id sid tid token lemma pos fcis-5218 1 1 frontiers frontier NOUN fcis-5218 1 2 in in ADP fcis-5218 1 3 computing computing NOUN fcis-5218 1 4 and and CCONJ fcis-5218 1 5 intelligent intelligent ADJ fcis-5218 1 6 systems system NOUN fcis-5218 1 7 issn issn VERB fcis-5218 1 8 : : PUNCT fcis-5218 1 9 2832 2832 NUM fcis-5218 1 10 - - SYM fcis-5218 1 11 6024 6024 NUM fcis-5218 1 12 | | NOUN fcis-5218 1 13 vol vol NOUN fcis-5218 1 14 . . PROPN fcis-5218 2 1 2 2 NUM fcis-5218 2 2 , , PUNCT fcis-5218 2 3 no no INTJ fcis-5218 2 4 . . NOUN fcis-5218 2 5 3 3 NUM fcis-5218 2 6 , , PUNCT fcis-5218 2 7 2022 2022 NUM fcis-5218 2 8 75 75 NUM fcis-5218 2 9 identification identification NOUN fcis-5218 2 10 of of ADP fcis-5218 2 11 tea tea NOUN fcis-5218 2 12 leaf leaf NOUN fcis-5218 2 13 diseases disease NOUN fcis-5218 2 14 based base VERB fcis-5218 2 15 on on ADP fcis-5218 2 16 deep deep ADJ fcis-5218 2 17 transfer transfer NOUN fcis-5218 2 18 learning learning PROPN fcis-5218 2 19 wei wei PROPN fcis-5218 2 20 wu wu PROPN fcis-5218 2 21 taishan taishan PROPN fcis-5218 2 22 university university PROPN fcis-5218 2 23 , , PUNCT fcis-5218 2 24 tai'an tai'an PROPN fcis-5218 2 25 271000 271000 NUM fcis-5218 2 26 , , PUNCT fcis-5218 2 27 china china PROPN fcis-5218 2 28 abstract abstract NOUN fcis-5218 2 29 : : PUNCT fcis-5218 2 30 tea tea NOUN fcis-5218 2 31 trees tree NOUN fcis-5218 2 32 are be AUX fcis-5218 2 33 often often ADV fcis-5218 2 34 vulnerable vulnerable ADJ fcis-5218 2 35 to to ADP fcis-5218 2 36 diseases disease NOUN fcis-5218 2 37 and and CCONJ fcis-5218 2 38 insect insect VERB fcis-5218 2 39 pests pest NOUN fcis-5218 2 40 in in ADP fcis-5218 2 41 the the DET fcis-5218 2 42 process process NOUN fcis-5218 2 43 of of ADP fcis-5218 2 44 growth growth NOUN fcis-5218 2 45 , , PUNCT fcis-5218 2 46 resulting result VERB fcis-5218 2 47 in in ADP fcis-5218 2 48 the the DET fcis-5218 2 49 decline decline NOUN fcis-5218 2 50 of of ADP fcis-5218 2 51 tea tea NOUN fcis-5218 2 52 production production NOUN fcis-5218 2 53 and and CCONJ fcis-5218 2 54 quality quality NOUN fcis-5218 2 55 . . PUNCT fcis-5218 3 1 it it PRON fcis-5218 3 2 is be AUX fcis-5218 3 3 of of ADP fcis-5218 3 4 great great ADJ fcis-5218 3 5 significance significance NOUN fcis-5218 3 6 to to PART fcis-5218 3 7 identify identify VERB fcis-5218 3 8 and and CCONJ fcis-5218 3 9 prevent prevent VERB fcis-5218 3 10 tea tea NOUN fcis-5218 3 11 leaf leaf NOUN fcis-5218 3 12 diseases disease NOUN fcis-5218 3 13 in in ADP fcis-5218 3 14 time time NOUN fcis-5218 3 15 to to PART fcis-5218 3 16 ensure ensure VERB fcis-5218 3 17 the the DET fcis-5218 3 18 steady steady ADJ fcis-5218 3 19 increase increase NOUN fcis-5218 3 20 of of ADP fcis-5218 3 21 tea tea NOUN fcis-5218 3 22 . . PUNCT fcis-5218 4 1 this this DET fcis-5218 4 2 study study NOUN fcis-5218 4 3 proposes propose VERB fcis-5218 4 4 a a DET fcis-5218 4 5 tea tea NOUN fcis-5218 4 6 leaf leaf NOUN fcis-5218 4 7 disease disease NOUN fcis-5218 4 8 identification identification NOUN fcis-5218 4 9 method method NOUN fcis-5218 4 10 based base VERB fcis-5218 4 11 on on ADP fcis-5218 4 12 deep deep ADJ fcis-5218 4 13 transfer transfer NOUN fcis-5218 4 14 learning learning NOUN fcis-5218 4 15 , , PUNCT fcis-5218 4 16 which which PRON fcis-5218 4 17 improves improve VERB fcis-5218 4 18 the the DET fcis-5218 4 19 recognition recognition NOUN fcis-5218 4 20 accuracy accuracy NOUN fcis-5218 4 21 of of ADP fcis-5218 4 22 the the DET fcis-5218 4 23 model model NOUN fcis-5218 4 24 through through ADP fcis-5218 4 25 knowledge knowledge NOUN fcis-5218 4 26 transfer transfer NOUN fcis-5218 4 27 . . PUNCT fcis-5218 5 1 besides besides SCONJ fcis-5218 5 2 , , PUNCT fcis-5218 5 3 for for ADP fcis-5218 5 4 the the DET fcis-5218 5 5 unbalanced unbalanced ADJ fcis-5218 5 6 distribution distribution NOUN fcis-5218 5 7 of of ADP fcis-5218 5 8 the the DET fcis-5218 5 9 number number NOUN fcis-5218 5 10 of of ADP fcis-5218 5 11 samples sample NOUN fcis-5218 5 12 , , PUNCT fcis-5218 5 13 the the DET fcis-5218 5 14 crossentropy crossentropy NOUN fcis-5218 5 15 loss loss NOUN fcis-5218 5 16 function function NOUN fcis-5218 5 17 is be AUX fcis-5218 5 18 replaced replace VERB fcis-5218 5 19 with with ADP fcis-5218 5 20 the the DET fcis-5218 5 21 focal focal ADJ fcis-5218 5 22 loss loss NOUN fcis-5218 5 23 function function NOUN fcis-5218 5 24 , , PUNCT fcis-5218 5 25 which which PRON fcis-5218 5 26 further far ADV fcis-5218 5 27 improves improve VERB fcis-5218 5 28 the the DET fcis-5218 5 29 identification identification NOUN fcis-5218 5 30 effect effect NOUN fcis-5218 5 31 of of ADP fcis-5218 5 32 the the DET fcis-5218 5 33 model model NOUN fcis-5218 5 34 . . PUNCT fcis-5218 6 1 the the DET fcis-5218 6 2 experiment experiment NOUN fcis-5218 6 3 shows show VERB fcis-5218 6 4 that that SCONJ fcis-5218 6 5 the the DET fcis-5218 6 6 identification identification NOUN fcis-5218 6 7 model model NOUN fcis-5218 6 8 of of ADP fcis-5218 6 9 tea tea NOUN fcis-5218 6 10 leaf leaf NOUN fcis-5218 6 11 disease disease NOUN fcis-5218 6 12 proposed propose VERB fcis-5218 6 13 in in ADP fcis-5218 6 14 this this DET fcis-5218 6 15 study study NOUN fcis-5218 6 16 can can AUX fcis-5218 6 17 achieve achieve VERB fcis-5218 6 18 the the DET fcis-5218 6 19 accuracy accuracy NOUN fcis-5218 6 20 of of ADP fcis-5218 6 21 more more ADJ fcis-5218 6 22 than than ADP fcis-5218 6 23 90.42 90.42 NUM fcis-5218 6 24 % % NOUN fcis-5218 6 25 , , PUNCT fcis-5218 6 26 which which PRON fcis-5218 6 27 verifies verify VERB fcis-5218 6 28 the the DET fcis-5218 6 29 effectiveness effectiveness NOUN fcis-5218 6 30 of of ADP fcis-5218 6 31 this this DET fcis-5218 6 32 research research NOUN fcis-5218 6 33 and and CCONJ fcis-5218 6 34 has have VERB fcis-5218 6 35 important important ADJ fcis-5218 6 36 theoretical theoretical ADJ fcis-5218 6 37 and and CCONJ fcis-5218 6 38 practical practical ADJ fcis-5218 6 39 significance significance NOUN fcis-5218 6 40 in in ADP fcis-5218 6 41 promoting promote VERB fcis-5218 6 42 the the DET fcis-5218 6 43 development development NOUN fcis-5218 6 44 of of ADP fcis-5218 6 45 intelligent intelligent ADJ fcis-5218 6 46 agriculture agriculture NOUN fcis-5218 6 47 . . PUNCT fcis-5218 7 1 keywords keyword NOUN fcis-5218 7 2 : : PUNCT fcis-5218 7 3 neural neural ADJ fcis-5218 7 4 network network NOUN fcis-5218 7 5 ; ; PUNCT fcis-5218 7 6 deep deep ADJ fcis-5218 7 7 learning learning NOUN fcis-5218 7 8 ; ; PUNCT fcis-5218 7 9 transfer transfer NOUN fcis-5218 7 10 learning learning NOUN fcis-5218 7 11 ; ; PUNCT fcis-5218 7 12 tea tea NOUN fcis-5218 7 13 leaf leaf NOUN fcis-5218 7 14 disease disease NOUN fcis-5218 7 15 . . PUNCT fcis-5218 8 1 1 1 X fcis-5218 8 2 . . X fcis-5218 8 3 introduction introduction NOUN fcis-5218 8 4 tea tea NOUN fcis-5218 8 5 tree tree NOUN fcis-5218 8 6 is be AUX fcis-5218 8 7 an an DET fcis-5218 8 8 important important ADJ fcis-5218 8 9 economic economic ADJ fcis-5218 8 10 crop crop NOUN fcis-5218 8 11 , , PUNCT fcis-5218 8 12 and and CCONJ fcis-5218 8 13 tea tea NOUN fcis-5218 8 14 leaf leaf NOUN fcis-5218 8 15 disease disease NOUN fcis-5218 8 16 is be AUX fcis-5218 8 17 one one NUM fcis-5218 8 18 of of ADP fcis-5218 8 19 the the DET fcis-5218 8 20 important important ADJ fcis-5218 8 21 factors factor NOUN fcis-5218 8 22 affecting affect VERB fcis-5218 8 23 the the DET fcis-5218 8 24 normal normal ADJ fcis-5218 8 25 growth growth NOUN fcis-5218 8 26 of of ADP fcis-5218 8 27 tea tea NOUN fcis-5218 8 28 . . PUNCT fcis-5218 9 1 it it PRON fcis-5218 9 2 is be AUX fcis-5218 9 3 one one NUM fcis-5218 9 4 of of ADP fcis-5218 9 5 the the DET fcis-5218 9 6 important important ADJ fcis-5218 9 7 measures measure NOUN fcis-5218 9 8 to to PART fcis-5218 9 9 ensure ensure VERB fcis-5218 9 10 the the DET fcis-5218 9 11 steady steady ADJ fcis-5218 9 12 growth growth NOUN fcis-5218 9 13 of of ADP fcis-5218 9 14 tea tea NOUN fcis-5218 9 15 production production NOUN fcis-5218 9 16 and and CCONJ fcis-5218 9 17 quality quality NOUN fcis-5218 9 18 , , PUNCT fcis-5218 9 19 which which PRON fcis-5218 9 20 has have VERB fcis-5218 9 21 important important ADJ fcis-5218 9 22 economic economic ADJ fcis-5218 9 23 value value NOUN fcis-5218 9 24 and and CCONJ fcis-5218 9 25 practical practical ADJ fcis-5218 9 26 significance significance NOUN fcis-5218 9 27 . . PUNCT fcis-5218 10 1 in in ADP fcis-5218 10 2 the the DET fcis-5218 10 3 past past ADJ fcis-5218 10 4 years year NOUN fcis-5218 10 5 , , PUNCT fcis-5218 10 6 the the DET fcis-5218 10 7 methods method NOUN fcis-5218 10 8 of of ADP fcis-5218 10 9 identifying identify VERB fcis-5218 10 10 tea tea NOUN fcis-5218 10 11 leaf leaf NOUN fcis-5218 10 12 disease disease NOUN fcis-5218 10 13 were be AUX fcis-5218 10 14 mainly mainly ADV fcis-5218 10 15 through through ADP fcis-5218 10 16 expert expert ADJ fcis-5218 10 17 diagnosis diagnosis NOUN fcis-5218 10 18 or or CCONJ fcis-5218 10 19 consulting consulting NOUN fcis-5218 10 20 experienced experienced ADJ fcis-5218 10 21 farmers farmer NOUN fcis-5218 10 22 . . PUNCT fcis-5218 11 1 however however ADV fcis-5218 11 2 , , PUNCT fcis-5218 11 3 expert expert ADJ fcis-5218 11 4 consultation consultation NOUN fcis-5218 11 5 may may AUX fcis-5218 11 6 have have VERB fcis-5218 11 7 the the DET fcis-5218 11 8 disadvantage disadvantage NOUN fcis-5218 11 9 of of ADP fcis-5218 11 10 high high ADJ fcis-5218 11 11 cost cost NOUN fcis-5218 11 12 and and CCONJ fcis-5218 11 13 lack lack NOUN fcis-5218 11 14 of of ADP fcis-5218 11 15 timeliness timeliness NOUN fcis-5218 11 16 . . PUNCT fcis-5218 12 1 due due ADP fcis-5218 12 2 to to ADP fcis-5218 12 3 individual individual ADJ fcis-5218 12 4 limitations limitation NOUN fcis-5218 12 5 , , PUNCT fcis-5218 12 6 the the DET fcis-5218 12 7 most most ADV fcis-5218 12 8 experienced experienced ADJ fcis-5218 12 9 farmers farmer NOUN fcis-5218 12 10 can can AUX fcis-5218 12 11 not not PART fcis-5218 12 12 accurately accurately ADV fcis-5218 12 13 identify identify VERB fcis-5218 12 14 all all DET fcis-5218 12 15 species specie NOUN fcis-5218 12 16 of of ADP fcis-5218 12 17 leaf leaf NOUN fcis-5218 12 18 pests pest NOUN fcis-5218 12 19 and and CCONJ fcis-5218 12 20 diseases disease NOUN fcis-5218 12 21 . . PUNCT fcis-5218 13 1 therefore therefore ADV fcis-5218 13 2 , , PUNCT fcis-5218 13 3 in in ADP fcis-5218 13 4 the the DET fcis-5218 13 5 context context NOUN fcis-5218 13 6 of of ADP fcis-5218 13 7 smart smart ADJ fcis-5218 13 8 agriculture agriculture NOUN fcis-5218 13 9 , , PUNCT fcis-5218 13 10 it it PRON fcis-5218 13 11 has have AUX fcis-5218 13 12 become become VERB fcis-5218 13 13 an an DET fcis-5218 13 14 important important ADJ fcis-5218 13 15 trend trend NOUN fcis-5218 13 16 to to PART fcis-5218 13 17 promote promote VERB fcis-5218 13 18 the the DET fcis-5218 13 19 identification identification NOUN fcis-5218 13 20 of of ADP fcis-5218 13 21 leaf leaf NOUN fcis-5218 13 22 diseases disease NOUN fcis-5218 13 23 and and CCONJ fcis-5218 13 24 pests pest NOUN fcis-5218 13 25 to to PART fcis-5218 13 26 promote promote VERB fcis-5218 13 27 the the DET fcis-5218 13 28 tea tea NOUN fcis-5218 13 29 production production NOUN fcis-5218 13 30 . . PUNCT fcis-5218 14 1 in in ADP fcis-5218 14 2 previous previous ADJ fcis-5218 14 3 years year NOUN fcis-5218 14 4 , , PUNCT fcis-5218 14 5 the the DET fcis-5218 14 6 traditional traditional ADJ fcis-5218 14 7 machine machine NOUN fcis-5218 14 8 learning learning NOUN fcis-5218 14 9 method method NOUN fcis-5218 14 10 was be AUX fcis-5218 14 11 mainly mainly ADV fcis-5218 14 12 used use VERB fcis-5218 14 13 to to PART fcis-5218 14 14 identify identify VERB fcis-5218 14 15 tea tea NOUN fcis-5218 14 16 leaf leaf NOUN fcis-5218 14 17 diseases disease NOUN fcis-5218 14 18 . . PUNCT fcis-5218 15 1 this this DET fcis-5218 15 2 method method NOUN fcis-5218 15 3 mainly mainly ADV fcis-5218 15 4 manually manually ADV fcis-5218 15 5 extracts extract NOUN fcis-5218 15 6 and and CCONJ fcis-5218 15 7 handles handle VERB fcis-5218 15 8 the the DET fcis-5218 15 9 leaf leaf NOUN fcis-5218 15 10 shape shape NOUN fcis-5218 15 11 , , PUNCT fcis-5218 15 12 color color NOUN fcis-5218 15 13 , , PUNCT fcis-5218 15 14 texture texture NOUN fcis-5218 15 15 and and CCONJ fcis-5218 15 16 other other ADJ fcis-5218 15 17 characteristics characteristic NOUN fcis-5218 15 18 of of ADP fcis-5218 15 19 different different ADJ fcis-5218 15 20 diseases disease NOUN fcis-5218 15 21 and and CCONJ fcis-5218 15 22 pests pest NOUN fcis-5218 15 23 , , PUNCT fcis-5218 15 24 and and CCONJ fcis-5218 15 25 then then ADV fcis-5218 15 26 adopts adopt VERB fcis-5218 15 27 methods method NOUN fcis-5218 15 28 such such ADJ fcis-5218 15 29 as as ADP fcis-5218 15 30 k k NOUN fcis-5218 15 31 - - PUNCT fcis-5218 15 32 means mean NOUN fcis-5218 15 33 , , PUNCT fcis-5218 15 34 support support VERB fcis-5218 15 35 vector vector NOUN fcis-5218 15 36 machines machine NOUN fcis-5218 15 37 ( ( PUNCT fcis-5218 15 38 svm svm PROPN fcis-5218 15 39 ) ) PUNCT fcis-5218 15 40 , , PUNCT fcis-5218 15 41 and and CCONJ fcis-5218 15 42 bayesian bayesian NOUN fcis-5218 15 43 to to PART fcis-5218 15 44 achieve achieve VERB fcis-5218 15 45 classification classification NOUN fcis-5218 15 46 . . PUNCT fcis-5218 16 1 for for ADP fcis-5218 16 2 example example NOUN fcis-5218 16 3 , , PUNCT fcis-5218 16 4 hossain hossain PROPN fcis-5218 16 5 et et PROPN fcis-5218 16 6 al al PROPN fcis-5218 16 7 . . PROPN fcis-5218 16 8 used use VERB fcis-5218 16 9 a a DET fcis-5218 16 10 support support NOUN fcis-5218 16 11 vector vector NOUN fcis-5218 16 12 machine machine NOUN fcis-5218 16 13 to to PART fcis-5218 16 14 classify classify VERB fcis-5218 16 15 and and CCONJ fcis-5218 16 16 identify identify VERB fcis-5218 16 17 the the DET fcis-5218 16 18 extracted extract VERB fcis-5218 16 19 image image NOUN fcis-5218 16 20 features feature NOUN fcis-5218 16 21 of of ADP fcis-5218 16 22 tea tea NOUN fcis-5218 16 23 leaf leaf NOUN fcis-5218 16 24 disease disease NOUN fcis-5218 16 25 . . PUNCT fcis-5218 17 1 shao shao PROPN fcis-5218 17 2 et et PROPN fcis-5218 17 3 al al PROPN fcis-5218 17 4 . . PROPN fcis-5218 17 5 used use VERB fcis-5218 17 6 the the DET fcis-5218 17 7 k k NOUN fcis-5218 17 8 - - PUNCT fcis-5218 17 9 means means NOUN fcis-5218 17 10 algorithm algorithm NOUN fcis-5218 17 11 for for ADP fcis-5218 17 12 tea tea NOUN fcis-5218 17 13 leaf leaf NOUN fcis-5218 17 14 bud bud NOUN fcis-5218 17 15 recognition recognition NOUN fcis-5218 17 16 . . PUNCT fcis-5218 18 1 the the DET fcis-5218 18 2 classification classification NOUN fcis-5218 18 3 accuracy accuracy NOUN fcis-5218 18 4 of of ADP fcis-5218 18 5 traditional traditional ADJ fcis-5218 18 6 machine machine NOUN fcis-5218 18 7 learning learning NOUN fcis-5218 18 8 methods method NOUN fcis-5218 18 9 largely largely ADV fcis-5218 18 10 depends depend VERB fcis-5218 18 11 on on ADP fcis-5218 18 12 the the DET fcis-5218 18 13 effect effect NOUN fcis-5218 18 14 of of ADP fcis-5218 18 15 manual manual ADJ fcis-5218 18 16 design design NOUN fcis-5218 18 17 features feature NOUN fcis-5218 18 18 . . PUNCT fcis-5218 19 1 however however ADV fcis-5218 19 2 , , PUNCT fcis-5218 19 3 in in ADP fcis-5218 19 4 some some DET fcis-5218 19 5 high high ADJ fcis-5218 19 6 noise noise NOUN fcis-5218 19 7 environment environment NOUN fcis-5218 19 8 with with ADP fcis-5218 19 9 complex complex ADJ fcis-5218 19 10 background background NOUN fcis-5218 19 11 , , PUNCT fcis-5218 19 12 the the DET fcis-5218 19 13 image image NOUN fcis-5218 19 14 recognition recognition NOUN fcis-5218 19 15 accuracy accuracy NOUN fcis-5218 19 16 will will AUX fcis-5218 19 17 be be AUX fcis-5218 19 18 greatly greatly ADV fcis-5218 19 19 reduced reduce VERB fcis-5218 19 20 . . PUNCT fcis-5218 20 1 with with ADP fcis-5218 20 2 the the DET fcis-5218 20 3 development development NOUN fcis-5218 20 4 of of ADP fcis-5218 20 5 big big ADJ fcis-5218 20 6 data datum NOUN fcis-5218 20 7 technology technology NOUN fcis-5218 20 8 and and CCONJ fcis-5218 20 9 the the DET fcis-5218 20 10 continuous continuous ADJ fcis-5218 20 11 improvement improvement NOUN fcis-5218 20 12 of of ADP fcis-5218 20 13 the the DET fcis-5218 20 14 computing computing NOUN fcis-5218 20 15 ability ability NOUN fcis-5218 20 16 of of ADP fcis-5218 20 17 hardware hardware NOUN fcis-5218 20 18 , , PUNCT fcis-5218 20 19 deep deep ADJ fcis-5218 20 20 learning learning NOUN fcis-5218 20 21 technology technology NOUN fcis-5218 20 22 has have AUX fcis-5218 20 23 been be AUX fcis-5218 20 24 quickly quickly ADV fcis-5218 20 25 applied apply VERB fcis-5218 20 26 to to ADP fcis-5218 20 27 the the DET fcis-5218 20 28 tea tea NOUN fcis-5218 20 29 leaf leaf NOUN fcis-5218 20 30 disease disease NOUN fcis-5218 20 31 identification identification NOUN fcis-5218 20 32 . . PUNCT fcis-5218 21 1 although although SCONJ fcis-5218 21 2 it it PRON fcis-5218 21 3 belongs belong VERB fcis-5218 21 4 to to ADP fcis-5218 21 5 the the DET fcis-5218 21 6 field field NOUN fcis-5218 21 7 of of ADP fcis-5218 21 8 machine machine NOUN fcis-5218 21 9 learning learning NOUN fcis-5218 21 10 , , PUNCT fcis-5218 21 11 deep deep ADJ fcis-5218 21 12 learning learning NOUN fcis-5218 21 13 emphasizes emphasize VERB fcis-5218 21 14 the the DET fcis-5218 21 15 use use NOUN fcis-5218 21 16 of of ADP fcis-5218 21 17 deep deep ADJ fcis-5218 21 18 network network NOUN fcis-5218 21 19 construction construction NOUN fcis-5218 21 20 and and CCONJ fcis-5218 21 21 convolution convolution NOUN fcis-5218 21 22 operation operation NOUN fcis-5218 21 23 to to PART fcis-5218 21 24 improve improve VERB fcis-5218 21 25 the the DET fcis-5218 21 26 nonlinear nonlinear ADJ fcis-5218 21 27 expression expression NOUN fcis-5218 21 28 ability ability NOUN fcis-5218 21 29 of of ADP fcis-5218 21 30 data datum NOUN fcis-5218 21 31 . . PUNCT fcis-5218 22 1 compared compare VERB fcis-5218 22 2 with with ADP fcis-5218 22 3 traditional traditional ADJ fcis-5218 22 4 machine machine NOUN fcis-5218 22 5 learning learning NOUN fcis-5218 22 6 methods method NOUN fcis-5218 22 7 , , PUNCT fcis-5218 22 8 deep deep ADJ fcis-5218 22 9 learning learning NOUN fcis-5218 22 10 has have VERB fcis-5218 22 11 better well ADJ fcis-5218 22 12 feature feature NOUN fcis-5218 22 13 learning learn VERB fcis-5218 22 14 ability ability NOUN fcis-5218 22 15 . . PUNCT fcis-5218 23 1 it it PRON fcis-5218 23 2 shows show VERB fcis-5218 23 3 an an DET fcis-5218 23 4 efficient efficient ADJ fcis-5218 23 5 recognition recognition NOUN fcis-5218 23 6 effect effect NOUN fcis-5218 23 7 even even ADV fcis-5218 23 8 when when SCONJ fcis-5218 23 9 dealing deal VERB fcis-5218 23 10 with with ADP fcis-5218 23 11 the the DET fcis-5218 23 12 classification classification NOUN fcis-5218 23 13 problem problem NOUN fcis-5218 23 14 of of ADP fcis-5218 23 15 complex complex ADJ fcis-5218 23 16 images image NOUN fcis-5218 23 17 , , PUNCT fcis-5218 23 18 including include VERB fcis-5218 23 19 its its PRON fcis-5218 23 20 successful successful ADJ fcis-5218 23 21 application application NOUN fcis-5218 23 22 in in ADP fcis-5218 23 23 the the DET fcis-5218 23 24 field field NOUN fcis-5218 23 25 of of ADP fcis-5218 23 26 tea tea NOUN fcis-5218 23 27 leaf leaf NOUN fcis-5218 23 28 disease disease NOUN fcis-5218 23 29 recognition recognition NOUN fcis-5218 23 30 . . PUNCT fcis-5218 24 1 for for ADP fcis-5218 24 2 example example NOUN fcis-5218 24 3 , , PUNCT fcis-5218 24 4 the the DET fcis-5218 24 5 identification identification NOUN fcis-5218 24 6 network network NOUN fcis-5218 24 7 of of ADP fcis-5218 24 8 tea tea NOUN fcis-5218 24 9 leaf leaf NOUN fcis-5218 24 10 diseases disease NOUN fcis-5218 24 11 can can AUX fcis-5218 24 12 achieve achieve VERB fcis-5218 24 13 high high ADJ fcis-5218 24 14 recognition recognition NOUN fcis-5218 24 15 accuracy accuracy NOUN fcis-5218 24 16 by by ADP fcis-5218 24 17 adjusting adjust VERB fcis-5218 24 18 the the DET fcis-5218 24 19 network network NOUN fcis-5218 24 20 parameters parameter NOUN fcis-5218 24 21 and and CCONJ fcis-5218 24 22 the the DET fcis-5218 24 23 addition addition NOUN fcis-5218 24 24 of of ADP fcis-5218 24 25 dropout dropout NOUN fcis-5218 24 26 . . PUNCT fcis-5218 25 1 a a DET fcis-5218 25 2 cnns cnns PROPN fcis-5218 25 3 model model NOUN fcis-5218 25 4 named name VERB fcis-5218 25 5 leafnet leafnet PROPN fcis-5218 25 6 was be AUX fcis-5218 25 7 developed develop VERB fcis-5218 25 8 to to PART fcis-5218 25 9 automatically automatically ADV fcis-5218 25 10 extract extract VERB fcis-5218 25 11 the the DET fcis-5218 25 12 features feature NOUN fcis-5218 25 13 of of ADP fcis-5218 25 14 tea tea NOUN fcis-5218 25 15 plant plant NOUN fcis-5218 25 16 diseases disease NOUN fcis-5218 25 17 from from ADP fcis-5218 25 18 images image NOUN fcis-5218 25 19 , , PUNCT fcis-5218 25 20 which which PRON fcis-5218 25 21 was be AUX fcis-5218 25 22 superior superior ADJ fcis-5218 25 23 in in ADP fcis-5218 25 24 the the DET fcis-5218 25 25 recognition recognition NOUN fcis-5218 25 26 of of ADP fcis-5218 25 27 tea tea NOUN fcis-5218 25 28 leaf leaf NOUN fcis-5218 25 29 diseases disease NOUN fcis-5218 25 30 compared compare VERB fcis-5218 25 31 to to ADP fcis-5218 25 32 the the DET fcis-5218 25 33 mlp mlp NOUN fcis-5218 25 34 and and CCONJ fcis-5218 25 35 svm svm ADJ fcis-5218 25 36 algorithms algorithm NOUN fcis-5218 25 37 . . PUNCT fcis-5218 26 1 deep deep ADJ fcis-5218 26 2 learning learning NOUN fcis-5218 26 3 requires require VERB fcis-5218 26 4 a a DET fcis-5218 26 5 large large ADJ fcis-5218 26 6 number number NOUN fcis-5218 26 7 of of ADP fcis-5218 26 8 labeled label VERB fcis-5218 26 9 samples sample NOUN fcis-5218 26 10 for for ADP fcis-5218 26 11 model model NOUN fcis-5218 26 12 training training NOUN fcis-5218 26 13 to to PART fcis-5218 26 14 prevent prevent VERB fcis-5218 26 15 overfitting overfitting NOUN fcis-5218 26 16 . . PUNCT fcis-5218 27 1 however however ADV fcis-5218 27 2 , , PUNCT fcis-5218 27 3 some some DET fcis-5218 27 4 samples sample NOUN fcis-5218 27 5 of of ADP fcis-5218 27 6 leaf leaf NOUN fcis-5218 27 7 disease disease NOUN fcis-5218 27 8 types type NOUN fcis-5218 27 9 are be AUX fcis-5218 27 10 not not PART fcis-5218 27 11 easy easy ADJ fcis-5218 27 12 to to PART fcis-5218 27 13 obtain obtain VERB fcis-5218 27 14 in in ADP fcis-5218 27 15 practice practice NOUN fcis-5218 27 16 , , PUNCT fcis-5218 27 17 resulting result VERB fcis-5218 27 18 in in ADP fcis-5218 27 19 the the DET fcis-5218 27 20 training training NOUN fcis-5218 27 21 process process NOUN fcis-5218 27 22 ca can AUX fcis-5218 27 23 n't not PART fcis-5218 27 24 effectively effectively ADV fcis-5218 27 25 converge converge VERB fcis-5218 27 26 . . PUNCT fcis-5218 28 1 it it PRON fcis-5218 28 2 's be AUX fcis-5218 28 3 important important ADJ fcis-5218 28 4 for for SCONJ fcis-5218 28 5 the the DET fcis-5218 28 6 neural neural ADJ fcis-5218 28 7 network network NOUN fcis-5218 28 8 model model NOUN fcis-5218 28 9 to to PART fcis-5218 28 10 be be AUX fcis-5218 28 11 trained train VERB fcis-5218 28 12 with with ADP fcis-5218 28 13 high high ADJ fcis-5218 28 14 recognition recognition NOUN fcis-5218 28 15 accuracy accuracy NOUN fcis-5218 28 16 even even ADV fcis-5218 28 17 though though SCONJ fcis-5218 28 18 the the DET fcis-5218 28 19 sample sample NOUN fcis-5218 28 20 size size NOUN fcis-5218 28 21 is be AUX fcis-5218 28 22 few few ADJ fcis-5218 28 23 , , PUNCT fcis-5218 28 24 which which PRON fcis-5218 28 25 is be AUX fcis-5218 28 26 the the DET fcis-5218 28 27 key key NOUN fcis-5218 28 28 to to PART fcis-5218 28 29 ensure ensure VERB fcis-5218 28 30 the the DET fcis-5218 28 31 further further ADJ fcis-5218 28 32 popularization popularization NOUN fcis-5218 28 33 of of ADP fcis-5218 28 34 deep deep ADJ fcis-5218 28 35 learning learning NOUN fcis-5218 28 36 in in ADP fcis-5218 28 37 agriculture agriculture NOUN fcis-5218 28 38 . . PUNCT fcis-5218 29 1 inspired inspire VERB fcis-5218 29 2 by by ADP fcis-5218 29 3 the the DET fcis-5218 29 4 characteristics characteristic NOUN fcis-5218 29 5 of of ADP fcis-5218 29 6 human human ADJ fcis-5218 29 7 learning learning NOUN fcis-5218 29 8 , , PUNCT fcis-5218 29 9 transfer transfer NOUN fcis-5218 29 10 learning learning NOUN fcis-5218 29 11 can can AUX fcis-5218 29 12 effectively effectively ADV fcis-5218 29 13 solve solve VERB fcis-5218 29 14 the the DET fcis-5218 29 15 problem problem NOUN fcis-5218 29 16 of of ADP fcis-5218 29 17 sample sample NOUN fcis-5218 29 18 insufficiency insufficiency NOUN fcis-5218 29 19 , , PUNCT fcis-5218 29 20 and and CCONJ fcis-5218 29 21 attract attract VERB fcis-5218 29 22 increasing increase VERB fcis-5218 29 23 attention attention NOUN fcis-5218 29 24 through through ADP fcis-5218 29 25 the the DET fcis-5218 29 26 years year NOUN fcis-5218 29 27 . . PUNCT fcis-5218 30 1 this this DET fcis-5218 30 2 method method NOUN fcis-5218 30 3 does do AUX fcis-5218 30 4 not not PART fcis-5218 30 5 require require VERB fcis-5218 30 6 a a DET fcis-5218 30 7 large large ADJ fcis-5218 30 8 number number NOUN fcis-5218 30 9 of of ADP fcis-5218 30 10 training training NOUN fcis-5218 30 11 samples sample NOUN fcis-5218 30 12 when when SCONJ fcis-5218 30 13 encountering encounter VERB fcis-5218 30 14 a a DET fcis-5218 30 15 new new ADJ fcis-5218 30 16 category category NOUN fcis-5218 30 17 , , PUNCT fcis-5218 30 18 but but CCONJ fcis-5218 30 19 rather rather ADV fcis-5218 30 20 summarizes summarize VERB fcis-5218 30 21 them they PRON fcis-5218 30 22 from from ADP fcis-5218 30 23 the the DET fcis-5218 30 24 existing exist VERB fcis-5218 30 25 knowledge knowledge NOUN fcis-5218 30 26 to to PART fcis-5218 30 27 achieve achieve VERB fcis-5218 30 28 knowledge knowledge NOUN fcis-5218 30 29 transfer transfer NOUN fcis-5218 30 30 , , PUNCT fcis-5218 30 31 which which PRON fcis-5218 30 32 is be AUX fcis-5218 30 33 therefore therefore ADV fcis-5218 30 34 very very ADV fcis-5218 30 35 close close ADJ fcis-5218 30 36 to to ADP fcis-5218 30 37 the the DET fcis-5218 30 38 human human ADJ fcis-5218 30 39 learning learning NOUN fcis-5218 30 40 style style NOUN fcis-5218 30 41 . . PUNCT fcis-5218 31 1 in in ADP fcis-5218 31 2 this this DET fcis-5218 31 3 study study NOUN fcis-5218 31 4 , , PUNCT fcis-5218 31 5 we we PRON fcis-5218 31 6 proposed propose VERB fcis-5218 31 7 a a DET fcis-5218 31 8 neural neural ADJ fcis-5218 31 9 network network NOUN fcis-5218 31 10 model model NOUN fcis-5218 31 11 based base VERB fcis-5218 31 12 on on ADP fcis-5218 31 13 transfer transfer NOUN fcis-5218 31 14 learning learn VERB fcis-5218 31 15 to to PART fcis-5218 31 16 identify identify VERB fcis-5218 31 17 tea tea NOUN fcis-5218 31 18 leaf leaf NOUN fcis-5218 31 19 diseases disease NOUN fcis-5218 31 20 . . PUNCT fcis-5218 32 1 the the DET fcis-5218 32 2 inception inception PROPN fcis-5218 32 3 - - PUNCT fcis-5218 32 4 v3 v3 NOUN fcis-5218 32 5 network network NOUN fcis-5218 32 6 is be AUX fcis-5218 32 7 used use VERB fcis-5218 32 8 to to PART fcis-5218 32 9 extract extract VERB fcis-5218 32 10 the the DET fcis-5218 32 11 features feature NOUN fcis-5218 32 12 of of ADP fcis-5218 32 13 tea tea NOUN fcis-5218 32 14 leaf leaf NOUN fcis-5218 32 15 image image NOUN fcis-5218 32 16 , , PUNCT fcis-5218 32 17 and and CCONJ fcis-5218 32 18 the the DET fcis-5218 32 19 loss loss NOUN fcis-5218 32 20 function function NOUN fcis-5218 32 21 is be AUX fcis-5218 32 22 improved improve VERB fcis-5218 32 23 to to PART fcis-5218 32 24 realize realize VERB fcis-5218 32 25 the the DET fcis-5218 32 26 promotion promotion NOUN fcis-5218 32 27 of of ADP fcis-5218 32 28 identification identification NOUN fcis-5218 32 29 accuracy accuracy NOUN fcis-5218 32 30 . . PUNCT fcis-5218 33 1 2 2 X fcis-5218 33 2 . . X fcis-5218 33 3 related relate VERB fcis-5218 33 4 theory theory NOUN fcis-5218 33 5 2.1 2.1 NUM fcis-5218 33 6 . . PUNCT fcis-5218 34 1 convolutional convolutional ADJ fcis-5218 34 2 neural neural ADJ fcis-5218 34 3 network network NOUN fcis-5218 34 4 convolutional convolutional ADJ fcis-5218 34 5 neural neural ADJ fcis-5218 34 6 network network NOUN fcis-5218 34 7 ( ( PUNCT fcis-5218 34 8 cnn cnn PROPN fcis-5218 34 9 ) ) PUNCT fcis-5218 34 10 is be AUX fcis-5218 34 11 an an DET fcis-5218 34 12 important important ADJ fcis-5218 34 13 foundation foundation NOUN fcis-5218 34 14 for for ADP fcis-5218 34 15 deep deep ADJ fcis-5218 34 16 learning learning NOUN fcis-5218 34 17 to to PART fcis-5218 34 18 achieve achieve VERB fcis-5218 34 19 breakthrough breakthrough NOUN fcis-5218 34 20 in in ADP fcis-5218 34 21 image image NOUN fcis-5218 34 22 classification classification NOUN fcis-5218 34 23 and and CCONJ fcis-5218 34 24 recognition recognition NOUN fcis-5218 34 25 . . PUNCT fcis-5218 35 1 different different ADJ fcis-5218 35 2 from from ADP fcis-5218 35 3 the the DET fcis-5218 35 4 traditional traditional ADJ fcis-5218 35 5 machine machine NOUN fcis-5218 35 6 learning learning NOUN fcis-5218 35 7 method method NOUN fcis-5218 35 8 , , PUNCT fcis-5218 35 9 the the DET fcis-5218 35 10 neural neural ADJ fcis-5218 35 11 network network NOUN fcis-5218 35 12 based base VERB fcis-5218 35 13 on on ADP fcis-5218 35 14 cnn cnn PROPN fcis-5218 35 15 can can AUX fcis-5218 35 16 realize realize VERB fcis-5218 35 17 the the DET fcis-5218 35 18 end end NOUN fcis-5218 35 19 - - PUNCT fcis-5218 35 20 to to ADP fcis-5218 35 21 - - PUNCT fcis-5218 35 22 end end VERB fcis-5218 35 23 automatic automatic ADJ fcis-5218 35 24 feature feature NOUN fcis-5218 35 25 extraction extraction NOUN fcis-5218 35 26 . . PUNCT fcis-5218 36 1 typical typical ADJ fcis-5218 36 2 cnn cnn PROPN fcis-5218 36 3 architecture architecture NOUN fcis-5218 36 4 mainly mainly ADV fcis-5218 36 5 includes include VERB fcis-5218 36 6 convolutional convolutional ADJ fcis-5218 36 7 layer layer NOUN fcis-5218 36 8 , , PUNCT fcis-5218 36 9 pooling pool VERB fcis-5218 36 10 layer layer NOUN fcis-5218 36 11 and and CCONJ fcis-5218 36 12 fully fully ADV fcis-5218 36 13 connected connected ADJ fcis-5218 36 14 layer layer NOUN fcis-5218 36 15 . . PUNCT fcis-5218 37 1 ( ( PUNCT fcis-5218 37 2 1 1 X fcis-5218 37 3 ) ) PUNCT fcis-5218 37 4 convolutional convolutional ADJ fcis-5218 37 5 layers layer NOUN fcis-5218 37 6 the the DET fcis-5218 37 7 convolutional convolutional ADJ fcis-5218 37 8 layer layer NOUN fcis-5218 37 9 performs perform VERB fcis-5218 37 10 the the DET fcis-5218 37 11 convolution convolution NOUN fcis-5218 37 12 76 76 NUM fcis-5218 37 13 operation operation NOUN fcis-5218 37 14 by by ADP fcis-5218 37 15 convolution convolution NOUN fcis-5218 37 16 kernel kernel PROPN fcis-5218 37 17 , , PUNCT fcis-5218 37 18 adding add VERB fcis-5218 37 19 an an DET fcis-5218 37 20 offset offset ADJ fcis-5218 37 21 vector vector NOUN fcis-5218 37 22 resulting result VERB fcis-5218 37 23 the the DET fcis-5218 37 24 output output NOUN fcis-5218 37 25 . . PUNCT fcis-5218 38 1 the the DET fcis-5218 38 2 model model NOUN fcis-5218 38 3 parameters parameter NOUN fcis-5218 38 4 of of ADP fcis-5218 38 5 the the DET fcis-5218 38 6 convolutional convolutional ADJ fcis-5218 38 7 layer layer NOUN fcis-5218 38 8 include include VERB fcis-5218 38 9 the the DET fcis-5218 38 10 convolution convolution NOUN fcis-5218 38 11 kernel kernel NOUN fcis-5218 38 12 part part NOUN fcis-5218 38 13 and and CCONJ fcis-5218 38 14 the the DET fcis-5218 38 15 offset offset ADJ fcis-5218 38 16 part part NOUN fcis-5218 38 17 . . PUNCT fcis-5218 39 1 convolutional convolutional ADJ fcis-5218 39 2 networks network NOUN fcis-5218 39 3 of of ADP fcis-5218 39 4 different different ADJ fcis-5218 39 5 complexity complexity NOUN fcis-5218 39 6 can can AUX fcis-5218 39 7 be be AUX fcis-5218 39 8 designed design VERB fcis-5218 39 9 by by ADP fcis-5218 39 10 the the DET fcis-5218 39 11 superposition superposition NOUN fcis-5218 39 12 of of ADP fcis-5218 39 13 multiple multiple ADJ fcis-5218 39 14 convolutional convolutional ADJ fcis-5218 39 15 layers layer NOUN fcis-5218 39 16 . . PUNCT fcis-5218 40 1 the the DET fcis-5218 40 2 characteristic characteristic ADJ fcis-5218 40 3 output output NOUN fcis-5218 40 4 of of ADP fcis-5218 40 5 the the DET fcis-5218 40 6 i i PROPN fcis-5218 40 7 - - PUNCT fcis-5218 40 8 th th X fcis-5218 40 9 layer layer NOUN fcis-5218 40 10 convolutional convolutional ADJ fcis-5218 40 11 network network NOUN fcis-5218 40 12 can can AUX fcis-5218 40 13 be be AUX fcis-5218 40 14 expressed express VERB fcis-5218 40 15 as as ADP fcis-5218 40 16 : : PUNCT fcis-5218 40 17 β„Žπ‘– β„Žπ‘– NOUN fcis-5218 40 18 = = PUNCT fcis-5218 40 19 𝑓(β„Žπ‘–βˆ’1π‘Šπ‘– 𝑓(β„Žπ‘–βˆ’1π‘Šπ‘– PROPN fcis-5218 40 20 + + X fcis-5218 40 21 𝑏𝑖 𝑏𝑖 NOUN fcis-5218 40 22 ) ) PUNCT fcis-5218 40 23 where where SCONJ fcis-5218 40 24 π‘Šπ‘– π‘Šπ‘– PROPN fcis-5218 40 25 is be AUX fcis-5218 40 26 the the DET fcis-5218 40 27 weight weight NOUN fcis-5218 40 28 of of ADP fcis-5218 40 29 the the DET fcis-5218 40 30 convolution convolution NOUN fcis-5218 40 31 kernel kernel NOUN fcis-5218 40 32 in in ADP fcis-5218 40 33 i i PROPN fcis-5218 40 34 - - PUNCT fcis-5218 40 35 th th X fcis-5218 40 36 layer layer NOUN fcis-5218 40 37 , , PUNCT fcis-5218 40 38 𝑏𝑖 𝑏𝑖 PROPN fcis-5218 40 39 is be AUX fcis-5218 40 40 the the DET fcis-5218 40 41 standard standard ADJ fcis-5218 40 42 deviation deviation NOUN fcis-5218 40 43 , , PUNCT fcis-5218 40 44 β„Žπ‘–βˆ’1 β„Žπ‘–βˆ’1 PROPN fcis-5218 40 45 is be AUX fcis-5218 40 46 the the DET fcis-5218 40 47 feature feature NOUN fcis-5218 40 48 input input NOUN fcis-5218 40 49 of of ADP fcis-5218 40 50 the the DET fcis-5218 40 51 convolutional convolutional ADJ fcis-5218 40 52 layer layer NOUN fcis-5218 40 53 ( ( PUNCT fcis-5218 40 54 β„Ž0 β„Ž0 NOUN fcis-5218 40 55 denotes denote VERB fcis-5218 40 56 the the DET fcis-5218 40 57 input input NOUN fcis-5218 40 58 image image NOUN fcis-5218 40 59 ) ) PUNCT fcis-5218 40 60 , , PUNCT fcis-5218 40 61 and and CCONJ fcis-5218 40 62 f f PROPN fcis-5218 40 63 is be AUX fcis-5218 40 64 the the DET fcis-5218 40 65 activation activation NOUN fcis-5218 40 66 function function NOUN fcis-5218 40 67 . . PUNCT fcis-5218 41 1 ( ( PUNCT fcis-5218 41 2 2 2 X fcis-5218 41 3 ) ) PUNCT fcis-5218 41 4 pooling pool VERB fcis-5218 41 5 layers layer NOUN fcis-5218 41 6 the the DET fcis-5218 41 7 function function NOUN fcis-5218 41 8 of of ADP fcis-5218 41 9 pooling pool VERB fcis-5218 41 10 layers layer NOUN fcis-5218 41 11 is be AUX fcis-5218 41 12 reducing reduce VERB fcis-5218 41 13 the the DET fcis-5218 41 14 excessive excessive ADJ fcis-5218 41 15 sensitivity sensitivity NOUN fcis-5218 41 16 of of ADP fcis-5218 41 17 the the DET fcis-5218 41 18 convolutional convolutional ADJ fcis-5218 41 19 layer layer NOUN fcis-5218 41 20 to to ADP fcis-5218 41 21 the the DET fcis-5218 41 22 position position NOUN fcis-5218 41 23 , , PUNCT fcis-5218 41 24 so so SCONJ fcis-5218 41 25 that that SCONJ fcis-5218 41 26 the the DET fcis-5218 41 27 model model NOUN fcis-5218 41 28 also also ADV fcis-5218 41 29 has have VERB fcis-5218 41 30 a a DET fcis-5218 41 31 good good ADJ fcis-5218 41 32 recognition recognition NOUN fcis-5218 41 33 ability ability NOUN fcis-5218 41 34 of of ADP fcis-5218 41 35 the the DET fcis-5218 41 36 same same ADJ fcis-5218 41 37 object object NOUN fcis-5218 41 38 in in ADP fcis-5218 41 39 different different ADJ fcis-5218 41 40 positions position NOUN fcis-5218 41 41 . . PUNCT fcis-5218 42 1 commonly commonly ADV fcis-5218 42 2 used use VERB fcis-5218 42 3 pooling pooling NOUN fcis-5218 42 4 methods method NOUN fcis-5218 42 5 include include VERB fcis-5218 42 6 maximum maximum ADJ fcis-5218 42 7 pooling pooling NOUN fcis-5218 42 8 and and CCONJ fcis-5218 42 9 average average ADJ fcis-5218 42 10 pooling pooling NOUN fcis-5218 42 11 , , PUNCT fcis-5218 42 12 which which PRON fcis-5218 42 13 calculate calculate VERB fcis-5218 42 14 the the DET fcis-5218 42 15 maximum maximum ADJ fcis-5218 42 16 and and CCONJ fcis-5218 42 17 average average ADJ fcis-5218 42 18 values value NOUN fcis-5218 42 19 in in ADP fcis-5218 42 20 the the DET fcis-5218 42 21 pooling pooling NOUN fcis-5218 42 22 window window NOUN fcis-5218 42 23 respectively respectively ADV fcis-5218 42 24 . . PUNCT fcis-5218 43 1 ( ( PUNCT fcis-5218 43 2 3 3 X fcis-5218 43 3 ) ) PUNCT fcis-5218 43 4 fully fully ADV fcis-5218 43 5 connected connect VERB fcis-5218 43 6 layer layer NOUN fcis-5218 43 7 after after ADP fcis-5218 43 8 feature feature NOUN fcis-5218 43 9 extraction extraction NOUN fcis-5218 43 10 in in ADP fcis-5218 43 11 the the DET fcis-5218 43 12 deep deep ADJ fcis-5218 43 13 network network NOUN fcis-5218 43 14 , , PUNCT fcis-5218 43 15 one one NUM fcis-5218 43 16 or or CCONJ fcis-5218 43 17 more more ADJ fcis-5218 43 18 layers layer NOUN fcis-5218 43 19 are be AUX fcis-5218 43 20 usually usually ADV fcis-5218 43 21 followed follow VERB fcis-5218 43 22 for for ADP fcis-5218 43 23 the the DET fcis-5218 43 24 image image NOUN fcis-5218 43 25 classification classification NOUN fcis-5218 43 26 task task NOUN fcis-5218 43 27 . . PUNCT fcis-5218 44 1 the the DET fcis-5218 44 2 final final ADJ fcis-5218 44 3 output output NOUN fcis-5218 44 4 layer layer NOUN fcis-5218 44 5 usually usually ADV fcis-5218 44 6 uses use VERB fcis-5218 44 7 the the DET fcis-5218 44 8 softmax softmax NOUN fcis-5218 44 9 function function NOUN fcis-5218 44 10 to to PART fcis-5218 44 11 map map VERB fcis-5218 44 12 the the DET fcis-5218 44 13 output output NOUN fcis-5218 44 14 values value NOUN fcis-5218 44 15 into into ADP fcis-5218 44 16 a a DET fcis-5218 44 17 probability probability NOUN fcis-5218 44 18 distribution distribution NOUN fcis-5218 44 19 in in ADP fcis-5218 44 20 which which PRON fcis-5218 44 21 all all DET fcis-5218 44 22 values value NOUN fcis-5218 44 23 are be AUX fcis-5218 44 24 positive positive ADJ fcis-5218 44 25 and and CCONJ fcis-5218 44 26 the the DET fcis-5218 44 27 sum sum NOUN fcis-5218 44 28 is be AUX fcis-5218 44 29 1 1 NUM fcis-5218 44 30 . . PUNCT fcis-5218 45 1 the the DET fcis-5218 45 2 j j PROPN fcis-5218 45 3 - - PUNCT fcis-5218 45 4 th th VERB fcis-5218 45 5 output output NOUN fcis-5218 45 6 result result NOUN fcis-5218 45 7 can can AUX fcis-5218 45 8 be be AUX fcis-5218 45 9 expressed express VERB fcis-5218 45 10 as as ADP fcis-5218 45 11 : : PUNCT fcis-5218 45 12 𝑂𝑗 𝑂𝑗 PROPN fcis-5218 45 13 = = PUNCT fcis-5218 45 14 π‘ π‘œπ‘“π‘‘π‘šπ‘Žπ‘₯(𝑧𝑗 π‘ π‘œπ‘“π‘‘π‘šπ‘Žπ‘₯(𝑧𝑗 PROPN fcis-5218 45 15 ) ) PUNCT fcis-5218 45 16 = = SYM fcis-5218 45 17 𝑒𝑧𝑗/βˆ‘ 𝑒𝑧𝑗/βˆ‘ ADJ fcis-5218 45 18 π‘’π‘§π‘˜ π‘’π‘§π‘˜ NOUN fcis-5218 46 1 𝐾 𝐾 PROPN fcis-5218 46 2 π‘˜=1 π‘˜=1 PROPN fcis-5218 46 3 where where SCONJ fcis-5218 46 4 j∈ j∈ PROPN fcis-5218 46 5 { { PUNCT fcis-5218 46 6 1 1 NUM fcis-5218 46 7 … … PUNCT fcis-5218 46 8 , , PUNCT fcis-5218 46 9 k k NOUN fcis-5218 46 10 } } PUNCT fcis-5218 46 11 , , PUNCT fcis-5218 46 12 k k PROPN fcis-5218 46 13 is be AUX fcis-5218 46 14 the the DET fcis-5218 46 15 output output NOUN fcis-5218 46 16 numbers number NOUN fcis-5218 46 17 of of ADP fcis-5218 46 18 the the DET fcis-5218 46 19 last last ADJ fcis-5218 46 20 connection connection NOUN fcis-5218 46 21 layer layer NOUN fcis-5218 46 22 , , PUNCT fcis-5218 46 23 and and CCONJ fcis-5218 46 24 z z NOUN fcis-5218 46 25 represents represent VERB fcis-5218 46 26 the the DET fcis-5218 46 27 output output NOUN fcis-5218 46 28 vector vector NOUN fcis-5218 46 29 of of ADP fcis-5218 46 30 the the DET fcis-5218 46 31 last last ADJ fcis-5218 46 32 connection connection NOUN fcis-5218 46 33 layer layer NOUN fcis-5218 46 34 . . PUNCT fcis-5218 47 1 2.2 2.2 NUM fcis-5218 47 2 . . PUNCT fcis-5218 47 3 transfer transfer NOUN fcis-5218 47 4 learning learn VERB fcis-5218 47 5 ordinary ordinary ADJ fcis-5218 47 6 deep deep ADJ fcis-5218 47 7 learning learning NOUN fcis-5218 47 8 models model NOUN fcis-5218 47 9 are be AUX fcis-5218 47 10 trained train VERB fcis-5218 47 11 for for ADP fcis-5218 47 12 specific specific ADJ fcis-5218 47 13 data datum NOUN fcis-5218 47 14 or or CCONJ fcis-5218 47 15 tasks task NOUN fcis-5218 47 16 . . PUNCT fcis-5218 48 1 when when SCONJ fcis-5218 48 2 these these DET fcis-5218 48 3 data datum NOUN fcis-5218 48 4 or or CCONJ fcis-5218 48 5 tasks task NOUN fcis-5218 48 6 change change NOUN fcis-5218 48 7 , , PUNCT fcis-5218 48 8 the the DET fcis-5218 48 9 learned learn VERB fcis-5218 48 10 models model NOUN fcis-5218 48 11 usually usually ADV fcis-5218 48 12 have have VERB fcis-5218 48 13 poor poor ADJ fcis-5218 48 14 generalization generalization NOUN fcis-5218 48 15 , , PUNCT fcis-5218 48 16 and and CCONJ fcis-5218 48 17 need need VERB fcis-5218 48 18 to to PART fcis-5218 48 19 be be AUX fcis-5218 48 20 trained train VERB fcis-5218 48 21 again again ADV fcis-5218 48 22 . . PUNCT fcis-5218 49 1 however however ADV fcis-5218 49 2 , , PUNCT fcis-5218 49 3 the the DET fcis-5218 49 4 knowledge knowledge NOUN fcis-5218 49 5 learned learn VERB fcis-5218 49 6 by by ADP fcis-5218 49 7 the the DET fcis-5218 49 8 model model NOUN fcis-5218 49 9 can can AUX fcis-5218 49 10 usually usually ADV fcis-5218 49 11 be be AUX fcis-5218 49 12 reused reuse VERB fcis-5218 49 13 by by ADP fcis-5218 49 14 similar similar ADJ fcis-5218 49 15 tasks task NOUN fcis-5218 49 16 . . PUNCT fcis-5218 50 1 transfer transfer NOUN fcis-5218 50 2 learning learning NOUN fcis-5218 50 3 is be AUX fcis-5218 50 4 a a DET fcis-5218 50 5 technique technique NOUN fcis-5218 50 6 to to PART fcis-5218 50 7 study study VERB fcis-5218 50 8 how how SCONJ fcis-5218 50 9 to to PART fcis-5218 50 10 reuse reuse VERB fcis-5218 50 11 the the DET fcis-5218 50 12 learned learn VERB fcis-5218 50 13 knowledge knowledge NOUN fcis-5218 50 14 , , PUNCT fcis-5218 50 15 so so SCONJ fcis-5218 50 16 as as SCONJ fcis-5218 50 17 to to PART fcis-5218 50 18 avoid avoid VERB fcis-5218 50 19 training training NOUN fcis-5218 50 20 model model NOUN fcis-5218 50 21 from from ADP fcis-5218 50 22 scratch scratch NOUN fcis-5218 50 23 , , PUNCT fcis-5218 50 24 and and CCONJ fcis-5218 50 25 improve improve VERB fcis-5218 50 26 the the DET fcis-5218 50 27 training training NOUN fcis-5218 50 28 efficiency efficiency NOUN fcis-5218 50 29 and and CCONJ fcis-5218 50 30 stability stability NOUN fcis-5218 50 31 of of ADP fcis-5218 50 32 the the DET fcis-5218 50 33 model model NOUN fcis-5218 50 34 . . PUNCT fcis-5218 51 1 a a DET fcis-5218 51 2 typical typical ADJ fcis-5218 51 3 transfer transfer NOUN fcis-5218 51 4 learning learn VERB fcis-5218 51 5 approach approach NOUN fcis-5218 51 6 is be AUX fcis-5218 51 7 knowledge knowledge NOUN fcis-5218 51 8 transfer transfer NOUN fcis-5218 51 9 from from ADP fcis-5218 51 10 the the DET fcis-5218 51 11 pre pre ADJ fcis-5218 51 12 - - ADJ fcis-5218 51 13 trained train VERB fcis-5218 51 14 model model NOUN fcis-5218 51 15 . . PUNCT fcis-5218 52 1 this this DET fcis-5218 52 2 method method NOUN fcis-5218 52 3 first first ADV fcis-5218 52 4 pre pre NOUN fcis-5218 52 5 - - NOUN fcis-5218 52 6 trains train VERB fcis-5218 52 7 the the DET fcis-5218 52 8 network network NOUN fcis-5218 52 9 model model NOUN fcis-5218 52 10 with with ADP fcis-5218 52 11 a a DET fcis-5218 52 12 large large ADJ fcis-5218 52 13 amount amount NOUN fcis-5218 52 14 of of ADP fcis-5218 52 15 data datum NOUN fcis-5218 52 16 related relate VERB fcis-5218 52 17 to to ADP fcis-5218 52 18 the the DET fcis-5218 52 19 target target NOUN fcis-5218 52 20 task task NOUN fcis-5218 52 21 , , PUNCT fcis-5218 52 22 and and CCONJ fcis-5218 52 23 the the DET fcis-5218 52 24 learned learn VERB fcis-5218 52 25 model model NOUN fcis-5218 52 26 parameters parameter NOUN fcis-5218 52 27 , , PUNCT fcis-5218 52 28 namely namely ADV fcis-5218 52 29 the the DET fcis-5218 52 30 knowledge knowledge NOUN fcis-5218 52 31 , , PUNCT fcis-5218 52 32 are be AUX fcis-5218 52 33 preserved preserve VERB fcis-5218 52 34 . . PUNCT fcis-5218 53 1 based base VERB fcis-5218 53 2 on on ADP fcis-5218 53 3 the the DET fcis-5218 53 4 learned learn VERB fcis-5218 53 5 knowledge knowledge NOUN fcis-5218 53 6 , , PUNCT fcis-5218 53 7 the the DET fcis-5218 53 8 target target NOUN fcis-5218 53 9 data datum NOUN fcis-5218 53 10 set set VERB fcis-5218 53 11 is be AUX fcis-5218 53 12 trained train VERB fcis-5218 53 13 with with ADP fcis-5218 53 14 finetuning finetuning NOUN fcis-5218 53 15 , , PUNCT fcis-5218 53 16 so so SCONJ fcis-5218 53 17 as as SCONJ fcis-5218 53 18 to to PART fcis-5218 53 19 learn learn VERB fcis-5218 53 20 the the DET fcis-5218 53 21 knowledge knowledge NOUN fcis-5218 53 22 information information NOUN fcis-5218 53 23 of of ADP fcis-5218 53 24 the the DET fcis-5218 53 25 new new ADJ fcis-5218 53 26 data datum NOUN fcis-5218 53 27 set set VERB fcis-5218 53 28 . . PUNCT fcis-5218 54 1 3 3 X fcis-5218 54 2 . . NUM fcis-5218 54 3 proposed propose VERB fcis-5218 54 4 approach approach NOUN fcis-5218 54 5 this this DET fcis-5218 54 6 study study NOUN fcis-5218 54 7 proposes propose VERB fcis-5218 54 8 a a DET fcis-5218 54 9 tea tea NOUN fcis-5218 54 10 leaf leaf NOUN fcis-5218 54 11 disease disease NOUN fcis-5218 54 12 identification identification NOUN fcis-5218 54 13 method method NOUN fcis-5218 54 14 based base VERB fcis-5218 54 15 on on ADP fcis-5218 54 16 deep deep ADJ fcis-5218 54 17 transfer transfer NOUN fcis-5218 54 18 learning learning NOUN fcis-5218 54 19 , , PUNCT fcis-5218 54 20 which which PRON fcis-5218 54 21 is be AUX fcis-5218 54 22 then then ADV fcis-5218 54 23 introduced introduce VERB fcis-5218 54 24 from from ADP fcis-5218 54 25 the the DET fcis-5218 54 26 aspects aspect NOUN fcis-5218 54 27 of of ADP fcis-5218 54 28 image image NOUN fcis-5218 54 29 preprocessing preprocessing NOUN fcis-5218 54 30 , , PUNCT fcis-5218 54 31 feature feature NOUN fcis-5218 54 32 extraction extraction NOUN fcis-5218 54 33 , , PUNCT fcis-5218 54 34 transfer transfer NOUN fcis-5218 54 35 learning learning NOUN fcis-5218 54 36 and and CCONJ fcis-5218 54 37 loss loss NOUN fcis-5218 54 38 function function NOUN fcis-5218 54 39 respectively respectively ADV fcis-5218 54 40 . . PUNCT fcis-5218 55 1 3.1 3.1 NUM fcis-5218 55 2 . . PUNCT fcis-5218 55 3 image image NOUN fcis-5218 55 4 preprocessing preprocesse VERB fcis-5218 55 5 the the DET fcis-5218 55 6 data data NOUN fcis-5218 55 7 sets set NOUN fcis-5218 55 8 of of ADP fcis-5218 55 9 tea tea NOUN fcis-5218 55 10 leaf leaf NOUN fcis-5218 55 11 disease disease NOUN fcis-5218 55 12 selected select VERB fcis-5218 55 13 in in ADP fcis-5218 55 14 this this DET fcis-5218 55 15 study study NOUN fcis-5218 55 16 were be AUX fcis-5218 55 17 all all PRON fcis-5218 55 18 collected collect VERB fcis-5218 55 19 in in ADP fcis-5218 55 20 the the DET fcis-5218 55 21 tea tea NOUN fcis-5218 55 22 garden garden NOUN fcis-5218 55 23 in in ADP fcis-5218 55 24 tai'an tai'an PROPN fcis-5218 55 25 , , PUNCT fcis-5218 55 26 shandong shandong PROPN fcis-5218 55 27 province province PROPN fcis-5218 55 28 , , PUNCT fcis-5218 55 29 with with ADP fcis-5218 55 30 a a DET fcis-5218 55 31 total total NOUN fcis-5218 55 32 of of ADP fcis-5218 55 33 694 694 NUM fcis-5218 55 34 images image NOUN fcis-5218 55 35 , , PUNCT fcis-5218 55 36 including include VERB fcis-5218 55 37 141 141 NUM fcis-5218 55 38 images image NOUN fcis-5218 55 39 of of ADP fcis-5218 55 40 tea tea NOUN fcis-5218 55 41 white white PROPN fcis-5218 55 42 star star PROPN fcis-5218 55 43 , , PUNCT fcis-5218 55 44 107 107 NUM fcis-5218 55 45 images image NOUN fcis-5218 55 46 of of ADP fcis-5218 55 47 tea tea NOUN fcis-5218 55 48 leaf leaf NOUN fcis-5218 55 49 blight blight NOUN fcis-5218 55 50 , , PUNCT fcis-5218 55 51 186 186 NUM fcis-5218 55 52 images image NOUN fcis-5218 55 53 of of ADP fcis-5218 55 54 tea tea NOUN fcis-5218 55 55 wheel wheel NOUN fcis-5218 55 56 spot spot NOUN fcis-5218 55 57 and and CCONJ fcis-5218 55 58 260 260 NUM fcis-5218 55 59 images image NOUN fcis-5218 55 60 of of ADP fcis-5218 55 61 normal normal ADJ fcis-5218 55 62 leaves leave NOUN fcis-5218 55 63 , , PUNCT fcis-5218 55 64 with with ADP fcis-5218 55 65 the the DET fcis-5218 55 66 corresponding correspond VERB fcis-5218 55 67 label label NOUN fcis-5218 55 68 of of ADP fcis-5218 55 69 0 0 NUM fcis-5218 55 70 to to PART fcis-5218 55 71 3 3 NUM fcis-5218 55 72 . . PUNCT fcis-5218 56 1 the the DET fcis-5218 56 2 labels label NOUN fcis-5218 56 3 were be AUX fcis-5218 56 4 converted convert VERB fcis-5218 56 5 into into ADP fcis-5218 56 6 one one NUM fcis-5218 56 7 - - PUNCT fcis-5218 56 8 hot hot ADJ fcis-5218 56 9 vector vector NOUN fcis-5218 56 10 to to PART fcis-5218 56 11 correspond correspond VERB fcis-5218 56 12 to to ADP fcis-5218 56 13 the the DET fcis-5218 56 14 output output NOUN fcis-5218 56 15 of of ADP fcis-5218 56 16 the the DET fcis-5218 56 17 last last ADJ fcis-5218 56 18 layer layer NOUN fcis-5218 56 19 of of ADP fcis-5218 56 20 the the DET fcis-5218 56 21 neural neural ADJ fcis-5218 56 22 network network NOUN fcis-5218 56 23 . . PUNCT fcis-5218 57 1 for for ADP fcis-5218 57 2 the the DET fcis-5218 57 3 original original ADJ fcis-5218 57 4 label label NOUN fcis-5218 57 5 i i PRON fcis-5218 57 6 , , PUNCT fcis-5218 57 7 the the DET fcis-5218 57 8 length length NOUN fcis-5218 57 9 of of ADP fcis-5218 57 10 the the DET fcis-5218 57 11 transformed transform VERB fcis-5218 57 12 one one NUM fcis-5218 57 13 - - PUNCT fcis-5218 57 14 hot hot ADJ fcis-5218 57 15 vector vector NOUN fcis-5218 57 16 is be AUX fcis-5218 57 17 4 4 NUM fcis-5218 57 18 , , PUNCT fcis-5218 57 19 the the DET fcis-5218 57 20 i i PROPN fcis-5218 57 21 - - PUNCT fcis-5218 57 22 th th VERB fcis-5218 57 23 element element NOUN fcis-5218 57 24 is be AUX fcis-5218 57 25 1 1 NUM fcis-5218 57 26 , , PUNCT fcis-5218 57 27 and and CCONJ fcis-5218 57 28 the the DET fcis-5218 57 29 remaining remain VERB fcis-5218 57 30 values value NOUN fcis-5218 57 31 are be AUX fcis-5218 57 32 0 0 NUM fcis-5218 57 33 . . PUNCT fcis-5218 58 1 for for ADP fcis-5218 58 2 example example NOUN fcis-5218 58 3 , , PUNCT fcis-5218 58 4 the the DET fcis-5218 58 5 original original ADJ fcis-5218 58 6 label label NOUN fcis-5218 58 7 of of ADP fcis-5218 58 8 tea tea NOUN fcis-5218 58 9 wheel wheel NOUN fcis-5218 58 10 spot spot NOUN fcis-5218 58 11 disease disease NOUN fcis-5218 58 12 is be AUX fcis-5218 58 13 1 1 NUM fcis-5218 58 14 , , PUNCT fcis-5218 58 15 and and CCONJ fcis-5218 58 16 the the DET fcis-5218 58 17 label label NOUN fcis-5218 58 18 after after ADP fcis-5218 58 19 conversion conversion NOUN fcis-5218 58 20 to to ADP fcis-5218 58 21 the the DET fcis-5218 58 22 one one NUM fcis-5218 58 23 - - PUNCT fcis-5218 58 24 hot hot ADJ fcis-5218 58 25 form form NOUN fcis-5218 58 26 is be AUX fcis-5218 58 27 [ [ X fcis-5218 58 28 0,0,1,0 0,0,1,0 NUM fcis-5218 58 29 ] ] PUNCT fcis-5218 58 30 . . PUNCT fcis-5218 59 1 classes class NOUN fcis-5218 59 2 number number NOUN fcis-5218 59 3 of of ADP fcis-5218 59 4 original original ADJ fcis-5218 59 5 images image NOUN fcis-5218 59 6 number number NOUN fcis-5218 59 7 of of ADP fcis-5218 59 8 images image NOUN fcis-5218 59 9 after after ADP fcis-5218 59 10 preprocessing preprocesse VERB fcis-5218 59 11 tea tea NOUN fcis-5218 59 12 white white PROPN fcis-5218 59 13 star star PROPN fcis-5218 59 14 disease disease PROPN fcis-5218 59 15 141 141 NUM fcis-5218 59 16 1242 1242 NUM fcis-5218 59 17 leaf leaf NOUN fcis-5218 59 18 blight blight NOUN fcis-5218 59 19 disease disease NOUN fcis-5218 59 20 107 107 NUM fcis-5218 59 21 819 819 NUM fcis-5218 59 22 tea tea NOUN fcis-5218 59 23 wheel wheel NOUN fcis-5218 59 24 spot spot NOUN fcis-5218 59 25 disease disease NOUN fcis-5218 59 26 186 186 NUM fcis-5218 59 27 1228 1228 NUM fcis-5218 59 28 normal normal ADJ fcis-5218 59 29 leaves leave VERB fcis-5218 59 30 260 260 NUM fcis-5218 59 31 1376 1376 NUM fcis-5218 59 32 figure figure NOUN fcis-5218 59 33 1 1 NUM fcis-5218 59 34 . . PUNCT fcis-5218 59 35 sample sample NOUN fcis-5218 59 36 data datum NOUN fcis-5218 59 37 size size NOUN fcis-5218 59 38 before before ADP fcis-5218 59 39 and and CCONJ fcis-5218 59 40 after after ADP fcis-5218 59 41 data datum NOUN fcis-5218 59 42 augmentation augmentation NOUN fcis-5218 59 43 deep deep ADJ fcis-5218 59 44 learning learning NOUN fcis-5218 59 45 methods method NOUN fcis-5218 59 46 require require VERB fcis-5218 59 47 large large ADJ fcis-5218 59 48 amounts amount NOUN fcis-5218 59 49 of of ADP fcis-5218 59 50 samples sample NOUN fcis-5218 59 51 for for ADP fcis-5218 59 52 training training NOUN fcis-5218 59 53 , , PUNCT fcis-5218 59 54 while while SCONJ fcis-5218 59 55 the the DET fcis-5218 59 56 actual actual ADJ fcis-5218 59 57 amount amount NOUN fcis-5218 59 58 of of ADP fcis-5218 59 59 sample sample NOUN fcis-5218 59 60 data datum NOUN fcis-5218 59 61 collected collect VERB fcis-5218 59 62 in in ADP fcis-5218 59 63 tea tea NOUN fcis-5218 59 64 garden garden NOUN fcis-5218 59 65 is be AUX fcis-5218 59 66 obviously obviously ADV fcis-5218 59 67 insufficient insufficient ADJ fcis-5218 59 68 . . PUNCT fcis-5218 60 1 in in ADP fcis-5218 60 2 order order NOUN fcis-5218 60 3 to to PART fcis-5218 60 4 reduce reduce VERB fcis-5218 60 5 the the DET fcis-5218 60 6 problem problem NOUN fcis-5218 60 7 of of ADP fcis-5218 60 8 overfitting overfitte VERB fcis-5218 60 9 due due ADP fcis-5218 60 10 to to ADP fcis-5218 60 11 data datum NOUN fcis-5218 60 12 insufficiency insufficiency NOUN fcis-5218 60 13 , , PUNCT fcis-5218 60 14 data data NOUN fcis-5218 60 15 enhancement enhancement NOUN fcis-5218 60 16 technology technology NOUN fcis-5218 60 17 such such ADJ fcis-5218 60 18 as as ADP fcis-5218 60 19 random random ADJ fcis-5218 60 20 rotation rotation NOUN fcis-5218 60 21 , , PUNCT fcis-5218 60 22 random random ADJ fcis-5218 60 23 cropping cropping NOUN fcis-5218 60 24 , , PUNCT fcis-5218 60 25 random random ADJ fcis-5218 60 26 color color NOUN fcis-5218 60 27 change change NOUN fcis-5218 60 28 and and CCONJ fcis-5218 60 29 brightness brightness NOUN fcis-5218 60 30 adjustment adjustment NOUN fcis-5218 60 31 were be AUX fcis-5218 60 32 used use VERB fcis-5218 60 33 to to PART fcis-5218 60 34 expand expand VERB fcis-5218 60 35 the the DET fcis-5218 60 36 collected collected ADJ fcis-5218 60 37 tea tea NOUN fcis-5218 60 38 leaf leaf NOUN fcis-5218 60 39 images image NOUN fcis-5218 60 40 . . PUNCT fcis-5218 61 1 the the DET fcis-5218 61 2 total total NOUN fcis-5218 61 3 of of ADP fcis-5218 61 4 4665 4665 NUM fcis-5218 61 5 images image NOUN fcis-5218 61 6 were be AUX fcis-5218 61 7 obtained obtain VERB fcis-5218 61 8 through through ADP fcis-5218 61 9 data datum NOUN fcis-5218 61 10 enhancement enhancement NOUN fcis-5218 61 11 . . PUNCT fcis-5218 62 1 images image NOUN fcis-5218 62 2 of of ADP fcis-5218 62 3 each each DET fcis-5218 62 4 type type NOUN fcis-5218 62 5 was be AUX fcis-5218 62 6 randomly randomly ADV fcis-5218 62 7 divided divide VERB fcis-5218 62 8 into into ADP fcis-5218 62 9 training training NOUN fcis-5218 62 10 set set NOUN fcis-5218 62 11 and and CCONJ fcis-5218 62 12 test test NOUN fcis-5218 62 13 set set VERB fcis-5218 62 14 in in ADP fcis-5218 62 15 a a DET fcis-5218 62 16 ratio ratio NOUN fcis-5218 62 17 of of ADP fcis-5218 62 18 7:3 7:3 NUM fcis-5218 62 19 . . PROPN fcis-5218 63 1 3.2 3.2 NUM fcis-5218 63 2 . . PUNCT fcis-5218 64 1 feature feature NOUN fcis-5218 64 2 extraction extraction NOUN fcis-5218 64 3 in in ADP fcis-5218 64 4 this this DET fcis-5218 64 5 study study NOUN fcis-5218 64 6 , , PUNCT fcis-5218 64 7 the the DET fcis-5218 64 8 inception inception ADJ fcis-5218 64 9 v3 v3 PROPN fcis-5218 64 10 model model NOUN fcis-5218 64 11 was be AUX fcis-5218 64 12 used use VERB fcis-5218 64 13 for for ADP fcis-5218 64 14 feature feature NOUN fcis-5218 64 15 extraction extraction NOUN fcis-5218 64 16 . . PUNCT fcis-5218 65 1 the the DET fcis-5218 65 2 inception inception NOUN fcis-5218 65 3 module module NOUN fcis-5218 65 4 used use VERB fcis-5218 65 5 in in ADP fcis-5218 65 6 this this DET fcis-5218 65 7 model model NOUN fcis-5218 65 8 does do AUX fcis-5218 65 9 not not PART fcis-5218 65 10 only only ADV fcis-5218 65 11 pursue pursue VERB fcis-5218 65 12 the the DET fcis-5218 65 13 depth depth NOUN fcis-5218 65 14 of of ADP fcis-5218 65 15 the the DET fcis-5218 65 16 network network NOUN fcis-5218 65 17 like like ADP fcis-5218 65 18 most most ADV fcis-5218 65 19 popular popular ADJ fcis-5218 65 20 cnn cnn PROPN fcis-5218 65 21 model model NOUN fcis-5218 65 22 , , PUNCT fcis-5218 65 23 but but CCONJ fcis-5218 65 24 also also ADV fcis-5218 65 25 balances balance VERB fcis-5218 65 26 the the DET fcis-5218 65 27 width width ADJ fcis-5218 65 28 and and CCONJ fcis-5218 65 29 depth depth NOUN fcis-5218 65 30 of of ADP fcis-5218 65 31 the the DET fcis-5218 65 32 network network NOUN fcis-5218 65 33 , , PUNCT fcis-5218 65 34 which which PRON fcis-5218 65 35 helps help VERB fcis-5218 65 36 to to PART fcis-5218 65 37 enhance enhance VERB fcis-5218 65 38 the the DET fcis-5218 65 39 feature feature NOUN fcis-5218 65 40 extraction extraction NOUN fcis-5218 65 41 ability ability NOUN fcis-5218 65 42 of of ADP fcis-5218 65 43 the the DET fcis-5218 65 44 network network NOUN fcis-5218 65 45 . . PUNCT fcis-5218 66 1 at at ADP fcis-5218 66 2 the the DET fcis-5218 66 3 same same ADJ fcis-5218 66 4 time time NOUN fcis-5218 66 5 , , PUNCT fcis-5218 66 6 it it PRON fcis-5218 66 7 divides divide VERB fcis-5218 66 8 large large ADJ fcis-5218 66 9 convolution convolution NOUN fcis-5218 66 10 kernel kernel NOUN fcis-5218 66 11 into into ADP fcis-5218 66 12 smaller small ADJ fcis-5218 66 13 ones one NOUN fcis-5218 66 14 , , PUNCT fcis-5218 66 15 which which PRON fcis-5218 66 16 improves improve VERB fcis-5218 66 17 the the DET fcis-5218 66 18 training training NOUN fcis-5218 66 19 efficiency efficiency NOUN fcis-5218 66 20 of of ADP fcis-5218 66 21 the the DET fcis-5218 66 22 model model NOUN fcis-5218 66 23 by by ADP fcis-5218 66 24 reducing reduce VERB fcis-5218 66 25 the the DET fcis-5218 66 26 number number NOUN fcis-5218 66 27 of of ADP fcis-5218 66 28 parameters parameter NOUN fcis-5218 66 29 and and CCONJ fcis-5218 66 30 reduces reduce VERB fcis-5218 66 31 the the DET fcis-5218 66 32 problem problem NOUN fcis-5218 66 33 of of ADP fcis-5218 66 34 model model NOUN fcis-5218 66 35 overfitting overfitte VERB fcis-5218 66 36 as as ADV fcis-5218 66 37 well well ADV fcis-5218 66 38 . . PUNCT fcis-5218 67 1 3.3 3.3 NUM fcis-5218 67 2 . . PUNCT fcis-5218 68 1 transfer transfer NOUN fcis-5218 68 2 learning learning NOUN fcis-5218 68 3 strategy strategy NOUN fcis-5218 68 4 the the DET fcis-5218 68 5 classification classification NOUN fcis-5218 68 6 model model NOUN fcis-5218 68 7 of of ADP fcis-5218 68 8 tea tea NOUN fcis-5218 68 9 leaf leaf NOUN fcis-5218 68 10 disease disease NOUN fcis-5218 68 11 proposed propose VERB fcis-5218 68 12 in in ADP fcis-5218 68 13 this this DET fcis-5218 68 14 study study NOUN fcis-5218 68 15 consists consist VERB fcis-5218 68 16 of of ADP fcis-5218 68 17 the the DET fcis-5218 68 18 feature feature NOUN fcis-5218 68 19 extraction extraction NOUN fcis-5218 68 20 part part NOUN fcis-5218 68 21 and and CCONJ fcis-5218 68 22 classification classification NOUN fcis-5218 68 23 part part NOUN fcis-5218 68 24 . . PUNCT fcis-5218 69 1 the the DET fcis-5218 69 2 feature feature NOUN fcis-5218 69 3 extraction extraction NOUN fcis-5218 69 4 part part NOUN fcis-5218 69 5 uses use VERB fcis-5218 69 6 the the DET fcis-5218 69 7 inception inception ADJ fcis-5218 69 8 v3 v3 PROPN fcis-5218 69 9 - - PUNCT fcis-5218 69 10 based base VERB fcis-5218 69 11 network network NOUN fcis-5218 69 12 model model NOUN fcis-5218 69 13 for for ADP fcis-5218 69 14 transfer transfer NOUN fcis-5218 69 15 learning learning NOUN fcis-5218 69 16 , , PUNCT fcis-5218 69 17 and and CCONJ fcis-5218 69 18 the the DET fcis-5218 69 19 classification classification NOUN fcis-5218 69 20 part part NOUN fcis-5218 69 21 is be AUX fcis-5218 69 22 implemented implement VERB fcis-5218 69 23 based base VERB fcis-5218 69 24 on on ADP fcis-5218 69 25 the the DET fcis-5218 69 26 softmax softmax NOUN fcis-5218 69 27 layer layer NOUN fcis-5218 69 28 . . PUNCT fcis-5218 70 1 the the DET fcis-5218 70 2 training training NOUN fcis-5218 70 3 process process NOUN fcis-5218 70 4 is be AUX fcis-5218 70 5 described describe VERB fcis-5218 70 6 as as SCONJ fcis-5218 70 7 follows follow VERB fcis-5218 70 8 : : PUNCT fcis-5218 70 9 ( ( PUNCT fcis-5218 70 10 1 1 X fcis-5218 70 11 ) ) PUNCT fcis-5218 70 12 the the DET fcis-5218 70 13 network network NOUN fcis-5218 70 14 layer layer NOUN fcis-5218 70 15 before before ADP fcis-5218 70 16 the the DET fcis-5218 70 17 linear linear ADJ fcis-5218 70 18 connection connection NOUN fcis-5218 70 19 layer layer NOUN fcis-5218 70 20 of of ADP fcis-5218 70 21 the the DET fcis-5218 70 22 inception inception PROPN fcis-5218 70 23 v3 v3 PROPN fcis-5218 70 24 model model NOUN fcis-5218 70 25 is be AUX fcis-5218 70 26 defined define VERB fcis-5218 70 27 as as ADP fcis-5218 70 28 the the DET fcis-5218 70 29 bottleneck bottleneck NOUN fcis-5218 70 30 layer layer NOUN fcis-5218 70 31 . . PUNCT fcis-5218 71 1 the the DET fcis-5218 71 2 parameters parameter NOUN fcis-5218 71 3 of of ADP fcis-5218 71 4 the the DET fcis-5218 71 5 bottleneck bottleneck NOUN fcis-5218 71 6 layer layer NOUN fcis-5218 71 7 are be AUX fcis-5218 71 8 pre pre ADJ fcis-5218 71 9 - - VERB fcis-5218 71 10 trained train VERB fcis-5218 71 11 using use VERB fcis-5218 71 12 the the DET fcis-5218 71 13 public public ADJ fcis-5218 71 14 dataset dataset NOUN fcis-5218 71 15 imagenet imagenet NOUN fcis-5218 71 16 and and CCONJ fcis-5218 71 17 kept keep VERB fcis-5218 71 18 fixed fix VERB fcis-5218 71 19 in in ADP fcis-5218 71 20 the the DET fcis-5218 71 21 following follow VERB fcis-5218 71 22 model model NOUN fcis-5218 71 23 training training NOUN fcis-5218 71 24 . . PUNCT fcis-5218 72 1 the the DET fcis-5218 72 2 model model NOUN fcis-5218 72 3 pre pre VERB fcis-5218 72 4 - - VERB fcis-5218 72 5 trained train VERB fcis-5218 72 6 by by ADP fcis-5218 72 7 imagenet imagenet NOUN fcis-5218 72 8 can can AUX fcis-5218 72 9 help help VERB fcis-5218 72 10 the the DET fcis-5218 72 11 target target NOUN fcis-5218 72 12 task task NOUN fcis-5218 72 13 converge converge VERB fcis-5218 72 14 quickly quickly ADV fcis-5218 72 15 and and CCONJ fcis-5218 72 16 enhance enhance VERB fcis-5218 72 17 the the DET fcis-5218 72 18 generalization generalization NOUN fcis-5218 72 19 ability ability NOUN fcis-5218 72 20 of of ADP fcis-5218 72 21 identification identification NOUN fcis-5218 72 22 . . PUNCT fcis-5218 73 1 ( ( PUNCT fcis-5218 73 2 2 2 X fcis-5218 73 3 ) ) PUNCT fcis-5218 73 4 three three NUM fcis-5218 73 5 consecutive consecutive ADJ fcis-5218 73 6 fully fully ADV fcis-5218 73 7 connected connect VERB fcis-5218 73 8 layers layer NOUN fcis-5218 73 9 and and CCONJ fcis-5218 73 10 one one NUM fcis-5218 73 11 softmax softmax NOUN fcis-5218 73 12 layer layer NOUN fcis-5218 73 13 are be AUX fcis-5218 73 14 connected connect VERB fcis-5218 73 15 behind behind ADP fcis-5218 73 16 the the DET fcis-5218 73 17 bottleneck bottleneck NOUN fcis-5218 73 18 layer layer NOUN fcis-5218 73 19 . . PUNCT fcis-5218 74 1 the the DET fcis-5218 74 2 processed process VERB fcis-5218 74 3 images image NOUN fcis-5218 74 4 of of ADP fcis-5218 74 5 tea tea NOUN fcis-5218 74 6 leaf leaf NOUN fcis-5218 74 7 disease disease NOUN fcis-5218 74 8 are be AUX fcis-5218 74 9 used use VERB fcis-5218 74 10 as as ADP fcis-5218 74 11 the the DET fcis-5218 74 12 input input NOUN fcis-5218 74 13 of of ADP fcis-5218 74 14 the the DET fcis-5218 74 15 model model NOUN fcis-5218 74 16 . . PUNCT fcis-5218 75 1 parameters parameter NOUN fcis-5218 75 2 of of ADP fcis-5218 75 3 feature feature NOUN fcis-5218 75 4 extraction extraction NOUN fcis-5218 75 5 part part NOUN fcis-5218 75 6 are be AUX fcis-5218 75 7 frozen freeze VERB fcis-5218 75 8 , , PUNCT fcis-5218 75 9 only only ADV fcis-5218 75 10 parameters parameter NOUN fcis-5218 75 11 of of ADP fcis-5218 75 12 the the DET fcis-5218 75 13 subsequent subsequent ADJ fcis-5218 75 14 classification classification NOUN fcis-5218 75 15 part part NOUN fcis-5218 75 16 are be AUX fcis-5218 75 17 finetuned finetune VERB fcis-5218 75 18 . . PUNCT fcis-5218 76 1 in in ADP fcis-5218 76 2 this this DET fcis-5218 76 3 way way NOUN fcis-5218 76 4 , , PUNCT fcis-5218 76 5 the the DET fcis-5218 76 6 neural neural ADJ fcis-5218 76 7 network network NOUN fcis-5218 76 8 of of ADP fcis-5218 76 9 tea tea NOUN fcis-5218 76 10 leaf leaf NOUN fcis-5218 76 11 disease disease NOUN fcis-5218 76 12 identification identification NOUN fcis-5218 76 13 can can AUX fcis-5218 76 14 be be AUX fcis-5218 76 15 trained train VERB fcis-5218 76 16 more more ADV fcis-5218 76 17 quickly quickly ADV fcis-5218 76 18 and and CCONJ fcis-5218 76 19 accurately accurately ADV fcis-5218 76 20 . . PUNCT fcis-5218 77 1 3.4 3.4 NUM fcis-5218 77 2 . . PUNCT fcis-5218 77 3 loss loss NOUN fcis-5218 77 4 function function VERB fcis-5218 77 5 the the DET fcis-5218 77 6 loss loss NOUN fcis-5218 77 7 function function NOUN fcis-5218 77 8 is be AUX fcis-5218 77 9 used use VERB fcis-5218 77 10 to to PART fcis-5218 77 11 measure measure VERB fcis-5218 77 12 the the DET fcis-5218 77 13 difference difference NOUN fcis-5218 77 14 between between ADP fcis-5218 77 15 the the DET fcis-5218 77 16 predicted predict VERB fcis-5218 77 17 results result NOUN fcis-5218 77 18 and and CCONJ fcis-5218 77 19 the the DET fcis-5218 77 20 true true ADJ fcis-5218 77 21 labels label NOUN fcis-5218 77 22 , , PUNCT fcis-5218 77 23 and and CCONJ fcis-5218 77 24 updates update VERB fcis-5218 77 25 the the DET fcis-5218 77 26 77 77 NUM fcis-5218 77 27 model model NOUN fcis-5218 77 28 parameters parameter NOUN fcis-5218 77 29 by by ADP fcis-5218 77 30 back back NOUN fcis-5218 77 31 - - PUNCT fcis-5218 77 32 propagation propagation NOUN fcis-5218 77 33 to to PART fcis-5218 77 34 reduce reduce VERB fcis-5218 77 35 the the DET fcis-5218 77 36 prediction prediction NOUN fcis-5218 77 37 error error NOUN fcis-5218 77 38 . . PUNCT fcis-5218 78 1 cross cross ADJ fcis-5218 78 2 - - ADJ fcis-5218 78 3 entropy entropy ADJ fcis-5218 78 4 loss loss NOUN fcis-5218 78 5 function function NOUN fcis-5218 78 6 is be AUX fcis-5218 78 7 commonly commonly ADV fcis-5218 78 8 used use VERB fcis-5218 78 9 in in ADP fcis-5218 78 10 the the DET fcis-5218 78 11 classification classification NOUN fcis-5218 78 12 task task NOUN fcis-5218 78 13 . . PUNCT fcis-5218 79 1 the the DET fcis-5218 79 2 smaller small ADJ fcis-5218 79 3 the the DET fcis-5218 79 4 value value NOUN fcis-5218 79 5 of of ADP fcis-5218 79 6 the the DET fcis-5218 79 7 cross cross NOUN fcis-5218 79 8 - - ADJ fcis-5218 79 9 entropy entropy NOUN fcis-5218 79 10 indicates indicate VERB fcis-5218 79 11 , , PUNCT fcis-5218 79 12 the the PRON fcis-5218 79 13 closer close ADV fcis-5218 79 14 the the DET fcis-5218 79 15 actual actual ADJ fcis-5218 79 16 output output NOUN fcis-5218 79 17 of of ADP fcis-5218 79 18 the the DET fcis-5218 79 19 model model NOUN fcis-5218 79 20 is be AUX fcis-5218 79 21 related relate VERB fcis-5218 79 22 to to ADP fcis-5218 79 23 the the DET fcis-5218 79 24 expected expect VERB fcis-5218 79 25 result result NOUN fcis-5218 79 26 . . PUNCT fcis-5218 80 1 however however ADV fcis-5218 80 2 , , PUNCT fcis-5218 80 3 the the DET fcis-5218 80 4 crossentropy crossentropy NOUN fcis-5218 80 5 loss loss NOUN fcis-5218 80 6 function function NOUN fcis-5218 80 7 has have VERB fcis-5218 80 8 the the DET fcis-5218 80 9 same same ADJ fcis-5218 80 10 loss loss NOUN fcis-5218 80 11 weight weight NOUN fcis-5218 80 12 for for ADP fcis-5218 80 13 all all DET fcis-5218 80 14 classes class NOUN fcis-5218 80 15 , , PUNCT fcis-5218 80 16 which which PRON fcis-5218 80 17 is be AUX fcis-5218 80 18 not not PART fcis-5218 80 19 superior superior ADJ fcis-5218 80 20 in in ADP fcis-5218 80 21 this this DET fcis-5218 80 22 study study NOUN fcis-5218 80 23 with with ADP fcis-5218 80 24 an an DET fcis-5218 80 25 unbalanced unbalanced ADJ fcis-5218 80 26 sample sample NOUN fcis-5218 80 27 type type NOUN fcis-5218 80 28 . . PUNCT fcis-5218 81 1 therefore therefore ADV fcis-5218 81 2 , , PUNCT fcis-5218 81 3 the the DET fcis-5218 81 4 focal focal ADJ fcis-5218 81 5 loss loss NOUN fcis-5218 81 6 function function NOUN fcis-5218 81 7 was be AUX fcis-5218 81 8 introduced introduce VERB fcis-5218 81 9 instead instead ADV fcis-5218 81 10 of of ADP fcis-5218 81 11 the the DET fcis-5218 81 12 cross cross ADJ fcis-5218 81 13 - - ADJ fcis-5218 81 14 entropy entropy ADJ fcis-5218 81 15 loss loss NOUN fcis-5218 81 16 function function NOUN fcis-5218 81 17 , , PUNCT fcis-5218 81 18 and and CCONJ fcis-5218 81 19 the the DET fcis-5218 81 20 final final ADJ fcis-5218 81 21 loss loss NOUN fcis-5218 81 22 function function NOUN fcis-5218 81 23 is be AUX fcis-5218 81 24 : : PUNCT fcis-5218 81 25 fl(ΞΈ fl(ΞΈ PUNCT fcis-5218 81 26 ) ) PUNCT fcis-5218 81 27 = = SYM fcis-5218 81 28 βˆ’βˆ‘ βˆ’βˆ‘ PROPN fcis-5218 81 29 βˆ‘ βˆ‘ PUNCT fcis-5218 81 30 𝛼(1 𝛼(1 ADP fcis-5218 81 31 βˆ’ βˆ’ PROPN fcis-5218 81 32 οΏ½ οΏ½ PROPN fcis-5218 81 33 Μ‚ Μ‚ NOUN fcis-5218 81 34 οΏ½ οΏ½ NOUN fcis-5218 81 35 𝑐 𝑐 PROPN fcis-5218 81 36 𝑖)𝛾𝑦𝑐 𝑖)𝛾𝑦𝑐 PROPN fcis-5218 81 37 π‘–π‘™π‘œπ‘” π‘–π‘™π‘œπ‘” NOUN fcis-5218 81 38 𝐾 𝐾 PROPN fcis-5218 81 39 𝑐=1 𝑐=1 PROPN fcis-5218 81 40 𝑁 𝑁 PROPN fcis-5218 81 41 𝑖=1 𝑖=1 PROPN fcis-5218 81 42 οΏ½ οΏ½ PROPN fcis-5218 81 43 Μ‚ Μ‚ NOUN fcis-5218 81 44 οΏ½ οΏ½ NOUN fcis-5218 81 45 𝑐 𝑐 NOUN fcis-5218 81 46 𝑖 𝑖 PUNCT fcis-5218 81 47 whereΞΈrepresents whereΞΈrepresent NOUN fcis-5218 81 48 the the DET fcis-5218 81 49 parameters parameter NOUN fcis-5218 81 50 of of ADP fcis-5218 81 51 the the DET fcis-5218 81 52 model model NOUN fcis-5218 81 53 training training NOUN fcis-5218 81 54 , , PUNCT fcis-5218 81 55 n n X fcis-5218 81 56 is be AUX fcis-5218 81 57 the the DET fcis-5218 81 58 number number NOUN fcis-5218 81 59 of of ADP fcis-5218 81 60 training training NOUN fcis-5218 81 61 samples sample NOUN fcis-5218 81 62 , , PUNCT fcis-5218 81 63 k k X fcis-5218 81 64 is be AUX fcis-5218 81 65 the the DET fcis-5218 81 66 total total ADJ fcis-5218 81 67 number number NOUN fcis-5218 81 68 of of ADP fcis-5218 81 69 classifications classification NOUN fcis-5218 81 70 , , PUNCT fcis-5218 81 71 y y PROPN fcis-5218 81 72 and and CCONJ fcis-5218 81 73 yΜ‚ yΜ‚ NUM fcis-5218 81 74 represents represent VERB fcis-5218 81 75 the the DET fcis-5218 81 76 true true ADJ fcis-5218 81 77 labels label NOUN fcis-5218 81 78 and and CCONJ fcis-5218 81 79 the the DET fcis-5218 81 80 predictions prediction NOUN fcis-5218 81 81 respectively respectively ADV fcis-5218 81 82 , , PUNCT fcis-5218 82 1 yc yc PRON fcis-5218 82 2 i i PRON fcis-5218 82 3 representing represent VERB fcis-5218 82 4 the the DET fcis-5218 82 5 c c NOUN fcis-5218 82 6 - - PUNCT fcis-5218 82 7 th th VERB fcis-5218 82 8 scalar scalar ADJ fcis-5218 82 9 value value NOUN fcis-5218 82 10 in in ADP fcis-5218 82 11 the the DET fcis-5218 82 12 true true ADJ fcis-5218 82 13 label label NOUN fcis-5218 82 14 corresponding correspond VERB fcis-5218 82 15 to to ADP fcis-5218 82 16 the the DET fcis-5218 82 17 i i PROPN fcis-5218 82 18 - - PUNCT fcis-5218 82 19 th th X fcis-5218 82 20 sample sample NOUN fcis-5218 82 21 , , PUNCT fcis-5218 82 22 and and CCONJ fcis-5218 82 23 yΜ‚c yΜ‚c PRON fcis-5218 82 24 i i PRON fcis-5218 82 25 representing represent VERB fcis-5218 82 26 the the DET fcis-5218 82 27 c c NOUN fcis-5218 82 28 - - PUNCT fcis-5218 82 29 th th VERB fcis-5218 82 30 scalar scalar ADJ fcis-5218 82 31 value value NOUN fcis-5218 82 32 in in ADP fcis-5218 82 33 the the DET fcis-5218 82 34 results result NOUN fcis-5218 82 35 predicted predict VERB fcis-5218 82 36 by by ADP fcis-5218 82 37 the the DET fcis-5218 82 38 i i PROPN fcis-5218 82 39 - - PUNCT fcis-5218 82 40 th th VERB fcis-5218 82 41 sample sample NOUN fcis-5218 82 42 . . PUNCT fcis-5218 83 1 the the DET fcis-5218 83 2 balance balance NOUN fcis-5218 83 3 factorΞ± factorΞ± NOUN fcis-5218 83 4 is be AUX fcis-5218 83 5 used use VERB fcis-5218 83 6 to to PART fcis-5218 83 7 solve solve VERB fcis-5218 83 8 the the DET fcis-5218 83 9 problem problem NOUN fcis-5218 83 10 of of ADP fcis-5218 83 11 unbalanced unbalanced ADJ fcis-5218 83 12 distribution distribution NOUN fcis-5218 83 13 of of ADP fcis-5218 83 14 different different ADJ fcis-5218 83 15 categories category NOUN fcis-5218 83 16 of of ADP fcis-5218 83 17 data datum NOUN fcis-5218 83 18 , , PUNCT fcis-5218 83 19 Ξ³ Ξ³ X fcis-5218 83 20 is be AUX fcis-5218 83 21 used use VERB fcis-5218 83 22 to to PART fcis-5218 83 23 reduce reduce VERB fcis-5218 83 24 the the DET fcis-5218 83 25 loss loss NOUN fcis-5218 83 26 of of ADP fcis-5218 83 27 easy easy ADJ fcis-5218 83 28 to to PART fcis-5218 83 29 distinguish distinguish VERB fcis-5218 83 30 samples sample NOUN fcis-5218 83 31 , , PUNCT fcis-5218 83 32 and and CCONJ fcis-5218 83 33 to to PART fcis-5218 83 34 improve improve VERB fcis-5218 83 35 the the DET fcis-5218 83 36 importance importance NOUN fcis-5218 83 37 of of ADP fcis-5218 83 38 difficult difficult ADJ fcis-5218 83 39 to to PART fcis-5218 83 40 identify identify VERB fcis-5218 83 41 samples sample NOUN fcis-5218 83 42 , , PUNCT fcis-5218 83 43 so so SCONJ fcis-5218 83 44 that that SCONJ fcis-5218 83 45 the the DET fcis-5218 83 46 model model NOUN fcis-5218 83 47 can can AUX fcis-5218 83 48 be be AUX fcis-5218 83 49 trained train VERB fcis-5218 83 50 to to PART fcis-5218 83 51 extract extract VERB fcis-5218 83 52 more more ADV fcis-5218 83 53 important important ADJ fcis-5218 83 54 information information NOUN fcis-5218 83 55 hidden hide VERB fcis-5218 83 56 in in ADP fcis-5218 83 57 the the DET fcis-5218 83 58 data datum NOUN fcis-5218 83 59 . . PUNCT fcis-5218 84 1 figure figure NOUN fcis-5218 84 2 2 2 NUM fcis-5218 84 3 . . PUNCT fcis-5218 84 4 comparison comparison NOUN fcis-5218 84 5 of of ADP fcis-5218 84 6 training training NOUN fcis-5218 84 7 accuracy accuracy NOUN fcis-5218 84 8 and and CCONJ fcis-5218 84 9 validation validation NOUN fcis-5218 84 10 accuracy accuracy NOUN fcis-5218 84 11 based base VERB fcis-5218 84 12 on on ADP fcis-5218 84 13 random random ADJ fcis-5218 84 14 parameters parameter NOUN fcis-5218 84 15 and and CCONJ fcis-5218 84 16 transfer transfer VERB fcis-5218 84 17 learning learn VERB fcis-5218 84 18 4 4 NUM fcis-5218 84 19 . . PUNCT fcis-5218 84 20 experimental experimental ADJ fcis-5218 84 21 results result NOUN fcis-5218 84 22 and and CCONJ fcis-5218 84 23 analysis analysis NOUN fcis-5218 84 24 to to PART fcis-5218 84 25 verify verify VERB fcis-5218 84 26 the the DET fcis-5218 84 27 algorithm algorithm NOUN fcis-5218 84 28 of of ADP fcis-5218 84 29 this this DET fcis-5218 84 30 study study NOUN fcis-5218 84 31 , , PUNCT fcis-5218 84 32 several several ADJ fcis-5218 84 33 experiments experiment NOUN fcis-5218 84 34 were be AUX fcis-5218 84 35 performed perform VERB fcis-5218 84 36 . . PUNCT fcis-5218 85 1 the the DET fcis-5218 85 2 input input NOUN fcis-5218 85 3 image image NOUN fcis-5218 85 4 size size NOUN fcis-5218 85 5 was be AUX fcis-5218 85 6 uniformly uniformly ADV fcis-5218 85 7 scaled scale VERB fcis-5218 85 8 to to ADP fcis-5218 85 9 299Γ—299 299Γ—299 NUM fcis-5218 85 10 , , PUNCT fcis-5218 85 11 the the DET fcis-5218 85 12 learning learning NOUN fcis-5218 85 13 rate rate NOUN fcis-5218 85 14 was be AUX fcis-5218 85 15 initialized initialize VERB fcis-5218 85 16 to to ADP fcis-5218 85 17 0.001 0.001 NUM fcis-5218 85 18 , , PUNCT fcis-5218 85 19 and and CCONJ fcis-5218 85 20 the the DET fcis-5218 85 21 loss loss NOUN fcis-5218 85 22 function function NOUN fcis-5218 85 23 was be AUX fcis-5218 85 24 optimized optimize VERB fcis-5218 85 25 by by ADP fcis-5218 85 26 using use VERB fcis-5218 85 27 the the DET fcis-5218 85 28 small small ADJ fcis-5218 85 29 - - PUNCT fcis-5218 85 30 batch batch NOUN fcis-5218 85 31 stochastic stochastic ADJ fcis-5218 85 32 gradient gradient ADJ fcis-5218 85 33 descent descent NOUN fcis-5218 85 34 algorithm algorithm NOUN fcis-5218 85 35 for for ADP fcis-5218 85 36 model model NOUN fcis-5218 85 37 training training NOUN fcis-5218 85 38 . . PUNCT fcis-5218 86 1 figure figure NOUN fcis-5218 86 2 2 2 NUM fcis-5218 86 3 shows show VERB fcis-5218 86 4 the the DET fcis-5218 86 5 comparison comparison NOUN fcis-5218 86 6 of of ADP fcis-5218 86 7 training training NOUN fcis-5218 86 8 accuracy accuracy NOUN fcis-5218 86 9 and and CCONJ fcis-5218 86 10 validation validation NOUN fcis-5218 86 11 accuracy accuracy NOUN fcis-5218 86 12 based base VERB fcis-5218 86 13 on on ADP fcis-5218 86 14 random random ADJ fcis-5218 86 15 parameters parameter NOUN fcis-5218 86 16 and and CCONJ fcis-5218 86 17 transfer transfer NOUN fcis-5218 86 18 learning learning NOUN fcis-5218 86 19 . . PUNCT fcis-5218 87 1 it it PRON fcis-5218 87 2 can can AUX fcis-5218 87 3 be be AUX fcis-5218 87 4 seen see VERB fcis-5218 87 5 that that SCONJ fcis-5218 87 6 transfer transfer NOUN fcis-5218 87 7 learning learning NOUN fcis-5218 87 8 is be AUX fcis-5218 87 9 trained train VERB fcis-5218 87 10 based base VERB fcis-5218 87 11 on on ADP fcis-5218 87 12 the the DET fcis-5218 87 13 learned learn VERB fcis-5218 87 14 knowledge knowledge NOUN fcis-5218 87 15 , , PUNCT fcis-5218 87 16 which which PRON fcis-5218 87 17 can can AUX fcis-5218 87 18 make make VERB fcis-5218 87 19 the the DET fcis-5218 87 20 model model NOUN fcis-5218 87 21 converge converge VERB fcis-5218 87 22 faster fast ADV fcis-5218 87 23 . . PUNCT fcis-5218 88 1 the the DET fcis-5218 88 2 training training NOUN fcis-5218 88 3 accuracy accuracy NOUN fcis-5218 88 4 and and CCONJ fcis-5218 88 5 verification verification NOUN fcis-5218 88 6 accuracy accuracy NOUN fcis-5218 88 7 are be AUX fcis-5218 88 8 higher high ADJ fcis-5218 88 9 , , PUNCT fcis-5218 88 10 which which PRON fcis-5218 88 11 verifies verify VERB fcis-5218 88 12 the the DET fcis-5218 88 13 effectiveness effectiveness NOUN fcis-5218 88 14 of of ADP fcis-5218 88 15 transfer transfer NOUN fcis-5218 88 16 learning learn VERB fcis-5218 88 17 strategy strategy NOUN fcis-5218 88 18 in in ADP fcis-5218 88 19 the the DET fcis-5218 88 20 identification identification NOUN fcis-5218 88 21 of of ADP fcis-5218 88 22 tea tea NOUN fcis-5218 88 23 leaf leaf NOUN fcis-5218 88 24 diseases disease NOUN fcis-5218 88 25 . . PUNCT fcis-5218 89 1 this this DET fcis-5218 89 2 effectiveness effectiveness NOUN fcis-5218 89 3 benefits benefit NOUN fcis-5218 89 4 from from ADP fcis-5218 89 5 the the DET fcis-5218 89 6 pre pre ADJ fcis-5218 89 7 - - ADJ fcis-5218 89 8 trained train VERB fcis-5218 89 9 model model NOUN fcis-5218 89 10 that that PRON fcis-5218 89 11 already already ADV fcis-5218 89 12 has have VERB fcis-5218 89 13 the the DET fcis-5218 89 14 ability ability NOUN fcis-5218 89 15 to to PART fcis-5218 89 16 identify identify VERB fcis-5218 89 17 the the DET fcis-5218 89 18 geometry geometry NOUN fcis-5218 89 19 , , PUNCT fcis-5218 89 20 shape shape NOUN fcis-5218 89 21 , , PUNCT fcis-5218 89 22 attitude attitude NOUN fcis-5218 89 23 and and CCONJ fcis-5218 89 24 other other ADJ fcis-5218 89 25 information information NOUN fcis-5218 89 26 of of ADP fcis-5218 89 27 the the DET fcis-5218 89 28 object object NOUN fcis-5218 89 29 . . PUNCT fcis-5218 90 1 this this DET fcis-5218 90 2 prior prior ADJ fcis-5218 90 3 knowledge knowledge NOUN fcis-5218 90 4 can can AUX fcis-5218 90 5 be be AUX fcis-5218 90 6 very very ADV fcis-5218 90 7 quickly quickly ADV fcis-5218 90 8 transferred transfer VERB fcis-5218 90 9 to to ADP fcis-5218 90 10 the the DET fcis-5218 90 11 recognition recognition NOUN fcis-5218 90 12 task task NOUN fcis-5218 90 13 of of ADP fcis-5218 90 14 tea tea NOUN fcis-5218 90 15 leaf leaf NOUN fcis-5218 90 16 disease disease NOUN fcis-5218 90 17 images image NOUN fcis-5218 90 18 . . PUNCT fcis-5218 91 1 table table NOUN fcis-5218 91 2 1 1 NUM fcis-5218 91 3 . . PUNCT fcis-5218 91 4 comparison comparison NOUN fcis-5218 91 5 of of ADP fcis-5218 91 6 the the DET fcis-5218 91 7 classification classification NOUN fcis-5218 91 8 accuracy accuracy NOUN fcis-5218 91 9 based base VERB fcis-5218 91 10 on on ADP fcis-5218 91 11 cross cross PROPN fcis-5218 91 12 entropy entropy PROPN fcis-5218 91 13 loss loss NOUN fcis-5218 91 14 function function NOUN fcis-5218 91 15 and and CCONJ fcis-5218 91 16 focal focal ADJ fcis-5218 91 17 loss loss NOUN fcis-5218 91 18 function function NOUN fcis-5218 91 19 loss loss NOUN fcis-5218 91 20 function function NOUN fcis-5218 91 21 type type NOUN fcis-5218 91 22 top1 top1 PROPN fcis-5218 91 23 accuracy accuracy NOUN fcis-5218 91 24 ( ( PUNCT fcis-5218 91 25 % % INTJ fcis-5218 91 26 ) ) PUNCT fcis-5218 91 27 cross cross ADJ fcis-5218 91 28 - - ADJ fcis-5218 91 29 entropy entropy ADJ fcis-5218 91 30 loss loss NOUN fcis-5218 91 31 function function NOUN fcis-5218 91 32 89.75 89.75 NUM fcis-5218 91 33 focus focus NOUN fcis-5218 91 34 loss loss NOUN fcis-5218 91 35 function function NOUN fcis-5218 91 36 90.42 90.42 NUM fcis-5218 91 37 table table NOUN fcis-5218 91 38 1 1 NUM fcis-5218 91 39 compares compare VERB fcis-5218 91 40 the the DET fcis-5218 91 41 top1 top1 PROPN fcis-5218 91 42 classification classification NOUN fcis-5218 91 43 accuracy accuracy NOUN fcis-5218 91 44 of of ADP fcis-5218 91 45 the the DET fcis-5218 91 46 tea tea NOUN fcis-5218 91 47 leaf leaf NOUN fcis-5218 91 48 disease disease NOUN fcis-5218 91 49 using use VERB fcis-5218 91 50 the the DET fcis-5218 91 51 cross cross ADJ fcis-5218 91 52 - - ADJ fcis-5218 91 53 entropy entropy ADJ fcis-5218 91 54 loss loss NOUN fcis-5218 91 55 function function NOUN fcis-5218 91 56 and and CCONJ fcis-5218 91 57 the the DET fcis-5218 91 58 focus focus NOUN fcis-5218 91 59 loss loss NOUN fcis-5218 91 60 function function NOUN fcis-5218 91 61 . . PUNCT fcis-5218 92 1 it it PRON fcis-5218 92 2 can can AUX fcis-5218 92 3 be be AUX fcis-5218 92 4 seen see VERB fcis-5218 92 5 that that SCONJ fcis-5218 92 6 the the DET fcis-5218 92 7 accuracy accuracy NOUN fcis-5218 92 8 using use VERB fcis-5218 92 9 the the DET fcis-5218 92 10 focal focal ADJ fcis-5218 92 11 loss loss NOUN fcis-5218 92 12 function function NOUN fcis-5218 92 13 is be AUX fcis-5218 92 14 promoted promote VERB fcis-5218 92 15 by by ADP fcis-5218 92 16 0.67 0.67 NUM fcis-5218 92 17 % % NOUN fcis-5218 92 18 compared compare VERB fcis-5218 92 19 with with ADP fcis-5218 92 20 the the DET fcis-5218 92 21 cross cross ADJ fcis-5218 92 22 - - ADJ fcis-5218 92 23 entropy entropy ADJ fcis-5218 92 24 loss loss NOUN fcis-5218 92 25 function function NOUN fcis-5218 92 26 , , PUNCT fcis-5218 92 27 which which PRON fcis-5218 92 28 proves prove VERB fcis-5218 92 29 the the DET fcis-5218 92 30 effectiveness effectiveness NOUN fcis-5218 92 31 of of ADP fcis-5218 92 32 the the DET fcis-5218 92 33 improved improve VERB fcis-5218 92 34 loss loss NOUN fcis-5218 92 35 function function NOUN fcis-5218 92 36 . . PUNCT fcis-5218 93 1 5 5 X fcis-5218 93 2 . . X fcis-5218 93 3 summary summary NOUN fcis-5218 93 4 based base VERB fcis-5218 93 5 on on ADP fcis-5218 93 6 the the DET fcis-5218 93 7 strategy strategy NOUN fcis-5218 93 8 of of ADP fcis-5218 93 9 deep deep ADJ fcis-5218 93 10 transfer transfer NOUN fcis-5218 93 11 learning learning NOUN fcis-5218 93 12 , , PUNCT fcis-5218 93 13 this this DET fcis-5218 93 14 study study NOUN fcis-5218 93 15 constructed construct VERB fcis-5218 93 16 a a DET fcis-5218 93 17 deep deep ADJ fcis-5218 93 18 learning learning NOUN fcis-5218 93 19 network network NOUN fcis-5218 93 20 for for ADP fcis-5218 93 21 the the DET fcis-5218 93 22 recognition recognition NOUN fcis-5218 93 23 of of ADP fcis-5218 93 24 tea tea NOUN fcis-5218 93 25 leaf leaf NOUN fcis-5218 93 26 disease disease NOUN fcis-5218 93 27 . . PUNCT fcis-5218 94 1 the the DET fcis-5218 94 2 focus focus NOUN fcis-5218 94 3 loss loss NOUN fcis-5218 94 4 function function NOUN fcis-5218 94 5 was be AUX fcis-5218 94 6 used use VERB fcis-5218 94 7 to to PART fcis-5218 94 8 solve solve VERB fcis-5218 94 9 the the DET fcis-5218 94 10 problem problem NOUN fcis-5218 94 11 of of ADP fcis-5218 94 12 unbalanced unbalanced ADJ fcis-5218 94 13 data datum NOUN fcis-5218 94 14 distribution distribution NOUN fcis-5218 94 15 in in ADP fcis-5218 94 16 samples sample NOUN fcis-5218 94 17 . . PUNCT fcis-5218 95 1 the the DET fcis-5218 95 2 accuracy accuracy NOUN fcis-5218 95 3 could could AUX fcis-5218 95 4 reach reach VERB fcis-5218 95 5 more more ADJ fcis-5218 95 6 than than ADP fcis-5218 95 7 90.42 90.42 NUM fcis-5218 95 8 % % NOUN fcis-5218 95 9 , , PUNCT fcis-5218 95 10 which which PRON fcis-5218 95 11 verified verify VERB fcis-5218 95 12 the the DET fcis-5218 95 13 effectiveness effectiveness NOUN fcis-5218 95 14 of of ADP fcis-5218 95 15 transfer transfer NOUN fcis-5218 95 16 learning learning NOUN fcis-5218 95 17 in in ADP fcis-5218 95 18 the the DET fcis-5218 95 19 application application NOUN fcis-5218 95 20 of of ADP fcis-5218 95 21 tea tea NOUN fcis-5218 95 22 leaf leaf NOUN fcis-5218 95 23 disease disease NOUN fcis-5218 95 24 recognition recognition NOUN fcis-5218 95 25 . . PUNCT fcis-5218 96 1 this this DET fcis-5218 96 2 study study NOUN fcis-5218 96 3 will will AUX fcis-5218 96 4 provide provide VERB fcis-5218 96 5 important important ADJ fcis-5218 96 6 theoretical theoretical ADJ fcis-5218 96 7 support support NOUN fcis-5218 96 8 for for ADP fcis-5218 96 9 further further ADJ fcis-5218 96 10 achievements achievement NOUN fcis-5218 96 11 transformation transformation NOUN fcis-5218 96 12 , , PUNCT fcis-5218 96 13 and and CCONJ fcis-5218 96 14 has have VERB fcis-5218 96 15 important important ADJ fcis-5218 96 16 theoretical theoretical ADJ fcis-5218 96 17 and and CCONJ fcis-5218 96 18 practical practical ADJ fcis-5218 96 19 significance significance NOUN fcis-5218 96 20 in in ADP fcis-5218 96 21 the the DET fcis-5218 96 22 development development NOUN fcis-5218 96 23 of of ADP fcis-5218 96 24 intelligent intelligent ADJ fcis-5218 96 25 agriculture agriculture NOUN fcis-5218 96 26 . . PUNCT fcis-5218 97 1 acknowledgments acknowledgment NOUN fcis-5218 97 2 this this DET fcis-5218 97 3 work work NOUN fcis-5218 97 4 was be AUX fcis-5218 97 5 supported support VERB fcis-5218 97 6 by by ADP fcis-5218 97 7 the the DET fcis-5218 97 8 tai'an tai'an PROPN fcis-5218 97 9 science science NOUN fcis-5218 97 10 and and CCONJ fcis-5218 97 11 technology technology NOUN fcis-5218 97 12 innovation innovation NOUN fcis-5218 97 13 development development NOUN fcis-5218 97 14 project project NOUN fcis-5218 97 15 under under ADP fcis-5218 97 16 grant grant NOUN fcis-5218 97 17 no.2021ns097 no.2021ns097 PROPN fcis-5218 97 18 . . PUNCT fcis-5218 98 1 references reference NOUN fcis-5218 98 2 [ [ X fcis-5218 98 3 1 1 X fcis-5218 98 4 ] ] X fcis-5218 98 5 hossain hossain PROPN fcis-5218 98 6 s s PROPN fcis-5218 98 7 , , PUNCT fcis-5218 98 8 mou mou PROPN fcis-5218 98 9 r r PROPN fcis-5218 98 10 m m PROPN fcis-5218 98 11 , , PUNCT fcis-5218 98 12 hasan hasan PROPN fcis-5218 98 13 m m PROPN fcis-5218 98 14 m m PROPN fcis-5218 98 15 , , PUNCT fcis-5218 98 16 et et NOUN fcis-5218 98 17 al.recognition al.recognition NOUN fcis-5218 98 18 and and CCONJ fcis-5218 98 19 detection detection NOUN fcis-5218 98 20 of of ADP fcis-5218 98 21 tea tea NOUN fcis-5218 98 22 leaf leaf NOUN fcis-5218 98 23 's 's PART fcis-5218 98 24 diseases disease NOUN fcis-5218 98 25 using use VERB fcis-5218 98 26 support support NOUN fcis-5218 98 27 vector vector NOUN fcis-5218 98 28 machine[c]//2018 machine[c]//2018 PROPN fcis-5218 98 29 ieee ieee NOUN fcis-5218 98 30 14th 14th ADJ fcis-5218 98 31 international international ADJ fcis-5218 98 32 colloquium colloquium NOUN fcis-5218 98 33 on on ADP fcis-5218 98 34 signal signal NOUN fcis-5218 98 35 processing processing NOUN fcis-5218 98 36 & & CCONJ fcis-5218 98 37 its its PRON fcis-5218 98 38 applications application NOUN fcis-5218 98 39 ( ( PUNCT fcis-5218 98 40 cspa).ieee cspa).ieee NOUN fcis-5218 98 41 , , PUNCT fcis-5218 98 42 2018 2018 NUM fcis-5218 98 43 : : PUNCT fcis-5218 98 44 150154 150154 NUM fcis-5218 98 45 . . PUNCT fcis-5218 99 1 [ [ X fcis-5218 99 2 2 2 NUM fcis-5218 99 3 ] ] X fcis-5218 99 4 shao shao PROPN fcis-5218 99 5 p p PROPN fcis-5218 99 6 , , PUNCT fcis-5218 99 7 wu wu PROPN fcis-5218 99 8 m m PROPN fcis-5218 99 9 , , PUNCT fcis-5218 99 10 wang wang PROPN fcis-5218 99 11 x x PROPN fcis-5218 99 12 , , PUNCT fcis-5218 99 13 et et PROPN fcis-5218 99 14 al.research al.research ADV fcis-5218 99 15 on on ADP fcis-5218 99 16 the the DET fcis-5218 99 17 tea tea NOUN fcis-5218 99 18 bud bud NOUN fcis-5218 99 19 recognition recognition NOUN fcis-5218 99 20 based base VERB fcis-5218 99 21 on on ADP fcis-5218 99 22 improved improved ADJ fcis-5218 99 23 k k PROPN fcis-5218 99 24 - - PUNCT fcis-5218 99 25 means mean VERB fcis-5218 99 26 algorithm[c]//matec algorithm[c]//matec PROPN fcis-5218 99 27 web web NOUN fcis-5218 99 28 of of ADP fcis-5218 99 29 conferences.edp conferences.edp X fcis-5218 99 30 sciences science NOUN fcis-5218 99 31 , , PUNCT fcis-5218 99 32 2018 2018 NUM fcis-5218 99 33 , , PUNCT fcis-5218 99 34 232 232 NUM fcis-5218 99 35 : : SYM fcis-5218 99 36 03050 03050 NUM fcis-5218 99 37 . . PUNCT fcis-5218 100 1 [ [ X fcis-5218 100 2 3 3 X fcis-5218 100 3 ] ] X fcis-5218 100 4 xiaoxiao xiaoxiao PROPN fcis-5218 100 5 s s PROPN fcis-5218 100 6 u u NOUN fcis-5218 100 7 n n CCONJ fcis-5218 100 8 , , PUNCT fcis-5218 100 9 shaomin shaomin ADJ fcis-5218 100 10 m m PROPN fcis-5218 100 11 u u NOUN fcis-5218 100 12 , , PUNCT fcis-5218 100 13 yongyu yongyu PROPN fcis-5218 100 14 x x SYM fcis-5218 100 15 u u NOUN fcis-5218 100 16 , , PUNCT fcis-5218 100 17 et et PROPN fcis-5218 100 18 al.image al.image NOUN fcis-5218 100 19 recognition recognition NOUN fcis-5218 100 20 of of ADP fcis-5218 100 21 tea tea NOUN fcis-5218 100 22 leaf leaf NOUN fcis-5218 100 23 diseases disease NOUN fcis-5218 100 24 based base VERB fcis-5218 100 25 on on ADP fcis-5218 100 26 convolutional convolutional ADJ fcis-5218 100 27 neural neural ADJ fcis-5218 100 28 network[c]//2018 network[c]//2018 PROPN fcis-5218 100 29 international international ADJ fcis-5218 100 30 conference conference NOUN fcis-5218 100 31 on on ADP fcis-5218 100 32 security security NOUN fcis-5218 100 33 , , PUNCT fcis-5218 100 34 pattern pattern NOUN fcis-5218 100 35 analysis analysis NOUN fcis-5218 100 36 , , PUNCT fcis-5218 100 37 and and CCONJ fcis-5218 100 38 cybernetics cybernetic NOUN fcis-5218 100 39 ( ( PUNCT fcis-5218 100 40 spac).ieee spac).ieee NOUN fcis-5218 100 41 , , PUNCT fcis-5218 100 42 2018 2018 NUM fcis-5218 100 43 : : PUNCT fcis-5218 100 44 304309 304309 NUM fcis-5218 100 45 . . PUNCT fcis-5218 101 1 [ [ X fcis-5218 101 2 4 4 X fcis-5218 101 3 ] ] X fcis-5218 101 4 chen chen PROPN fcis-5218 101 5 j j PROPN fcis-5218 101 6 , , PUNCT fcis-5218 101 7 liu liu PROPN fcis-5218 101 8 q q PROPN fcis-5218 101 9 , , PUNCT fcis-5218 101 10 gao gao PROPN fcis-5218 101 11 l.visual l.visual ADJ fcis-5218 101 12 tea tea NOUN fcis-5218 101 13 leaf leaf NOUN fcis-5218 101 14 disease disease NOUN fcis-5218 101 15 recognition recognition NOUN fcis-5218 101 16 using use VERB fcis-5218 101 17 a a DET fcis-5218 101 18 convolutional convolutional ADJ fcis-5218 101 19 neural neural ADJ fcis-5218 101 20 network network NOUN fcis-5218 101 21 model[j].symmetry model[j].symmetry NOUN fcis-5218 101 22 , , PUNCT fcis-5218 101 23 2019 2019 NUM fcis-5218 101 24 , , PUNCT fcis-5218 101 25 11(3 11(3 NUM fcis-5218 101 26 ): ): PUNCT fcis-5218 101 27 343 343 NUM fcis-5218 101 28 . . PUNCT fcis-5218 102 1 [ [ X fcis-5218 102 2 5 5 X fcis-5218 102 3 ] ] PUNCT fcis-5218 102 4 szegedy szegedy VERB fcis-5218 102 5 c c NOUN fcis-5218 102 6 , , PUNCT fcis-5218 102 7 vanhoucke vanhoucke NOUN fcis-5218 102 8 v v NOUN fcis-5218 102 9 , , PUNCT fcis-5218 102 10 ioffe ioffe PROPN fcis-5218 102 11 s s PART fcis-5218 102 12 , , PUNCT fcis-5218 102 13 et et NOUN fcis-5218 102 14 al.rethinking al.rethinke VERB fcis-5218 102 15 the the DET fcis-5218 102 16 inception inception ADJ fcis-5218 102 17 architecture architecture NOUN fcis-5218 102 18 for for ADP fcis-5218 102 19 computer computer NOUN fcis-5218 102 20 vision[c].computer vision[c].computer NOUN fcis-5218 102 21 vision vision NOUN fcis-5218 102 22 and and CCONJ fcis-5218 102 23 pattern pattern NOUN fcis-5218 102 24 recognition.ieee recognition.ieee NOUN fcis-5218 102 25 , , PUNCT fcis-5218 102 26 2016 2016 NUM fcis-5218 102 27 : : PUNCT fcis-5218 102 28 2818 2818 NUM fcis-5218 102 29 - - SYM fcis-5218 102 30 2826 2826 NUM fcis-5218 102 31 . . PUNCT fcis-5218 103 1 [ [ X fcis-5218 103 2 6 6 NUM fcis-5218 103 3 ] ] PUNCT fcis-5218 103 4 lin lin PROPN fcis-5218 103 5 t t PROPN fcis-5218 103 6 y y PROPN fcis-5218 103 7 , , PUNCT fcis-5218 103 8 goyal goyal PROPN fcis-5218 103 9 p p NOUN fcis-5218 103 10 , , PUNCT fcis-5218 103 11 girshick girshick ADJ fcis-5218 103 12 r r NOUN fcis-5218 103 13 , , PUNCT fcis-5218 103 14 et et NOUN fcis-5218 103 15 al.focal al.focal ADJ fcis-5218 103 16 loss loss NOUN fcis-5218 103 17 for for ADP fcis-5218 103 18 dense dense ADJ fcis-5218 103 19 object object NOUN fcis-5218 103 20 detection[c]//proceedings detection[c]//proceeding NOUN fcis-5218 103 21 of of ADP fcis-5218 103 22 the the DET fcis-5218 103 23 ieee ieee NOUN fcis-5218 103 24 international international PROPN fcis-5218 103 25 conference conference NOUN fcis-5218 103 26 on on ADP fcis-5218 103 27 computer computer NOUN fcis-5218 103 28 vision.2017 vision.2017 NOUN fcis-5218 103 29 : : PUNCT fcis-5218 103 30 2980 2980 NUM fcis-5218 103 31 - - SYM fcis-5218 103 32 2988 2988 NUM fcis-5218 103 33 . . PUNCT