst-12714 139 21 for for ADP ajst-12714 139 22 oil oil NOUN ajst-12714 139 23 and and CCONJ ajst-12714 139 24 gas gas NOUN ajst-12714 139 25 pipeline pipeline NOUN ajst-12714 139 26 safety safety NOUN ajst-12714 139 27 based base VERB ajst-12714 139 28 on on ADP ajst-12714 139 29 autonomous autonomous ADJ ajst-12714 139 30 deep deep ADJ ajst-12714 139 31 learning learning NOUN ajst-12714 139 32 network[j network[j PROPN ajst-12714 139 33 ] ] PUNCT ajst-12714 139 34 . . PUNCT ajst-12714 140 1 computer computer NOUN ajst-12714 140 2 communications communication NOUN ajst-12714 140 3 , , PUNCT ajst-12714 140 4 2022 2022 NUM ajst-12714 140 5 , , PUNCT ajst-12714 140 6 195 195 NUM ajst-12714 140 7 : : SYM ajst-12714 140 8 14 14 NUM ajst-12714 140 9 - - SYM ajst-12714 140 10 26 26 NUM ajst-12714 140 11 . . PUNCT ajst-12714 141 1 [ [ X ajst-12714 141 2 8 8 NUM ajst-12714 141 3 ] ] PUNCT ajst-12714 141 4 a. a. NOUN ajst-12714 141 5 carvalhoa carvalhoa NOUN ajst-12714 141 6 , , PUNCT ajst-12714 141 7 j.m.a j.m.a NOUN ajst-12714 141 8 . . PUNCT ajst-12714 141 9 rebelloa rebelloa NOUN ajst-12714 141 10 , , PUNCT ajst-12714 141 11 l.v.s.sagrilob l.v.s.sagrilob PROPN ajst-12714 141 12 , , PUNCT ajst-12714 141 13 et et PROPN ajst-12714 141 14 al al PROPN ajst-12714 141 15 . . PUNCT ajst-12714 141 16 mfl mfl PROPN ajst-12714 141 17 signals signal NOUN ajst-12714 141 18 and and CCONJ ajst-12714 141 19 artificial artificial ADJ ajst-12714 141 20 neural neural ADJ ajst-12714 141 21 networks network NOUN ajst-12714 141 22 applied apply VERB ajst-12714 141 23 to to ADP ajst-12714 141 24 detection detection NOUN ajst-12714 141 25 and and CCONJ ajst-12714 141 26 classification classification NOUN ajst-12714 141 27 of of ADP ajst-12714 141 28 pipe pipe NOUN ajst-12714 141 29 weld weld NOUN ajst-12714 141 30 defects[j defects[j PROPN ajst-12714 141 31 ] ] PUNCT ajst-12714 141 32 . . PUNCT ajst-12714 142 1 ndt$e ndt$e PROPN ajst-12714 142 2 internationa1 internationa1 PROPN ajst-12714 142 3 , , PUNCT ajst-12714 142 4 2006 2006 NUM ajst-12714 142 5 , , PUNCT ajst-12714 142 6 39 39 NUM ajst-12714 142 7 : : PUNCT ajst-12714 142 8 661 661 NUM ajst-12714 142 9 - - SYM ajst-12714 142 10 667 667 NUM ajst-12714 142 11 . . PUNCT ajst-12714 143 1 [ [ X ajst-12714 143 2 9 9 X ajst-12714 143 3 ] ] X ajst-12714 143 4 zheng zheng PROPN ajst-12714 143 5 biaohua biaohua PROPN ajst-12714 143 6 , , PUNCT ajst-12714 143 7 he he PRON ajst-12714 143 8 wen wen PROPN ajst-12714 143 9 , , PUNCT ajst-12714 143 10 zhou zhou PROPN ajst-12714 143 11 songqiang songqiang PROPN ajst-12714 143 12 , , PUNCT ajst-12714 143 13 zhang zhang PROPN ajst-12714 143 14 chunlei chunlei PROPN ajst-12714 143 15 , , PUNCT ajst-12714 143 16 gao gao PROPN ajst-12714 143 17 zhong zhong PROPN ajst-12714 143 18 . . PUNCT ajst-12714 144 1 3d 3d PROPN ajst-12714 144 2 finite finite PROPN ajst-12714 144 3 element element PROPN ajst-12714 144 4 simulation simulation NOUN ajst-12714 144 5 analysis analysis NOUN ajst-12714 144 6 of of ADP ajst-12714 144 7 magnetic magnetic ADJ ajst-12714 144 8 leakage leakage NOUN ajst-12714 144 9 detection detection NOUN ajst-12714 144 10 of of ADP ajst-12714 144 11 pipeline pipeline NOUN ajst-12714 144 12 defects[j defects[j PROPN ajst-12714 144 13 ] ] PUNCT ajst-12714 144 14 . . PUNCT ajst-12714 145 1 china china PROPN ajst-12714 145 2 safety safety PROPN ajst-12714 145 3 science science PROPN ajst-12714 145 4 journal journal PROPN ajst-12714 145 5 , , PUNCT ajst-12714 145 6 2013 2013 NUM ajst-12714 145 7 , , PUNCT ajst-12714 145 8 23(12 23(12 NUM ajst-12714 145 9 ): ): PUNCT ajst-12714 145 10 35 35 NUM ajst-12714 145 11 - - SYM ajst-12714 145 12 41 41 NUM ajst-12714 145 13 . . PUNCT ajst-12714 146 1 [ [ X ajst-12714 146 2 10 10 NUM ajst-12714 146 3 ] ] PUNCT ajst-12714 146 4 ding ding NOUN ajst-12714 146 5 zhanwu zhanwu NOUN ajst-12714 146 6 , , PUNCT ajst-12714 146 7 he he PRON ajst-12714 146 8 renyang renyang PROPN ajst-12714 146 9 , , PUNCT ajst-12714 146 10 liu liu PROPN ajst-12714 146 11 zhong zhong PROPN ajst-12714 146 12 . . PUNCT ajst-12714 147 1 simulation simulation NOUN ajst-12714 147 2 analysis analysis NOUN ajst-12714 147 3 and and CCONJ ajst-12714 147 4 quantification quantification NOUN ajst-12714 147 5 model model NOUN ajst-12714 147 6 of of ADP ajst-12714 147 7 pipeline pipeline PROPN ajst-12714 147 8 magnetic magnetic ADJ ajst-12714 147 9 leakage leakage NOUN ajst-12714 147 10 detection detection NOUN ajst-12714 147 11 defect defect VERB ajst-12714 147 12 signal[j signal[j NOUN ajst-12714 147 13 ] ] PUNCT ajst-12714 147 14 . . PUNCT ajst-12714 148 1 nondestructive nondestructive ADJ ajst-12714 148 2 testing testing NOUN ajst-12714 148 3 , , PUNCT ajst-12714 148 4 2013 2013 NUM ajst-12714 148 5 , , PUNCT ajst-12714 148 6 35(03 35(03 NUM ajst-12714 148 7 ): ): PUNCT ajst-12714 148 8 30 30 NUM ajst-12714 148 9 - - SYM ajst-12714 148 10 33 33 NUM ajst-12714 148 11 . . PUNCT ajst-12714 149 1 [ [ X ajst-12714 149 2 11 11 NUM ajst-12714 149 3 ] ] X ajst-12714 149 4 wu wu PROPN ajst-12714 149 5 zhenning zhenning PROPN ajst-12714 149 6 , , PUNCT ajst-12714 149 7 wang wang PROPN ajst-12714 149 8 lixing lixing PROPN ajst-12714 149 9 , , PUNCT ajst-12714 149 10 liu liu PROPN ajst-12714 149 11 jinhai jinhai PROPN ajst-12714 149 12 . . PUNCT ajst-12714 150 1 reconstruction reconstruction NOUN ajst-12714 150 2 method method NOUN ajst-12714 150 3 for for ADP ajst-12714 150 4 complex complex ADJ ajst-12714 150 5 defects defect NOUN ajst-12714 150 6 of of ADP ajst-12714 150 7 magnetic magnetic ADJ ajst-12714 150 8 field field NOUN ajst-12714 150 9 leakage leakage NOUN ajst-12714 150 10 detection detection NOUN ajst-12714 150 11 based base VERB ajst-12714 150 12 on on ADP ajst-12714 150 13 spatial spatial ADJ ajst-12714 150 14 mapping mapping NOUN ajst-12714 150 15 for for ADP ajst-12714 150 16 uniform uniform ADJ ajst-12714 150 17 speed speed NOUN ajst-12714 150 18 sampling sample VERB ajst-12714 150 19 [ [ X ajst-12714 150 20 j j X ajst-12714 150 21 ] ] X ajst-12714 150 22 . . PUNCT ajst-12714 151 1 chinese chinese ADJ ajst-12714 151 2 journal journal PROPN ajst-12714 151 3 of of ADP ajst-12714 151 4 scientific scientific ADJ ajst-12714 151 5 instrument instrument NOUN ajst-12714 151 6 , , PUNCT ajst-12714 151 7 2018 2018 NUM ajst-12714 151 8 , , PUNCT ajst-12714 151 9 039(007 039(007 NUM ajst-12714 151 10 ): ): PUNCT ajst-12714 151 11 164 164 NUM ajst-12714 151 12 - - SYM ajst-12714 151 13 172 172 NUM ajst-12714 151 14 . . PUNCT ajst-12714 152 1 [ [ X ajst-12714 152 2 12 12 NUM ajst-12714 152 3 ] ] X ajst-12714 152 4 jiao jiao PROPN ajst-12714 152 5 jingpin jingpin NOUN ajst-12714 152 6 , , PUNCT ajst-12714 152 7 chang chang PROPN ajst-12714 152 8 yu yu PROPN ajst-12714 152 9 , , PUNCT ajst-12714 152 10 li li PROPN ajst-12714 152 11 guanghai guanghai PROPN ajst-12714 152 12 , , PUNCT ajst-12714 152 13 he he PRON ajst-12714 152 14 cunfu cunfu VERB ajst-12714 152 15 , , PUNCT ajst-12714 152 16 wu wu PROPN ajst-12714 152 17 bin bin PROPN ajst-12714 152 18 . . PUNCT ajst-12714 153 1 research research NOUN ajst-12714 153 2 on on ADP ajst-12714 153 3 low low ADJ ajst-12714 153 4 - - PUNCT ajst-12714 153 5 frequency frequency NOUN ajst-12714 153 6 magnetic magnetic ADJ ajst-12714 153 7 leakage leakage NOUN ajst-12714 153 8 detection detection NOUN ajst-12714 153 9 technology technology NOUN ajst-12714 153 10 for for ADP ajst-12714 153 11 internal internal ADJ ajst-12714 153 12 and and CCONJ ajst-12714 153 13 external external ADJ ajst-12714 153 14 surface surface NOUN ajst-12714 153 15 cracks crack NOUN ajst-12714 153 16 of of ADP ajst-12714 153 17 ferromagnetic ferromagnetic ADJ ajst-12714 153 18 components[j components[j PROPN ajst-12714 153 19 ] ] X ajst-12714 153 20 . . PUNCT ajst-12714 154 1 chinese chinese ADJ ajst-12714 154 2 journal journal PROPN ajst-12714 154 3 of of ADP ajst-12714 154 4 scientific scientific ADJ ajst-12714 154 5 instrument instrument NOUN ajst-12714 154 6 , , PUNCT ajst-12714 154 7 2016 2016 NUM ajst-12714 154 8 , , PUNCT ajst-12714 154 9 37(08 37(08 NUM ajst-12714 154 10 ): ): PUNCT ajst-12714 154 11 1808 1808 NUM ajst-12714 154 12 - - SYM ajst-12714 154 13 1817 1817 NUM ajst-12714 154 14 . . PUNCT ajst-12714 155 1 [ [ X ajst-12714 155 2 13 13 NUM ajst-12714 155 3 ] ] X ajst-12714 155 4 guo guo PROPN ajst-12714 155 5 zijian zijian PROPN ajst-12714 155 6 , , PUNCT ajst-12714 155 7 ran run VERB ajst-12714 155 8 lin lin PROPN ajst-12714 155 9 . . PUNCT ajst-12714 156 1 defect defect VERB ajst-12714 156 2 identification identification NOUN ajst-12714 156 3 method method NOUN ajst-12714 156 4 for for ADP ajst-12714 156 5 magnetic magnetic ADJ ajst-12714 156 6 flux flux NOUN ajst-12714 156 7 leakage leakage NOUN ajst-12714 156 8 detection detection NOUN ajst-12714 156 9 of of ADP ajst-12714 156 10 ansys ansys PROPN ajst-12714 156 11 urban urban ADJ ajst-12714 156 12 natural natural ADJ ajst-12714 156 13 gas gas NOUN ajst-12714 156 14 pipelines[j pipelines[j PROPN ajst-12714 156 15 ] ] PUNCT ajst-12714 156 16 . . PUNCT ajst-12714 157 1 neijiang neijiang PROPN ajst-12714 157 2 science science PROPN ajst-12714 157 3 and and CCONJ ajst-12714 157 4 technology technology NOUN ajst-12714 157 5 , , PUNCT ajst-12714 157 6 2011 2011 NUM ajst-12714 157 7 , , PUNCT ajst-12714 157 8 32(03 32(03 NUM ajst-12714 157 9 ): ): PUNCT ajst-12714 157 10 115 115 NUM ajst-12714 157 11 + + SYM ajst-12714 157 12 37 37 NUM ajst-12714 157 13 . . PUNCT ajst-12714 158 1 [ [ X ajst-12714 158 2 14 14 NUM ajst-12714 158 3 ] ] X ajst-12714 158 4 anitha anitha PROPN ajst-12714 158 5 , , PUNCT ajst-12714 158 6 r r PROPN ajst-12714 158 7 , , PUNCT ajst-12714 158 8 renuka renuka PROPN ajst-12714 158 9 , , PUNCT ajst-12714 158 10 s s PROPN ajst-12714 158 11 , , PUNCT ajst-12714 158 12 abudhahir abudhahir ADJ ajst-12714 158 13 , , PUNCT ajst-12714 158 14 a. a. NOUN ajst-12714 158 15 multi multi PROPN ajst-12714 158 16 sensor sensor PROPN ajst-12714 158 17 data datum NOUN ajst-12714 158 18 fusion fusion NOUN ajst-12714 158 19 algorithms algorithm NOUN ajst-12714 158 20 for for ADP ajst-12714 158 21 target target NOUN ajst-12714 158 22 tracking tracking NOUN ajst-12714 158 23 using use VERB ajst-12714 158 24 multiple multiple ADJ ajst-12714 158 25 measurements[c]//ieee measurements[c]//ieee PROPN ajst-12714 158 26 international international ADJ ajst-12714 158 27 conference conference NOUN ajst-12714 158 28 on on ADP ajst-12714 158 29 computational computational ADJ ajst-12714 158 30 intelligence intelligence NOUN ajst-12714 158 31 & & CCONJ ajst-12714 158 32 computing computing PROPN ajst-12714 158 33 research research NOUN ajst-12714 158 34 . . PUNCT ajst-12714 159 1 ieee ieee NOUN ajst-12714 159 2 , , PUNCT ajst-12714 159 3 2017 2017 NUM ajst-12714 159 4 : : PUNCT ajst-12714 159 5 978 978 NUM ajst-12714 159 6 - - SYM ajst-12714 159 7 1 1 NUM ajst-12714 159 8 - - PUNCT ajst-12714 159 9 4799 4799 NUM ajst-12714 159 10 - - PUNCT ajst-12714 159 11 1597 1597 NUM ajst-12714 159 12 - - SYM ajst-12714 159 13 2 2 NUM ajst-12714 159 14 . . PUNCT ajst-12714 160 1 [ [ X ajst-12714 160 2 15 15 NUM ajst-12714 160 3 ] ] X ajst-12714 160 4 baskaran baskaran ADJ ajst-12714 160 5 n n PROPN ajst-12714 160 6 s s PROPN ajst-12714 160 7 m. m. NOUN ajst-12714 160 8 analysis analysis NOUN ajst-12714 160 9 of of ADP ajst-12714 160 10 finite finite ADJ ajst-12714 160 11 element element NOUN ajst-12714 160 12 modeling modeling NOUN ajst-12714 160 13 of of ADP ajst-12714 160 14 magnetic magnetic ADJ ajst-12714 160 15 flux flux NOUN ajst-12714 160 16 leakage leakage NOUN ajst-12714 160 17 technique technique NOUN ajst-12714 160 18 in in ADP ajst-12714 160 19 plates plate NOUN ajst-12714 160 20 with with ADP ajst-12714 160 21 defect[j defect[j PROPN ajst-12714 160 22 ] ] PUNCT ajst-12714 160 23 . . PUNCT ajst-12714 161 1 international international ADJ ajst-12714 161 2 journal journal PROPN ajst-12714 161 3 of of ADP ajst-12714 161 4 engineering engineering PROPN ajst-12714 161 5 sciences sciences PROPN ajst-12714 161 6 & & CCONJ ajst-12714 161 7 research research PROPN ajst-12714 161 8 technology technology NOUN ajst-12714 161 9 , , PUNCT ajst-12714 161 10 2014 2014 NUM ajst-12714 161 11 , , PUNCT ajst-12714 161 12 3 3 NUM ajst-12714 161 13 ( ( PUNCT ajst-12714 161 14 2 2 NUM ajst-12714 161 15 ): ): PUNCT ajst-12714 161 16 997 997 NUM ajst-12714 161 17 - - SYM ajst-12714 161 18 1000 1000 NUM ajst-12714 161 19 . . PUNCT ajst-12714 162 1 [ [ X ajst-12714 162 2 16 16 NUM ajst-12714 162 3 ] ] X ajst-12714 162 4 amineh amineh PROPN ajst-12714 162 5 r r PROPN ajst-12714 162 6 k k PROPN ajst-12714 162 7 , , PUNCT ajst-12714 162 8 ravan ravan NOUN ajst-12714 162 9 m m PROPN ajst-12714 162 10 , , PUNCT ajst-12714 162 11 sadeghi sadeghi PROPN ajst-12714 162 12 s s VERB ajst-12714 162 13 h h NOUN ajst-12714 162 14 h h NOUN ajst-12714 162 15 , , PUNCT ajst-12714 162 16 et et PROPN ajst-12714 162 17 al al PROPN ajst-12714 162 18 . . PUNCT ajst-12714 162 19 using use VERB ajst-12714 162 20 ac ac PROPN ajst-12714 162 21 field field PROPN ajst-12714 162 22 measurement measurement NOUN ajst-12714 162 23 data datum NOUN ajst-12714 162 24 at at ADP ajst-12714 162 25 an an DET ajst-12714 162 26 arbitrary arbitrary ADJ ajst-12714 162 27 liftoff liftoff NOUN ajst-12714 162 28 distance distance NOUN ajst-12714 162 29 to to PART ajst-12714 162 30 size size VERB ajst-12714 162 31 long long ADJ ajst-12714 162 32 surface surface NOUN ajst-12714 162 33 - - PUNCT ajst-12714 162 34 breaking break VERB ajst-12714 162 35 cracks crack NOUN ajst-12714 162 36 in in ADP ajst-12714 162 37 ferrous ferrous ADJ ajst-12714 162 38 metals[j metals[j NOUN ajst-12714 162 39 ] ] X ajst-12714 162 40 . . PUNCT ajst-12714 163 1 ndt&e ndt&e PROPN ajst-12714 163 2 international international ADJ ajst-12714 163 3 , , PUNCT ajst-12714 163 4 2008 2008 NUM ajst-12714 163 5 , , PUNCT ajst-12714 163 6 41(3 41(3 NUM ajst-12714 163 7 ): ): PUNCT ajst-12714 163 8 169 169 NUM ajst-12714 163 9 - - SYM ajst-12714 163 10 177 177 NUM ajst-12714 163 11 . . PUNCT ajst-12714 164 1 [ [ X ajst-12714 164 2 17 17 NUM ajst-12714 164 3 ] ] X ajst-12714 164 4 yang yang PROPN ajst-12714 164 5 lijian lijian PROPN ajst-12714 164 6 , , PUNCT ajst-12714 164 7 yu yu PROPN ajst-12714 164 8 wenlai wenlai PROPN ajst-12714 164 9 , , PUNCT ajst-12714 164 10 gao gao PROPN ajst-12714 164 11 songwei songwei PROPN ajst-12714 164 12 , , PUNCT ajst-12714 164 13 et et PROPN ajst-12714 164 14 al al PROPN ajst-12714 164 15 . . PROPN ajst-12714 164 16 defect defect VERB ajst-12714 164 17 identification identification NOUN ajst-12714 164 18 technology technology NOUN ajst-12714 164 19 of of ADP ajst-12714 164 20 pipeline pipeline PROPN ajst-12714 164 21 magnetic magnetic PROPN ajst-12714 164 22 flux flux PROPN ajst-12714 164 23 leakage leakage PROPN ajst-12714 164 24 detection[j detection[j PROPN ajst-12714 164 25 ] ] PUNCT ajst-12714 164 26 . . PUNCT ajst-12714 165 1 journal journal PROPN ajst-12714 165 2 of of ADP ajst-12714 165 3 shenyang shenyang PROPN ajst-12714 165 4 university university PROPN ajst-12714 165 5 of of ADP ajst-12714 165 6 technology technology NOUN ajst-12714 165 7 , , PUNCT ajst-12714 165 8 2010 2010 NUM ajst-12714 165 9 , , PUNCT ajst-12714 165 10 32(01 32(01 NUM ajst-12714 165 11 ): ): PUNCT ajst-12714 165 12 6569 6569 NUM ajst-12714 165 13 . . PUNCT ajst-12714 166 1 [ [ X ajst-12714 166 2 18 18 NUM ajst-12714 166 3 ] ] X ajst-12714 166 4 yang yang PROPN ajst-12714 166 5 lijian lijian PROPN ajst-12714 166 6 , , PUNCT ajst-12714 166 7 bi bi PROPN ajst-12714 166 8 dawei dawei PROPN ajst-12714 166 9 , , PUNCT ajst-12714 166 10 gao gao PROPN ajst-12714 166 11 songwei songwei PROPN ajst-12714 166 12 . . PUNCT ajst-12714 167 1 research research NOUN ajst-12714 167 2 on on ADP ajst-12714 167 3 defect defect ADJ ajst-12714 167 4 quantification quantification NOUN ajst-12714 167 5 technology technology NOUN ajst-12714 167 6 for for ADP ajst-12714 167 7 magnetic magnetic ADJ ajst-12714 167 8 leakage leakage NOUN ajst-12714 167 9 detection detection NOUN ajst-12714 167 10 of of ADP ajst-12714 167 11 oil oil NOUN ajst-12714 167 12 and and CCONJ ajst-12714 167 13 gas gas NOUN ajst-12714 167 14 pipelines[j pipelines[j PROPN ajst-12714 167 15 ] ] PUNCT ajst-12714 167 16 . . PUNCT ajst-12714 168 1 computer computer NOUN ajst-12714 168 2 measurement measurement NOUN ajst-12714 168 3 and and CCONJ ajst-12714 168 4 control control NOUN ajst-12714 168 5 , , PUNCT ajst-12714 168 6 2009 2009 NUM ajst-12714 168 7 , , PUNCT ajst-12714 168 8 17(08 17(08 NUM ajst-12714 168 9 ): ): PUNCT ajst-12714 168 10 1489 1489 NUM ajst-12714 168 11 - - SYM ajst-12714 168 12 1491 1491 NUM ajst-12714 168 13 . . PUNCT ajst-12714 169 1 [ [ X ajst-12714 169 2 19 19 NUM ajst-12714 169 3 ] ] X ajst-12714 169 4 mehdi mehdi NOUN ajst-12714 169 5 pourahmadi pourahmadi PROPN ajst-12714 169 6 , , PUNCT ajst-12714 169 7 mesbah mesbah PROPN ajst-12714 169 8 saybani saybani PROPN ajst-12714 169 9 . . PUNCT ajst-12714 170 1 reliability reliability NOUN ajst-12714 170 2 analysis analysis NOUN ajst-12714 170 3 with with ADP ajst-12714 170 4 corrosion corrosion NOUN ajst-12714 170 5 defects defect NOUN ajst-12714 170 6 in in ADP ajst-12714 170 7 submarine submarine NOUN ajst-12714 170 8 pipeline pipeline NOUN ajst-12714 170 9 case case NOUN ajst-12714 170 10 study study NOUN ajst-12714 170 11 : : PUNCT ajst-12714 170 12 oil oil NOUN ajst-12714 170 13 pipeline pipeline NOUN ajst-12714 170 14 in in ADP ajst-12714 170 15 ab ab PROPN ajst-12714 170 16 - - PUNCT ajst-12714 170 17 khark khark ADJ ajst-12714 170 18 island[j island[j NOUN ajst-12714 170 19 ] ] PUNCT ajst-12714 170 20 . . PUNCT ajst-12714 171 1 ocean ocean PROPN ajst-12714 171 2 engineering engineering PROPN ajst-12714 171 3 , , PUNCT ajst-12714 171 4 2022 2022 NUM ajst-12714 171 5 , , PUNCT ajst-12714 171 6 249 249 NUM ajst-12714 171 7 : : PUNCT ajst-12714 171 8 110 110 NUM ajst-12714 171 9 - - SYM ajst-12714 171 10 885 885 NUM ajst-12714 171 11 . . PUNCT ajst-12714 172 1 [ [ X ajst-12714 172 2 20 20 NUM ajst-12714 172 3 ] ] PUNCT ajst-12714 172 4 shuai shuai PROPN ajst-12714 172 5 hao hao PROPN ajst-12714 172 6 , , PUNCT ajst-12714 172 7 pengpeng pengpeng PROPN ajst-12714 172 8 shi shi PROPN ajst-12714 172 9 , , PUNCT ajst-12714 172 10 sanqing sanqe VERB ajst-12714 172 11 su su PROPN ajst-12714 172 12 , , PUNCT ajst-12714 172 13 et et PROPN ajst-12714 172 14 al al PROPN ajst-12714 172 15 . . PUNCT ajst-12714 172 16 evaluation evaluation NOUN ajst-12714 172 17 of of ADP ajst-12714 172 18 defect defect ADJ ajst-12714 172 19 depth depth NOUN ajst-12714 172 20 in in ADP ajst-12714 172 21 ferromagnetic ferromagnetic ADJ ajst-12714 172 22 materials material NOUN ajst-12714 172 23 via via ADP ajst-12714 172 24 magnetic magnetic ADJ ajst-12714 172 25 flux flux PROPN ajst-12714 172 26 leakage leakage NOUN ajst-12714 172 27 method method NOUN ajst-12714 172 28 with with ADP ajst-12714 172 29 a a DET ajst-12714 172 30 double double ADJ ajst-12714 172 31 hall hall NOUN ajst-12714 172 32 sensor[j sensor[j NOUN ajst-12714 172 33 ] ] PUNCT ajst-12714 172 34 . . PUNCT ajst-12714 173 1 journal journal PROPN ajst-12714 173 2 of of ADP ajst-12714 173 3 magnetism magnetism NOUN ajst-12714 173 4 and and CCONJ ajst-12714 173 5 magnetic magnetic ADJ ajst-12714 173 6 materials material NOUN ajst-12714 173 7 , , PUNCT ajst-12714 173 8 2022 2022 NUM ajst-12714 173 9 , , PUNCT ajst-12714 173 10 555 555 NUM ajst-12714 173 11 : : SYM ajst-12714 173 12 169 169 NUM ajst-12714 173 13 - - SYM ajst-12714 173 14 341 341 NUM ajst-12714 173 15 . . PUNCT