Int ern at i onal  Journ al of Ele ctrical  an d  Co mput er  En gin eeri ng   (IJ E C E)   Vo l.   15 ,  No.   1 ,  Febr uary   20 25 , pp.  908 ~ 920   IS S N:  20 88 - 8708 , DO I: 10 .11 591/ij ece.v 15 i 1 . pp 908 - 920           908       Journ al h om e page :  http: // ij ece.i aesc or e.c om   Harnessi ng   de ep   l earning   for   me dicin al   plant   resea rch:     a   co m pr ehensi ve   study       Vidya  Hull eke re Anan da 1 ,  Narasi mha M u rthy  Madiw ala  S athyan ar ay an a R ao 2 ,   Thar a  D harm ap ur a Kri shn amu r th y 3   1 Dep artm en t   of   Co m p u ter   Sci en ce   an d   Eng in eering ,   Kal p ataru I n stitu te of  Techn o lo g y ,  Tiptur ,   Visv esv araya  T e ch n o lo g ical Univ e rsity ,  Belag av i,  Ind ia   2 Dep artm en t   of   Inf o rm atio n   Scien ce   a n d   Eng in eering ,   BMS   Ins titu te   of   Te ch n o lo g y   an d   Manag em en t,   Ben g alu r u ,   Visv esv araya  Techn o lo g ical   Un iv ersity ,  Belag av i,  I n d ia   3 Dep artm en t   of   Inf o rm atio n   Scien ce   a n d   Eng in eering ,   Ch an n ab asav esh wara   Ins titu te   of   Tech n o lo g y ,   Gu b b i,   Visv esv araya  Techn o lo g ical Univ ersity ,  Belag av i,  I n d ia       Art ic le   In f o     ABSTR A CT   Art ic le   hist or y:   Re cei ved   Feb  23, 202 4   Re vised   Ju l  17,  2024   Accepte d   Se p 3, 2 024         In   toda y’s   world ,   peop le   are   mor e   prone   to   dise ase s   due   to   food   a dult er at ion   and   pol lut ion   in   the   envi ron ment,   and   p eopl e   have   found   a   way   of   using   her bal   m edi c ine   as   an   a lt ern at i ve   to   al lop at hi c   m edi c ine,   espe ci a ll y   since   cor onavi rus  dise ase   2019   (COV ID - 19) .   Medicin al   pl ant s   ar e   th e   source   of   her bal   me d ic in es   tha t   inc r ea se   the   im mun it y   of   hu ma ns.   Medi ci n al   pla nts   ar e   used   in   ma ny   appl i ca t ions,   li k e   phar mace ut icals,   cosmetic s,   and  drugs.   Medic in al   pla n t s   are   of   gre at   i mport an ce ,   and   henc e   th is   work   pre sen ts   a   rev ie w   of   the   m edi c ina l   p la nts   g rown   in   Karna t a ka   State,   Ind ia.   The   work   al so   high li ghts   spec ie s   ide n ti fi c at ion   and   dise a se   de te c ti on   of   m edi c ina l   pla nts   em ploy in g   machin e   le arn i ng   and   de ep   l ea r ning   appr o ac h es.   The   pap er   provide s   informati on   about   d at a sets   ava i la bl e   f or   var ious   m edicinal   pla n t   leaf   images.   Th e   de ep   le a rning   mode ls   used   for   spec i es   id ent if i ca t ion   and   disea se   de tecti o n   in   me d ic in al   pla nts   hav e   be e n   discussed   al o ng   with   the   result s .   Ke yw or d s :   Deep l ear ning   Disease  detect ion   M ac hin e lea rn i ng   M e dicinal  p la nt s   Sp eci es i den ti f ic at ion   This   is   an   open   acc ess   arti cl e   un der   the   CC   BY - SA   l ic ense .     Corres pond in g   Aut h or :   Vidya  Hu ll eke re  An a nda   Dep a rtme nt   of   Com pu te r   Scie nce   a nd   E ng i ne erin g,   Kalpata ru Insti tute o f Te ch no l ogy ,  V isvesva ra ya  Tech no l og ic al   Un i ver sit y   NH 20 6,  B H  R oad, Ti ptu r - 572201 ,   I nd ia   Emai l:   vidya. ha @gmai l.com       1.   INTROD U CTION   M e dicinal   pla nt s   play   a   pivot al   ro le   in   to da y’ s   w or l d   as   t he y   are   the   main   c ompone nt   of   A yur ve da.   Ayurve da   s olve s   healt h   prob l ems   thr ough   a   ho li sti c   ap proa ch   by   pr e ve nting   man y   disea ses   an d   hel ping   face   po te ntial   chall eng e s   in   t his   world   [1] .   Mo dern   me dicine   wea ke ns   the   immu nity   po we r   of   the   huma n   body,   wh e reas   t he   A yur ved ic   a ppr oa ch   to   healt h   he lps   mai ntain   good   healt h   by   i ncr easi ng   the   i mmunit y   of   the   body   natu rall y.   N ow adays,   we   ca n   see   that   man y   Ayurve dic   ther apies   treat   dise ases   that   are   ha rd   to   c ur e   in   m od e r n   medici ne .   In   man y   cases ,   A yur ved a   av oids   surge ries   a nd   heals   a   va riet y   of   diseases .   Accurat e   ide nt ific at ion   and   m on it or in g   of   me dicinal   plant   sp eci es   is   essenti al   to   ens ur e   t he   qua li ty,   ef ficacy,   and   safet y   of   herbal   pro du ct s   an d   t rad it io nal   me di ci nes,   as   me di ci nal   plants   ha ve   man y   us es   in   medici ne .   Howe ver,   the   visu al   simi la rity   an d   sp ect r um   of   medici nal   pl ant   bi od i versi ty   pose   sig nifi cant   ch al le ng e s   f or   co nvent ion al   identific at ion   methods ,   oft en   relyin g   on   t he   exp e rtise   of   hu man   e xp e rts.   A lso,   the   ti mely   detect ion   of   di seases   in   medici nal   plants   is   cr ucial   to   mainta in ing   the   pote nc y   an d   inte gr it y   of   the   plant s   for   their   i ntend e d   medici nal   a pp l ic at ion s.   Unde te ct ed   plant   diseases   can   le a d   to   si gn i fican t   crop   l os ses   a nd   t he   product ion   of   su bst an dard   he rb al   pro du ct s ,   unde rmin i ng   the   reli abili ty   an d   tr us t   in   tra diti on al   me dicine .   Evaluation Warning : The document was created with Spire.PDF for Python.
In t J  Elec  &  C omp E ng     IS S N:   20 88 - 8708         H ar ne ssin g de ep  le arni ng for  m e dicinal  pl ant rese ar c h:   a compre he ns iv e stu dy   ( Vidy a  Hu ll ekere  A na nda )   909   In   t his   work,   we   ai m   to   le ve rag e   the   pow er   of   dee p   le ar ning,   a   cutti ng - ed ge   a rtific ia l   intel li gen ce   te chn iq ue,   to   address   t hese   chall en ges   a nd   re vo l utio nize   the   way   medici nal   plants   are   ide ntifie d   a n d   monit or e d.   S pe ci fical ly,   we   se ek   to   de vel op   r obus t   an d   acc urat e   dee p   le arni ng - base d   a utomat ed   cl assifi c at ion   sy ste ms   e qu i pped   to   ha nd le   the   i den ti ficat ion   of   a   ric h   as so rtme nt   of   m edici nal   plant   var ie ti es.   E xp l or e   the   impleme ntati on   of   dee p   le ar ni ng   a ppr oac hes   to   the   ea rly   an d   pr eci se   detec ti on   of   disease s   in   me dicinal   plants,   enab li ng   proac ti ve   inter ven ti on   an d   mana ge ment.   The   w ork   e mphasiz e s   the   si gn i ficance   of   le ve rag i ng   dat a   augmentat io n   appr oach es   to   al le viate   the   li mit at ion s   of   small   medici na l   plant   datas et s   an d   e nh a nc e   the   gen e rali zat ion   of   dee p   le a rn i ng   m od el s .   Investi gate   the   pro fici ency   of   t he   deep   le ar nin g - based   ap pr oach e s   and   c ompa re   t hem   to   co nven ti on a l   mac hin e   le arn in g   meth od s ,   dem on st ra ti ng   the   a dvan cements   ma de   in   this   fiel d.   By   ac hieving   these   ob je ct iv es,   we   ai m   to   co ntribute   to   the   im prov e m ent   of   t he   me dicinal   plant   industr y ’ s   ef fi ci ency ,   qu al it y   co ntr ol,   a nd   su sta ina bili ty.   The   dee p   le a r ning - powe red   too ls   an d   te c hniq ues   dev el op e d   in   this   stu dy   ca n   be   inte gr at e d   into   the   w or kf l ow s   of   cult ivators ,   pro ducers,   an d   heal thcare   pr act it ion e rs,   e mpowe rin g   th em   with   reli a bl e   and   scal a ble   so luti ons   for   medici nal   pla nt   identific at io n   an d   disease   m onit ori ng.   T his   rese arch   re pr ese nts   a   cr ucial   le ap   in   ad opti ng   de ep   le ar ning   al gorith ms   in   me di ci nal   plant   ap plica ti on s ,   with   t he   po te ntial   to   transform   tra diti on al   healt hca re   pr act ic es,   e nh a nce   the   reli a bili ty   of   herbal   product s,   a nd   pro mo te   the   s us ta ina bl e   mana ge ment   of   me dicinal   pl ant   res ources .   The   pri ma ry   mo ti vatio n   of   this   wor k   is   to   le verage   the   power   of   dee p   le arn i ng   to   nav i gate   the   diff ic ulti es   of   accurate   i den ti ficat ion   a nd   ti mely   detect ion   of   diseases   in   me dicinal   plants   to   a mp li f y   the   eff ic ie nc y,   qu al it y   con tr ol,   and   su sta ina bi li ty   of   the   me dicinal   pla nt   industr y   a nd   t r aditi on al   healt hcar e   pr act ic es.   T he   pro blem   ad dre ssed   in   this   stu dy   is   to   ma xim iz e   the   us e   of   ayur ved ic   me di ci nes,   so   we   ne ed   to   identif y   t he   me dicinal   plants   a nd   s uppress   the   at ta ck   of   dise ases   on   medici nal   plants   to   sa ve   their   c om m un it y .   Deep   le ar ning   is   pivotal   in   i de ntify i ng   me dicinal   pla nt   sp e ci es   an d   detect ing   t heir   diseas es.   India   is   know n   as   popula r   f or   its   t rad it io na l   healt h   s ys te ms   that   i nclu de ,   A yur ve da,   Yoga,   U na ni,   Siddha,   an d   S owa - Ri gpa  al ong   with   H ome opat hy.   Ba se d   on   this   s ys te m ,   nowa days   we   can   see   man y   po li ci es   and   s ys te ms   na med   A Y USH   wh ic h   is   sho w n   in   Fi gure   1.   An ci e nt   huma n   ci vili zat ion   use d   t he   Sid dh a   syst em   arou nd   800 - 700   BC E,   the   U nan i   s ys te m   of   medici ne   was   us e d   ar ound   460 - 377   BC E,   Ayurve da   wa s   us e d   arou nd   900 - 800   BC E,   Home op at hy   ar ound   1850   CE,   a nd   Yoga   a nd   nat uro path y   was   use d   pa st   ma ny   de cades   wh ic h   are   nat ur al   heali ng   s ys te ms   [2] .   T he   num be r   of   plant   s pecies   us ed   in   these   sy ste ms   is   s how n   in   Figure   2   [3] .   India   is   one   of   the   rese rvoir s   of   biodive rsity   in   t he   w or l d.   Plants   a re   f ound   mai nly   in   the   Wester n   Gh at s ,   N or t h - Ea ste r n   I ndia ,   a nd   the   Himala ya n   reg i on.   T her e   are   a rou nd   7 , 000   medici nal   plan t   sp eci es   in   I nd ia ,   and   a rou nd   19 00   in   Ka rn at a ka   [ 4] .   P eop le   belo w   the   po ver ty   line   cannot   af for d   the   exp e ns i ve   heal thcare   se rv ic es   pro vid e d   to da y   an d   t hey   do   not   e ven   ha ve   acce ss   to   the   healt hcar e   ser vi ces,   especial ly   in   r emote   a reas.   Alte rn at ive   wa ys   nee d   to   be   fou nd   to   meet   the   c halle nge s   face d   by   t he se   po or   people.   M e dicinal   plants   offe r   reme dies   for   this   pro blem   a nd   man y   healt h   issues   c an   be   so lve d   if   ta ke n   care   of   at   the   rig ht   ti me   with   the   rig ht   medicat i on.   T he   pre dominant   portio n   of   t he   me dicinal   pla nts   is   f ound   in   forests.   Me dici nal   plants   pro vi de   treat me nt   f or   the   poor   pe op le   at   an   a ffo rd a ble   pr ic e   a nd   t hey   al s o   ge ner at e   income   a nd   e mp lo ym e nt   if   fo c us e d   pro pe r ly.   Acc ordin g   to   t he   W or l d   Healt h   O rg a ni zat ion   ( WHO ),   ab out   80%   of   t he   gl ob al   popula ti on   reli es   on   tra diti on al   medici ne,   w hich   ai m s   to   prom ote   t he   well - bei ng   of   both   people   a nd   the   plan et   [ 5] .   90%   of   t he   me di ci nal   pla nts   are   us e d   as   ra w   dru g   mate rial   in   the   Indian   me dicina l   sy ste m   [ 6] .   40 %  of   the   pha r maceuti cal   in dustrie s   a re   us in g   me dicinal   pl ants   [7] .               Figure   1.   Tra di ti on al   me dicin e   sy ste m     AYUS H   Figure   2.   I nd ia n   tra diti on al   m edici ne   s ys te m   with   history   a nd   numb e r   of   plant   s pecies   Evaluation Warning : The document was created with Spire.PDF for Python.
                          IS S N :   2088 - 8708   In t J  Elec   &  C omp E ng,  V ol.  15 , No .  1 ,  Febr uary   20 25 :   908 - 920   910   Unfortu natel y,   the   sect or   of   medici nal   plan ts   in   our   c ount ry   is   not   well   orga nized   a nd   po te ntial ly   util iz ed.   Nati onal   an d   sta te - l evel   orga nizat ion s   ha ve   to   de sign   poli ci es   relat ed   to   the   medici nal   plan t   sect or .   The   medici nal   plant   gro wing   ha bitat s   ha ve   to   be   i ncr eas ed   with   gr eat   su pe r visio n,   a nd   the   ra re   m edici na l   plant   sp eci es   t hat   a re   extinct   need   to   be   pr otect ed.   So   t hat   I nd ia   ca n   e xpor t   nat ur al   me dicines   to   i nter national   mar kets   [ 8] .   We   can   fin d   the   app li cat io n   of   medici nal   plan ts   in   va rio us   fiel ds .   Fi gure   3   sh ows   t he   a pp l ic at io n   of   me dicinal   pl ants   in   va rio us   fiel ds   in   the   m ark et .           Figure   3.   M e di ci nal   pla nts   in   var i ou s   fiel ds       Ar ti fici al   intel li gen ce  is   a   ne w   te ch nolo gy   of   c omp uter   sc ie nce   that   e vo l ved   67   year s   a go.   A rtific ia l  intel li gen ce   wa s   in ven te d   in   1956   by   Joh n   Mc Ca rthy   [9] .   B ut   it   has   b ec ome   popula r   in   the   past   few   de cades.   It   is   widel y   use d   in   al m os t   a ll   app li cat ions   tod a y.   M ac hi ne   le ar ning   (ML)   a nd   dee p   le arn i ng   ( DL )   a r e   the   su bfi el ds   of   Ar ti fici al   In te l li gen ce   an d   a re   fin ding   great   app li cat io n   in   ma ny   fi el ds   li ke   heal thcare,   agr ic ultur e ,   ba nk i ng,   s ocial   media,   c yber   secur it y,   rob otics,   e - c om me r ce,   ed ucati on,   an d  gaming.   In   t his   work,   deep   le a rn i ng   te c hn i ques   us e d   in   sp e ci es   identific at ion   a nd   diseas e   identific at io n   in   me dicinal   plant s   are   discuss e d.   Natu re   offe rs   us   a   rema r kabl e   meth od   of   natu rall y   heali ng   var i ous   il ln esses   a nd   i njurie s   t hroug h   medici nal   plan ts.   M e dicinal   plants   play   a   r emar kab le   r ole   in   re du ci ng   t he   use   of   to xic   su bst ances   in   dru g   pro du ct io n.   Me dicinal   pla nts   hav e   a ntimi crobial ,   antibi otic - resist ant ,   and   a ntibact er ia l   pr operti es.   Th e   medici nal   pro pe rtie s   of   pla nts   come   from   va r iou s   pa rts   of   t he   plants   su c h   as   le aves,   ste ms ,   ba rk,   root,   flo wer s ,   fruit ,   gum ,   r hiz om e,   seed ,   tu be r,   a nd   w ood.   M e dicinal   pla nt s   can   be   use d   to   incre ase   im munit y,   es peci al ly   in   childre n.   We   c an   us e   t hem   in   treat ing   co m mon   diseases   l ike   col ds ,   co ughs,   fev e r,   dys enter y,   vomit ing,   a nd   man y   oth e rs.   T his   s ub - sect io n   summariz es   va rio us   diseases   treat ed   us in g   medici nal   plan ts.   O n e   of   t h e   k e y   d i s e a s e s   t r e a t e d   u s i n g   m e d i c i n a l   p l a n t s   is   r e l a t e d   to   s k i n .   M a n y   a l l o p a t h y   d r u g s   e s p e c i a l l y   r e l a t e d   to   s k i n   a r e   m a d e   of   n a t u r a l   c o m p o n e n t s   a v a i l a b l e   from   m e d i c i n a l   p l a n t s .   T h e   f i v e   m o s t   u s e d   p l a n t   f a m i l i e s   in   I n d i a n   E t h n o d e r m a t o l o g y   a r e   E u p h o r b i a c e a e ,   F a b a c e a e ,   A p o c yn a c e a e ,   A s t e r a c e a e ,   a n d   Z i n g i b e r a c e a e .   T h e s e   f a m i l i e s   a r e   u s e d   to   c u r e   d i s e a s e s   s u c h   as   c u t s   a n d   w o u n d s ,   r i n g w o r m ,   s k i n   d i s e a s e s ,   e c z e m a ,   s o r e s ,   s c a b i e s ,   l e u k o d e r m a ,   b u r n i n g   a r e a s ,   w a r t s ,   h e r p e s ,   i n f e c t i v e   h e p a t i t i s ,   c a r b u n c l e ,   i t c h i n g ,   p i m p l e ,   s o l e   on   l e g s ,   l e p r o s y ,   p s o r i a s i s ,   s k i n   i n f e c t i o n ,   i n f l a m m a t o r y   d i s e a s e s ,   a s t r i n g e n t ,   b o i l s ,   d a n d r u f f ,   d e r m a t i t i s ,   s k i n   a l l e r g i e s ,   s w e l l i n g   on   h a n d s   a n d   l e g s ,   e a r   a c h e ,   a b d o m i n a l   c r a m p s ,   r a s h ,   s n a k e   b i t e s ,   s y p h i l i s ,   a c n e ,   b i t e s ,   w h i t e   s p o t   of   s k i n ,   w o r m y   sk in   s o r e s ,   h y p e r p i g m e n t a t i o n ,   l e u k o d e r m a ,   p r e m a t u r e   s k i n   w r i n k l e ,   s k i n   a l l e r g y   c a u s e d   by   i n s e c t ,   b i t e s   or   m i c r o b e s   [ 5 ] .   T h e y   a r e   a l s o   u s e d   to   t r e a t   d i s e a s e s   r e l a t e d   to   e n d o c r i n e   d i s o r d e r s ,   d i a b e t e s   m e l l i t u s ,   t h y r o i d a l   a n d   h o r m o n a l   i m b a l a n c e s .   O n c e   t h e   p l a n t   is   u s e d   for   m e d i c i n a l   p u r p o s e s ,   t h e   o t h e r   p a r t s   of   p l a n t s   f o r m   r e s i d u a l   b i o m a s s   u s e d   as   r e n e w a b l e   e n e r g y   r e s o u r c e s   on   t h e   e a r t h   [10] .   The   me dicinal   plants   of   t he   C ucur bitac eae   fa mil y   c on ta in   ri ch   phyt och e mica ls   that   giv e   thera pe utic   eff ect s.   T hese   pla nts   e xhibit   the   properti es   of   a ntihype r gly cemi c ,   a ntidiabeti c,   antic ancer ,   a ntimi c robial ,   antioxi dan t ,   a nalgesic,   a nti - inflammat ory,   anti - st ress,   a nd   imm unom odulato ry   e ff e ct s.   T hese   pla nts   a re   nu t riti on al ,   ec onom ic al ,   a nd   et hn ov et eri na ry   in   natu re   [ 11] .   T he   medici nal   pla nts   namel y   L ic an ia   ma cr ophy ll a   (l eaf,   bark),   M anil kar a   el at a   ( bark),   a nd   Vo uaca poua   A mer ic ana   ( ba rk)   e f fici ently   fi gh t   against   herpes   a nd   c hi kungun ya   dise ases   avail able   in   the   Amaz on i an   re gion   [ 12] .   M edici nal   pla nts   ha ve   the   be nef it s   of   reducin g   r espirato r y   infe ct ion s,   a ntidia betic   po te ntial ,   impro ve d   immu nity,   a ntiviral   pro pe rtie s,   anti - inflammat ory,   imp rove d   sle ep   qual it y,   normali zed   pulmo na ry   f un ct ion in g,   anti ox i dan t,   m uc ocili ary   c le aran ce,   an d   augmenti ng   ph ago c ytosi s   [ 13] .   Anothe r   im por ta nt   ap plica ti on   of   medici nal   plants   i nclu de s   synthesiz i ng   sil ver   nano par t ic le s.   Sil ve r   nano par ti cl es   ( AgNPs)   a re   bio synt hesized   from   herbal   plan ts   and   act   as   th erap e utic   age nt s   f or   bacteria ,   f ungi ,   and   tu mors   [ 14] .   M e dicinal   plants   play   an   imp or ta nt   r ol e   in   re du ci ng   inflammat io n   in   mem ory   c og niti ve                                       a     a       t     a   s   o o d   a d d   t       s   a t     a     d y   s     a   t   y     o o d s     a d   t   o   a         d                   a     t   a     a   t   d           d   d       s   o s     t     s Evaluation Warning : The document was created with Spire.PDF for Python.
In t J  Elec  &  C omp E ng     IS S N:   20 88 - 8708         H ar ne ssin g de ep  le arni ng for  m e dicinal  pl ant rese ar c h:   a compre he ns iv e stu dy   ( Vidy a  Hu ll ekere  A na nda )   911   diseases   li ke   A z       ’ s   di sease,   vascu la r   deme ntia,     a k    s o  ’s   dis ease,   tra um at i c   inju ry   in   the   brai n,   stroke,   an d   ce rebral   mala ria   [ 15] .   Re centl y   it   has   bee n   ob s er ved   that   bio act ive   co m pone nts   in   me dicinal   plants   ha ve   an   impact   on   oc ul ar   dis order s   like   infecti ons   in   the   eye,   e ye   diso r de rs,   a nd   vi sion   dama ge   w hich   include s   gla uc om a   a nd   di plopia   [ 16] .   C olon   cancer   is   a   dis ease   cause d   by   gut   micr ob i ota.   The   disease   can   be   treat ed   in   Ayu rv e da   with   the   us e   of   pla nts   mainl y   with   po l yphe no li c   c ompou nds   in   cur c um i n   (Cur cum a   longa/T urme ric).   It   has   che m opre ven ti on   prop e rtie s;   it   is   antioxi dan t   a nd   anti - infla mmat ory.   Aloe   ve ra   helps   pr e ve nt   the   gr ow t h   of   harmf ul   bacteria   cl ost ridiu m   perfr i ng e ns.   T he   Gl ycyrrhiza   glab ra   (lic or ic e) ,   P or t ulaca   oleracea,   a nd   Eu phorbia   la t hyrism   a re   t he   ot her   pla nts   us e d   in   t he   t reatm ent   of   c olon   ca ncer   [ 17] .   Para lysis   is   ano t her   im port ant   disease   t ha t   is   f ound   in   c om m on   nowa da ys   an d   the   m edic inal   plants   distri bu te d   ov er   25   famil ie s   are   use d   to   treat   diff e ren t   t yp es   of   pa ralysis   [18 ] .   Ov e r   200   pla nt   sp eci es   in   the   Tibet o - B urma n  et hnic   gro up   are   us e d   for   tre at ing   va rio us   he al th   issues   sp eci fical ly   in   women ,   li ke   uri nary   prob le m s,   prepa rtum   c are,   ge nital   prob le m s,   co ns ti pation,   m at er na l   diet,   increase d   ap pe ti te ,   br ea st   mil k   sti mu la nt,   be auty,   bloo d   ci r culat ion ,   bloo d   press ur e ,   ba by   healt h   ca re,   uter us   involuti on,   bl ood   no ur is hm e nt,   c hild birth ,   dizzi ness,   me nst ru al   c ycle,   a nd   fe rtil it y.   T he   th ree   plant s   us e d   extensi vely   inc lud e d   Bl um ea   B al samifera   of   the   A ste racea e   fa mil y,   C le r odde ndr um   C ol ebrookian um   of   the   L amia ceae   fa mil y,   B uddleja   A sia ti ca   of   the   Scr ophula ria ceae   famil y   [ 19] .   Co nsumi ng   medici na l   pla nts   treat s   fati gu e   ca us e d   due   to   post - c oro nav ir us   dis ease  ( post - C O VID )   [20] .   Cr ud e   e xtracts   of   medici nal   plants   are   al so   us ed   in   c uri ng   ma ny   ne urol og ic al   dis order s   li ke   e pilepsy   w hich   is   f ound   in   low - i nc om e   c ountrie s   [21] .   M e dicinal   plan ts   al so   fin d   ap plica ti on   in   veterinar y   pr act ic es   in   man y   part s   of   I ndia .   One   su ch   ca se   stu dy   is   done   in   M e gh a la ya   in   N or t h   East   I nd ia   [ 22] .   M edici nal   pla nts   im pro ve   a ni mal   healt h   a nd   reduce   t he   ne ed   for   sy nt hetic   antibi otics   an d   horm on e s   us ed   in   a ni mals   for   t heir   grow t h   a nd   re duce   the   m or ta li ty   rate   [23 ] .   In   this   view ,   m edici nal   pla nts   sti ll   need   to   be   exp l or e d   m or e   to   fin d   the   po t entia l   of   me dic inal   plants   in   c ur i ng   hum anita rian   disea ses.   With   this   ex plorat io n   a nd   inte gr at io n   of   t rad it io nal   me dicine   wit h   t he   moder n   me dic ine   sy ste m ,   the   existi ng   heal thcare   s ys te m   will   be   imp roved   with   the   best   res ults   an d   ha ve   few e r   si de   ef fe ct s.   Tra diti on al   medici ne   ai ds   in   co ns e rv i ng   harmo nic   c onf ormit y   in   nat ure.   P l a n t s   a r e   t h e   m a j o r   s o u r c e s   of   p r e s e r v i n g   b i o d i v e r s i t y   in   na t u r e .   I n d i a   c on t r i b u t e s   7%   of   b i o d i v e r s i t y   in   t h e   w o r l d .   M e d i c i n a l   p l a n t s   p r i o r i t i z e   b i o d i v e r s i t y   c o n s e r v a t i o n   t o o ,   a l o n g   w i t h   t h e   d i s c o v e r y   of   f u t u r e   l i f e - s a v i n g   c o m p o u n d s   [ 2 4 ] .   T h e r e   a r e   17 , 0 0 0   to   18 , 0 0 0   f l o w e r i n g   p l a n t   s p e c i e s   in   I n d i a ,   out   of   w h i c h   a r o u n d   7 , 000   a r e   u s e d   as   m e d i c i n a l   pl a n t s   [ 2 5 ] .   T h e r e   a r e   a r o u n d   1 , 900   s p e c i e s   of   m e d i c i n a l   p l a n t s   f o un d   in   n a t u r a l   f o r e s t s   in   K a r n a t a k a   a c c or d i n g   to   K a r n a t a k a   s t a t e   m e di c i n a l   p l a nt s   a u t h o r i t y   ( K a M p A )   [4].   T h e   l i s t   of   v a r i o u s   m e d i c i n a l   p l a n t s   a v a i l a b l e   in   K a r n a t a k a   [26]   a n d   t h e   p l a n t s   u s e d   to   t r e a t   di f f e r e n t   di s e a s e s   is   m a d e   a v a i l a b l e   in   t he   l i n k   h t t p s : / / g i t h u b . c o m / v i d y a h a 1 8 / M e d i c i n a l - p l a n t s   u s i n g   t h e   r e f e r e n c e s   [ 5 ] ,   [ 1 1 ] ,   [ 1 8 ] ,   [ 2 7 ]   w i t h   t h e   f i l e   n a m e   “ L   s t   of   v a r i o u s   m e d i c i n a l   p l a n t s   of   K a r n a t a k a   u s e d   in   c u r i n g   h u m a n   d i s e a s e s . ”   T h e   t a b l e   in   t h e   l i n k   s h o w s   m e d i c i n a l   p l a n t s   in   K a r n a t a k a   t h a t   c a n   be   f o u n d   e a s i l y   a r o u n d   us   to   t r e a t   c o m m o n   d i s e a s e s .         2.   LIT ERATUR E  REVIE W   This  subsect io n  of  the  pa pe r  pro vid es  a  com pr e he ns ive   ov e rv ie w  of  the  existi ng  r esearch  a nd   sch olarly   w ork   on  t wo  main   areas - s pecies  i den ti ficat io n  a nd  disease   detect ion   of  me di ci nal  pla nts.   S pecies  identific at ion  is  cr ucial   in  me dicinal   pla nts  t o  e ns ure  t he  sa fety  a nd  ef fici ent  use   of  medi ci nal  plants .  D ise ase   detect ion   is  i mporta nt,  as  we  need  to  c on s er ve  me dic inal  plants  a ga inst  dif fer e nt   bacteria l  a nd  fun gal   diseases.  S pecies i de ntific at ion  a nd d ise ase  det ect ion   helps  i n pro per   util iz at ion   of me dicin al  p la nts.     2. 1.     Specie s   identific at i on   in   medi ci n al   pl an t s   Sp eci es   i den ti ficat ion   is   cr uc ia l   in   me dicinal   plants   to   ens ur e   the   saf et y   a nd   ef fici ent   us e   of   medici nal   plan ts.   Di ff e ren t   m edici nal   plants   co ntain   dif fere nt   c ompou nds   an d   are   use d   to   t reat   a   va riet y   of   diseases.   Acc urat e   ide ntific at ion   of   me dicina l   plants   ens ure s   the   co rr ect   usa ge   of   me dicin al   plants   to   t re at   the   disease   a nd   the   risks   of   side   e ff ect s   ca n   be   r edu ce d.   It   al s o   help s   in   kn owing   t he   e xtinct   medici nal   plan ts   an d   the   co ns e rvat io n   of   su c h   me dicinal   pla nts   an d   th us   hel ps   in   ens ur i ng   the   a vaila bili ty   of   medici nal   plan ts   to   a   long   e xtent.   Ma ny   dee p   le a rni ng   models   a re   us e d   in   the   spe ci es   identific a ti on   of   va rio us   plants .   S om e   work   done   in   this   fie ld   an d   t he   dee p   le arn i ng   m od e ls   us e d   in   this   app li cat io n   are   li ste d   he re.   The   m odel s   use d   f or   i den ti f yi ng   medici nal   pl ant   sp eci es   in   the   w ork   gi ve n   by   B orkatull a   et   al .   [ 28] ,   include   Ima ge Net   pre - trai ne d   Re sNet 50,   D ense Net2 01,   V GG1 6,   a nd   I nc eptionV 3   with   RMS prop   opti mize r.   The   ex pe rimen t   is   ca rr ie d   out   f or   10   e poch s   with   a   le ar ning   rate   of   0.0 001.   T he   meas ure   of   accu rac y   of   the   models   inclu de s   72%,   97 % ,   96 % ,   an d   95%   resp e ct ively .   Islam   et   al .   [29] ,   us e   t wo   conv olu ti onal   neura l   netw ork   (C N N)  pre - trai ne d   m odel s   name ly   De ns eNet 201   a nd   I ncep t ion Re s NetV 2.   T he   accu rac y   f or   t he   trai ning   datase t   of   both   the   models   is   98. 46%   a nd   92.93 %,   validat ion   accurac y   is   96. 30%   a nd   90. 10%,   te st   accurac y   is   80. 69%   a nd   90. 09%,   validat io n   pr eci sio n   is   96 .81%   an d   90. 83%,   validat ion   recall   is   95.43 %   a nd   87.72%   an d   va li dation   F1 - S c or e   is   96. 10%   and   88. 94%   re sp ect ively ,   for   bo t h   the   m odel s.   Evaluation Warning : The document was created with Spire.PDF for Python.
                          IS S N :   2088 - 8708   In t J  Elec   &  C omp E ng,  V ol.  15 , No .  1 ,  Febr uary   20 25 :   908 - 920   912   Azadnia   a nd   Kh ei rali pour   [ 30] ,   e mp l oy   a rtific ia l   neural   netw orks   in   their   w ork .   T he   trai ni ng,   te sti ng ,   a nd   va li dation   data   are   60%,   20 %,   a nd   20%   resp ect ivel y.   The   model   w as   able   to   ide ntify   si x   medici nal   pla nt s,   he nce   the re   we re   six   ne uron s   in   the   ou t pu t   la ye r.   The   m od el   us e s   6 , 000   A yur ved ic   pla nt   samples.   10   ne uro ns   wer e   us e d   in   t he   hidde n   la yer.   T he   pe rformance   me asur e s   of   the   model   a re   c orr el at ion   coeffic ie nt - 1,   100%   cl assifi c at ion   rate,   a nd   2.35 × 10 - 6   me an   squa re   er ror   w hich   is   op ti mal.   A   deep   C NN   is   us e d   by   Push pa   an d  Ra ni   [31] ,   to   cl assify   40   Ayu rv e dic   plant   s pe ci es.   T he   m odel   her e   is   na me d   AyurPla nt Net,   al so   the   pre - trai ned   model s   Re sNet5 0   a nd   De ns e N et 121   are   c ompa red   with   Re s N et 34,   VGG 16,   M obil eNetV 3_Lar ge ,   an d   Ef fici entNetw ork _B4   a nd   the   model   r ecorde d   an   acc ur ac y   of   92. 27 %.   Viet   et   al .   [ 32] ,   in   thei r   w ork,   us e d   a   l a rg e   da ta set   of   me dic inal   pla nt   ima ge s.   T he   stu dy   a ssessed   a nd   op ti mize d   t he   fed e rated   le ar ni ng   f rame wor k   usi ng   t wo   fe de rated   le ar ning   ap proac hes,   F edAv g   a nd   Fe dP r ox,   and   f our   sta te - of - the - art   dee p   le arn i ng   net w orks   f or   t he   ta s k   of   cat eg or iz i ng   me dicinal   pl ants.   T he   trai ni ng   se t   was   distrib uted   in   t wo   f orm s:   ind e pende nt ly  an d  ide ntic al ly  distrib ute d  ( IID )   an d   non - in de pende nt ly  an d   identic al ly  dist rib uted  ( non - IID ).   The   opti mal   fed erate d   le arn i ng   s ys te m   achieve d   an   a c cur ac y   of   94.51%   a nd   82.65%   ov er   the   baseli ne   on   IID   data   a nd   Non - IID   data,   resp ect ivel y.   F uture   res earc h   cou l d   in vestig at e   the   us e   of   ot her,   m or e   co mp le x   f e der at e d   le ar ning   al go rithms.   Govinda pr a bh u  an d  S um at hi   [33]   pro po s ed   en sem ble  Eff ic ie nt Net   and   Xce ption   with   Re sNe t   (EE XR)   model   an d   em ploye d   the   M e nd el e y   dataset   that   c onta ins   1 , 800   images   of   18   m edici nal   pla nt   s pecies   in   their   work.   The   model   ac hieve d   an   a cc ur ac y   of   96.71 %   an d   er ror   r at e   of   3.24 %   f or   the   pla nt   da ta set .   Dey   et   al .   [34]   trai ne d   an d   evaluate d   the   performa nce   of   se ven   dee p   le arn i ng   m od el s   V GG1 6,   V G G19,   Den s eNet 201,   Re sNet5 0V2,   Xcep ti on,   In c e ption Re s NetV 2,   a nd   In ce ptionV 3   to   ide ntify   a nd   cl assif y   inter - famil y   a nd   int er - s pecies   var i at ion s   of   me dicinal   plants .   D ense Net   sho w ed   promisin g   resu lt s   with   99. 64%   accurac y.   T he   f utu re   w ork   addresses   the   chall enges   of   automatic   plan t   identific at io n   in   div e rse   re gions,   exp a ndin g   the   dataset ,   en ha nc ing   the   r obust ness   of   the   a ppr oac h,   a nd   t r anslat ing   t he   r esearch   fin dings   int o   pr act ic al   real - world   ide ntific at ion   s olu ti ons.     Wang   et   al .   [ 35]   ide ntifie d   and   qua ntifie d   two   medici na l   plant   spe ci es,   an d   the   a uthors   us e d   a   com bin at io n   of   dee p   le ar ning   and   unman ne d   aerial   ve hicle   rem ote   se ns in g   ( U AV RS )   to   achieve   qu a ntit at ive   detect ion   of   the   fl ow e rs   of   t hese   two   pla nt   s pecies.   Y OL Ov7   a nd   YO L Ov5 n   we re   e mp lo ye d   in   the   work   wh e re   YOLO v7   s howe d   best   resu lt s   of   97%   accurac y   a nd   YO L O v5n   wit h   93.40%   accu racy.   Sh a rma   a nd  V ardha n  [ 36]   propose d   the   MTJNet   m odel   to   e valuate   an   I nd ia n   me dicin al   le af   datase t   in   their   stu dy.   The   pr opos e d   M TJ Net   achie ved   a   pr eci si on   of   99.60%,   re cal l   of   99. 62% ,   accur acy   of   99. 71%,   and   F1 - Sc ore   of   99. 58%.   T he   e xp e rime ntal   resu lt s   s ho wed   that   the   M TJ Net   sta ti sti cal ly   ou tp er f ormed   pr e valent   mod el s.   Fu t ur e   w ork   incl ud e s   de pl oy i ng   the   mod el   in   real - w or l d   a pp li cat ions   and   ex plorin g   t ran s fer   le arn in g   a nd   domain   ada ptati on.   Fr om   the   c urre nt   sta te - of - t he - art,   it   is   evide nt   that   dee p   le arn i ng   models   are   wi dely   us e d   in   s pecies   identific at ion   wh ic h   overc ome s   t he   c onve ntion al   mac hi ne   le ar ning   te chn i qu e s   [37 ] .   The   popula r   dee p   le arn in g   m odel s   us e d   in   t his   a rea   inclu de   A NN,   CN N,   a nd   pre - trai ne d   m od el s   li ke   Re s Net,   De ns e Net ,   V GG,   In ce ptio nV3,   M TJ Net,   Xce pt ion ,   an d   Mo bileNet   out   of   th ese   De ns e Net   sh ows   pr om isi ng   re su lt s.   Th e   sta te - of - the - art   pre - trai ned   dee p   le arn in g   mode l   Eff ic ie nt NetB 0,   E ff ic ie ntNetV 2 - S,   vision   tra ns f orme r,   a n d   bid irect io nal   e ncode r   ima ge   t ran s f or me r   are   al so   us e d   [ 38] .   Acc ordin g   to   the   s urve y,   we   f ound   that   the   la ck   of   dataset s   is   the   ga p   i den ti fi ed   an d   oth e r   models   li ke   fe der at e d   le ar ning,   a nd   t ransfe r   le arn in g   ca n   be   use d   oth e r   tha n   pre - trai ned   m od el s .     2. 2.     Dise as e   d etectio n   in   me dici na l   pla nt s   Eve n   t hough   medici nal   pla nt s   are   use d   to   treat   man y   di seases,   t hey   a re   pro ne   to   di seases   li ke   human s.   M e di ci nal   pla nts   are   majo rly   a f fected   by   f ungal   di seases   [ 39] .   T he re   a re   ma ny   ot her   diseases   c ause d   by   pests   in   me dicinal   plants ,   inf or mati on   a bout   im portant   medici nal   pla nt s   that   are   gro wn   in   Ka rn at a ka   sta te   in   India   al ong   with   pests   that   cause   the   disease   and   par ts   of   the   plant   th at   are   aff ect ed   is   giv en   in   t he   link   https: // git hub. c om /v idy aha1 8/Medici nal - pl ants   with   the   f il e   name   “L  st   of   imp or ta nt   medici nal   pla nt s   of   Karnata ka   at ta cked   by   va rio us   p sts” .   It   is   importa nt   to   mana ge   th ese   diseases   to   ensure   the   he al th   and   pro du ct ivit y   of   medi ci nal   plants   and   to   protect   the   bi oacti ve   co mpo unds   in   t hem.   Id e ntific at ion   of   dis eases   in   plant s   is   very   imp or ta nt;   especial ly   in   t he   cat e gor y   of   me dicinal   plants.   By   detect ing   the   disease   in   me dicinal   plants,   we   ca n   ta ke   appr opriat e   m e asur e s   to   c on tr ol   the   same   a nd   sav e   the   plan t   fr om   bei ng   da mage d   or   dy i ng.   Als o,   as   m edici nal   plants   are   us e d   in   ma ny   a ppli cat ion s   as   dis cusse d   in   the   intr oductio n   se ct ion ,   the   qual it y   of   t he   pla nt   is   of   gr eat   im portan ce   to   preser ve   the   qual it y   of   the   products   li ke   herbal   medici nes   a nd   oth e r   pro duct s   li ke   cosmeti cs.   Ea rly   detect io n   of   disease   hel ps   in   a voidin g   the   sp rea ding   of   the   disease   to   ot her   par ts   of   the   pla nt.   It   al so   helps   in   av oid in g   t he   use   of   fe rtil iz ers   to   co ntr ol   plant   disease   by   usi ng   bio l og ic al ly   avail able   m anures   li ke   cow   dung   and   oth e r   pr oducts   from   nee m   plants .   The r eby,   the   qu al it y   of   the   plant   as   well   as   the   so il   is   al so   preser ve d.   Deep   le ar ning   te chnolo gy   ha s   its   eff ect   in   the   fiel d   of   plant   disease   detect ion   no wad a ys.   Few   Evaluation Warning : The document was created with Spire.PDF for Python.
In t J  Elec  &  C omp E ng     IS S N:   20 88 - 8708         H ar ne ssin g de ep  le arni ng for  m e dicinal  pl ant rese ar c h:   a compre he ns iv e stu dy   ( Vidy a  Hu ll ekere  A na nda )   913   works   t hat   ha ve   bee n   done   in   ide ntify i ng   pla nt   disease s   an d   the   dee p   le a rn i ng   m odel s   us e d   in   disease   detect ion   of   plants   are   disc usse d   her e .   Andrew   et   al .   [40]   in   t heir   st udy   us e d   the   publicl y   a vaila ble   Plant Vill age   dataset ,   whic h   co ntains   54,30 5   ima ge   samples   of   dif fer e nt   pla nt   disease   s pecies   in   38   cl asses.   The   researc he r s   us e d   De ns eN et - 121,   Re sNet - 50,   V GG - 16,   a nd   I ncep ti on V4.   The   De ns e Net - 121   model   a chieve d   t he   hi gh est   cl assifi cat io n   accurac y   of   99. 81%   an d   F1 - Sc or e   of   0.998,   ou t performi ng   t he   ot her   pre - trai ned   mod el s.   The   Re s N et - 50   model   achie ve d   an   acc ur ac y   of   99. 83%   a nd   a   m od el   l os s   of   0.0 27.   In   c omparis on,   the   I nc eptio nV4   model   reache d   97. 59%   accu racy,   a nd   the   VGG - 16   model   ha d   a   lo wer   accu racy   of   84.27%.     Diykh   et   al .   [41]   us e d   RGB   d r one   ima gery   data   from   P rince   E dwar d   Is la nd ,   Ca na da   to   pr e dict   the   normali zed   di f fer e nce  veget at ion   i nd e x  ( NDVI).   The   resea rch e rs   pro pose d   a   novel   fr a m ewor k   that   i ntegr at es   empirical   cu r ve le t   transform   and   a   De ns e Net   dee p   le ar ning   m od el .   T he   pr opos e d   Den s eNet - base d   m od el   achieve d   t he   hi gh est   str uctural   simi la rity   i ndex   (S S IM = 0.98)   an d   t he   lo west   mean   s quared   er r or   ( M S E=1 20 )   in   NDVI   pr e di ct ion .   T he   model   al s o   s howe d   an   accu rac y   of   97%   in   pr e dicti ng   N D VI   from   the   RGB   dro ne   imager y.   Th e   draw bac k   is   that   the   m odel   w as   only   te ste d   on   a   small   dataset .   A lom   et   al .   [ 42]   in   t he   work,   us e d   B rassic a   Na pus   ra pes eed   s pecies,   c ollec te d   f r om   agr ic ultur a l   fiel ds .   The   st udy   ad opte d   f ive   c on te m po r ary   dee p   le ar ni ng - base d   CN N   models:   De ns eN et 201,   V GG1 9,   In ce ptio nV3,   Xcep ti on,   an d   Re sNet5 0.   T he   De ns e Net2 01   model   has   the   hi gh est   a ccur ac y   of   10 0%   for   flo wer s   a nd   97%   for   both   pac kets   a nd   le aves .   T he   s ys te m   is   a   bi nary   cl assifi cat ion   an d   ca nnot   sp eci fy   m ulti - cl ass   s pecies.   Aishwar ya   an d  Re ddy   [ 43]   in   the   st udy,   c reated   a   c omp reh e ns ive   data set   of   gro undn ut   le af   images.   The   s tudy   util iz ed   a   tri - CN N   ar c hitec ture   c onsist ing   of   De nse Net1 69,   I nce ption,   a nd   Xc eption,   wh ic h   are   pre - trai ned   on   the   Ima geN et   dataset .   The   pro posed   met hod   ac hieve d   an   accu racy   rate   of   98.46% .   On   the   pota to   le af   dataset ,   t he   accu racy   rate   was   96. 05 % ,   and   on   t he   gra pe   le af   dataset ,   it   was   99.32% .   The   pro po se d   e ns e mb le   strat e gy   dem onstrat ed   s up e rio r   perfor mance   c ompa r ed   to   tra diti on a l   te chn iq ues .   To   s ummari ze ,   mac hin e   le a rn i ng   al gorith ms   use d   f or   disease   detect ion   inclu de   suppo rt   vect or   machine s,   random   f orest s,   a rtific ia l   neural   netw orks ,   de ep   belie f   ne tw orks,   an d   dee p   CN N   with   t raine d   models   s uc h   as   V G G16,   In c eptionV 4,   Re s Net5 0,   Re sNet 101,   Re s Net1 52,   Ale xN et ,   G oogleNet,   DenseNet   [40],  [ 44] ,  [ 45 ] .   Re ce ntly,   t r ansf e r   le ar ning   eme r ged   as   a   ne w   fiel d   to   en han ce   the   performa nce   of   dee p   le arn in g   m ode ls   [46] .   Ne w   te chnolo gy   in   DL   li ke   at t ention   mec ha nis ms   can   be   us e d   an d   these   at te ntion   mecha nisms   w ere   i nteg rated   into   Mo bileNe tV2,   E ff ic ie nt NetV 2,   an d   S huff le NetV 2   [ 47] .   T he   w ork   in   the   fiel d   of   pla nt   disease   detect ion   ha s   been   c arr ie d   out   on   man y   a vaila bl e   dataset s   li ke   plant  vill age   dataset ,   PlantD oc ,   a nd   Kaggle ,   a nd   done   f or   man y   oth e r   cr ops,   but   not   sp eci fical ly   done   f or   medici nal   plan ts.   It   is   good   to   de velo p   m or e   e ff ic ie nt   so l utions   by   evaluati ng   the   model   on   a   la rg e r   an d   more   div e rse   da ta set   with   diff e re nt   cr ops.   Howe ver,   t he   work   in   t he   fi el d   of   me dici nal   plants   is   minimal   a nd   r equ i res   at te ntion.   Pr e vious   stud ie s   sho w   t hat   the re   a re   li mit ed   dataset s   for   me dicinal   plants   es pecial ly   f or   disease   detect ion.   Add it ion al   dataset s   nee d   to   be   buil t   to   acce le rate   the   researc h   in   the   fiel ds   of   me dicinal   plants.   Data   au gme nt at ion   te chn iq ues   ma y   be   us ed   to   enh a nce   t he   da ta set   siz e   ava il able   for   me di ci nal   plants,   wh ic h   im pro ve s   the   gen e rali zat ion   of   deep   le ar ning   models.   By   a pp l ying   t he   de ep   le a rn i ng   te c hnolog y,   we   c an   ma ximize   t he   us e   of   medici nal   plants   as   the y   ha ve   am ple   us e   in   the   fiel d   of   medici ne .   Be yond   dee p   le ar ning   te c hn i qu e s,   phyto c hemical   analy sis   co uld   be   us e d   to   get   accu rate   res ults   f or   disease   de te ct ion   bu t,   it   ta kes   m or e   ti m e   an d   effor t   to   im ple ment.   By   le ve rag i ng   t he   ca pa bili ti es   of   de ep   le ar ning   for   sp eci es   i den t ific at ion   an d   di sease   detect ion,   me dicinal   plant - base d   industri es   can   en ha nc e   their   eff ic ie ncy ,   qual it y,   and   s us ta ina bili ty,   ulti mate ly   be ne fiti ng   both   the   pro ducers   an d   the   c on s ume rs   of   he r bal   m edici nes   a nd   pro duct s.   In   the   f utu r e,   we   ma y   w ork   on   m od el s   li ke   gr a ph - base d   de ep   le ar ning   a nd   lo ng  s hort - te rm mem ory ne tworks  ( L ST M) .       3.   METHO D     This   sect io n   pro vid es   an   e xp e rime ntal   set up   that   c onsist s   of   di ff e r ent   ste ps   us e d   f or   s pecies   identific at ion   a nd   disease   det ect ion   of   medi ci nal   plants .   Fi gure   4   prov i de s   the   ov e rv ie w   of   the   process .   The   workflo w   m odel   f or  the   pr opos e d  s ys te m   c on ta ini ng  s peci es  ide ntific at io n  a nd  disease   detect ion  is  show n  in   Figure  4.  T he   model  sta rts   w it h  the   data   co ll ect ion   sta ge ,  the  data  ca n  be   ta ke n  from   t he  sta ndar d  pu blic  dataset s  a nd  al so   real - ti me   da ta   can   b e   us ed   he nce   data   ca n  be   colle ct ed   from   m ulti so ur ce.  T he   ne xt  st ep  is   data  prep r oces sing   w hich  inc lud es  data  a ugmentat io n,   nor mali zat ion ,  a nd  se gm e ntati on.  The  t hir d  ste p  is  the   model  sel ect io n  phase,   a ny  ne ur al   net wor k  model  ca n  be   e mp lo ye d  he re.   The   f our th   ste p  is  t he   trai ni ng  ph a se   wh ic h  incl ud es   trai ning  the  se le ct ed  ne ur al   ne twork  model  by   sp li tt ing   t he   dataset   into  t r ai nin g  an d  validat ion  dataset s.  The  f ifth   ste p  is  t he  e valuati on   phas e  w he re  t he  t ra ined   ne ural  n et work model  is  test ed  a nd  e valuate d  fo r  go od   pe rfo rma nce  us i ng  metri cs  li ke  co nfusion  met rics,  F1 - Sc ore ,   a nd   a rea  unde r  curve.  O n  su c cessf ul  te sti ng ,   the   m od el   can   be  use d  f or  s pecies  ide ntific at ion  an d  disease   de te ct ion .  If  the   m od el   pe rfo r mance   needs  to   be   imp rove d  f urt he r,   t he  trai ni ng   of  t he   m od el   is  require d  a nd  hen ce   we   ha ve  the   lo op  f r om   t he   evaluati on  pha se to t he  trai nin g p hase .   Evaluation Warning : The document was created with Spire.PDF for Python.
                          IS S N :   2088 - 8708   In t J  Elec   &  C omp E ng,  V ol.  15 , No .  1 ,  Febr uary   20 25 :   908 - 920   914       Figure   4.   W ork flo w   diagr a m       3 . 1.     D atase ts   for   medi ci na l  plants   It   is   very   hard   and   re quires   a   lot   of   ti me   a nd   effo rt   to   colle ct   the   data   re quire d   to   trai n   t he   m achi ne   le arn in g   an d   de ep   le a rn i ng   m od el s .   Fe w   res earche rs   ha ve   worked   on   this   an d   c ontrib ute d   to   the   datase ts   f or   medici nal   plan ts.   This   sect io n   giv es   detai ls   ab ou t   the   dat aset s   that   ar e   exclusi vely   a va il able   for   me dicinal   plants.     3 . 1.1.   Kag gle   da t as e t   This   dataset   c on ta in s   images   of   m edici nal   plant   s pecies   Ca tharan t hus   R os eus ,   Kala nc ho e   P in nar a ,   Lo ng e vity   S pi nach,   Te rmin al ia   B el li rica,   Terminali a   C he bula ,   Ce ntell a   A sit ic a,   Azad ira chta   I nd ic a,   O ci mu m   T en nifloru m,   Cym bopogo n   C it ratus,   E upho r bia   T it hymal oid es .   Th e   dat aset   con ta in s   5 , 000   im ages   of   wh ic h   70 %   is   us e d   as   trai ning   dat a   co ntainin g   3 , 500   image s,   20 %   of   data   is   us e d   as   te sti ng   data   c onta ining     1 , 000   ima ges ,   and   10 %   as   validat ion   data   c onta inin g   500   i mages   [ 28].     Tit le :   M edici na l   plant   raw   URL:   http s:/ /www.k a ggle .com/ dsv/4 510170     DOI:   10. 34 740/ KAGG L E/DS V/4510 170       3 . 1 . 2.   Men dele y   data se t   The   dataset   c onsist s   of   two   c la sses,   the   me dicinal   le af   dataset ,   an d   the   medici nal   plan t   dataset   wit h   80   plant   s peci es,   6 , 900   imag es,   a nd   40   pla nt   s pecies,   5 , 900   ima ges   res pecti vely .   T he   pla nts   incl ud e   Al oe   Evaluation Warning : The document was created with Spire.PDF for Python.
In t J  Elec  &  C omp E ng     IS S N:   20 88 - 8708         H ar ne ssin g de ep  le arni ng for  m e dicinal  pl ant rese ar c h:   a compre he ns iv e stu dy   ( Vidy a  Hu ll ekere  A na nda )   915   barba den sis   mil le r,   P hy ll a nthus   E mb li c a,  Mo rin da   Ci trifolia ,  Tin os po ra   Co r dif olia,  Ficus   R el igiosa,  Eu phorbia   Hi rta,  Ba mbus oid e ae ,  Pipe r   Be tl e,  Ba co pa   M on nieri,  Ecl ipta   P ro st rate,  Ci nna mo m um   Ca mphora,   Ri ci nu s   Com munis,   Ci trus   M e dica,  C offe a ,  Mur ray a   K oen i gii,  C os tu s   I gne us ,   A rto carpus   Heterophyll us ,   Jasmin um ,  Zin gib e r   O ff ic inal e,  Psidi um   G ua ja va,   La wson ia   In er mis,  Hibiscus   a nd   ma ny   oth e rs.   T he   data   was   c ollec te d   in   M ysu ru   a nd   Ker al a   reg i ons   [48] .     Tit le :   In dia n   m edici nal lea ves  image  dataset s   URL:   http s:/ /data .mendele y.c om /data set s/ 748f8jk phb       DOI:   10. 17 632/ 748f8jk phb.3     3 . 1 . 3.   ME D11 7_Medici n al   p lant l eaf d atas et   The   dataset   c onta ins   117   different   sp eci es   of   77,50 0   t otal   medici nal   plan t   images.   T he   U - N et   m ode l   was   use d   for   t he   se gm e ntati on   of   the   ima ge s   in   the   pre pro cessi ng   sta ge.   Waters hed   seg mentat io n   te ch niques   wer e   al so   us e d   al ong   with   U - Net.   T he   cl ass   of   each   pla nt   c on ta in s   179   to   1 , 300   s pecies   [ 49] .     Database   name s:   MED 117_L eaf   Sp eci es   I   t he   databa se   w hich   ha s   t wo   s ubf old er s   nam el y,   Ra w   le af   image   set   of   Me dicinal   plants _v2   a nd   Se gme nted   le a f   set   usi ng   U - NE T   se gm e ntati on.   URL:   http s:/ /data .mendele y.c om /data set s/ dtv bwr hz nz/4   DOI:   10. 17 632/ dtv bw rh z nz.4     2. 1 . 4.   Medici n al   le af   d atase t   The   dataset   c on sist s   of   30   sp eci es   of   medici nal   plants   suc h   as   S antal um   Albu m,  M unti ngia   Ca la bu ra ,  Ama ran t hu s   Viri di s,  Aza dirac hta   Indica,  Ci tr us   Lemo n,  Ficus   Auriculat a ,   a nd   ma ny   more .   Each   sp eci es   has   60   to   100   ima ges   of   hi gh   qual it y.   It   is   a vaila ble   on   the   M e ndel ey   platfo rm   [ 50] .     URL:   http s:/ /data .mendele y.c om /data set s/ nnytj2v3 n5 / 1     DOI:   10. 17 632/ nnytj2v3 n5.1     3 . 1 . 5.   B D Medi Le av es :   a   le af   im ag es   d atase t   f or   B angla de shi   medi ci na l   pla nts   iden tifi cat i on   The   dataset   c on sist s   of   2 , 029   ori gi nal   i mages   an d   38 , 606   a ugmente d   ima ges   of   t he   le a ves   of   commo nly   f ound   te n   me dicinal   plants   in   Ba ng la desh.   T he   dataset   is   in   the   M en dele y   platfo rm .   It   con ta in s   two   cat eg ori es   namely   BD Me diLeaves   au gme nted   d at aset   and   B D M edi Leaves   or i gin a l   dataset   for   tr ai nin g,   te sti ng ,   a nd   va li dation.   T he   pl ants   inclu de   Hibisc us   Rosa - Sinensi s,  Ce nt el la   Asiat ic a,  Ph yllant uhus   E mb li ca,  Kalanc ho e   Pi nnat a,  M ika nia   M ic ra nth a,   Az adirac hta   I ndic a,  Te rmin al ia   Ar j un a ,  J us ti ci a   A dh at oda,  O ci mu m   Ten uifloru m,  a nd  Ca lotr op is   Giga ntean   [ 29] .     URL:   http s:/ /data .mendele y.c om /data set s/ gk5x6k8x r5 / 1   DOI:   10. 17 632/ gk5x6k8x r5.1     3 . 2.   Deep   le ar ning  m odel s   Deep   le a rn i ng   was   propose d   by   Ri na   Dec hter   in   1986.   The   fi rst   dee p - le arn i ng   al gor it hm s   we r e   publishe d   by   I vakh nenk o   a nd   Lapa   in   1967   [ 51] .   It   c onsist s   of   th ree   la yer s ,   an   in pu t   la ye r,   a   hidden   la ye r ,   a nd   an   outp ut   la ye r,   an d   ope rates   with   i nput   a nd   weig hts.   N ow a da ys ,   deep   le ar ning   has   become   incr ea sing l y   pr e valent   a nd   is   being   a ppli ed   in   al m os t   al l   app li cat ions.   Plant   s pecies   identific at ion   a nd   pla nt   diseas e   detect ion   are   vi ta l   app li cat ions   that   us e   dee p   le ar ning   models.   T he re   a r e   man y   dee p   l earn i ng   arc hitec tures   desig ne d   by   va rio us   pionee r s   who   are   w or king   in   t his   fi el d.   The   top   de ep   le a rn i ng   a rch it ect ures   i nc lud e:   i)  co nvolu ti on al   neural  net w orks  (CN N) ,  ii )  arti fici al   ne ural   netw orks  ( ANN) ,  ii i)  lo ng  s hort - te r m   memor y   netw orks  (LST M ) , iv) r ec urre nt  ne ur al   netw orks  (RN N) , v)  g e ner at ive  a dv ersar ia l n et w orks  (GAN) , vi)   rad ia l   basis  functi on  netw orks   (RB FN ) ,  vii)  mu lt i la yer   pe rcep t r on   ( M L P) ,  viii )  sel f - orga nizi ng   ma ps  (SO M ) ,  i x)   deep  belie f  n et works  ( DB N) ,   x) r est rict ed   B oltzma nn   mach ines  (RB M ) , a nd x i )  a uto e nc od e rs  (AE) .   M a ny   machi ne   le arn in g   a nd   deep   le a rn i ng   al gorithms   a r e   us e d,   to   na me   a   fe w   s upports   vect or   machine   ( SVM),   ra ndom   forest   (RF ),   per ce ptr on ,   ba ckpr op a gatio n   al gorithm ,   r egr es sio n   te ch niques,   Ba yesian   cl ass ifie r,   a nd   ne ural   netw orks.   T he   neural   net works   gaine d   high   im portan ce   an d   c onvoluti on a l   neural  netw orks  a re   popula r ly   use d   in   sp e ci es   identific at ion   a nd   diseas e   detect io n   in   plant s.   M a ny   models   work   based   on   the   a bove - me ntion e d   arc hitec tures.   In   t his   work,   we   ha ve   co ns ide re d   m edici nal   plants   an d   discusse d   w ork   do ne   in   sp ec ie s   identific at ion   an d   disease   detect ion   in   medici nal   plan ts.   The   majo rity   of   the   pap e rs   in vo l ve d   in   t he   st udy   hav e   us ed   C N N   models   [ 52] .   A   few   of   the   CNN   m od el s   use d   in   dee p   le a rn i n g   and   thei r   years   of   i nv e ntio n   a re   giv e n   in   Fig ur e   5.   Amo ng   the   m odel s   i de ntifie d   f rom   t he   stu dy,   De ns e Net   is   more   popula r   as   it   obta ins   good   resu lt s   for   our   ob je ct ives.   Evaluation Warning : The document was created with Spire.PDF for Python.
                          IS S N :   2088 - 8708   In t J  Elec   &  C omp E ng,  V ol.  15 , No .  1 ,  Febr uary   20 25 :   908 - 920   916       Figure   5.   Dee p   le arn i ng   m od e ls   base d   on   co nvol ution al   ne ural  n et w orks       4.   RESU LT S   A ND   DI SCUS S ION   In   this   sect io n,   we   discuss   t he   res ults   of   var i ou s   machi ne   le arn i ng   a nd   dee p   le a rn i ng   m od el s   us e d   in   sp eci es   ide ntif ic at ion   an d   di sease   detect io n   of   me dicina l   plants.   M a ny   co nventio na l   machine   le a rn i ng   te chn iq ues   li ke   rand om   f orest ,   suppo rt   vecto r   mac hin e,   a nd   oth e rs   ha ve   be en   use d.   H owever,   dee p   le ar nin g   models   giv e   pr om isi ng   res ults   co mp a red   to   machine   le ar ni ng   m odel s,   part ic ularly   ne ur al   net work s .   T he   stu dy   fou nd   that   m odel s   li ke   Re s Net,   De ns e Net,   VGG 16,   I nc eptionV 3,   A N N,   an d   AyurP la ntNet   a re   m aj or ly   employe d   f or   plant   spe ci es   identific at io n.   Among   these   Den s eNet   out pe rforms   al l   othe r   models   with   99.64%   accurac y,   a nd   M TJ Net   with   99.71%   acc ur a cy.   F or   plant   di sease   detect ion   va rio us   dee p   le ar ning   ap proach e s   li ke   I nce ptionV4,   Re s Net,   V GGNet,   Ale xNet ,   G oogleN et ,   O verfeat ,   V GG,   a nd   Dens eNet   ar e   em pl oy e d.   In   these   m odel s,   Den s eNet   outp erforms   al l   ot he r   m odel s   w ith   99. 8%   acc ur ac y.   M a ny   w orks   ha ve   been   do ne   on   s pecies   id entifi cat ion   but   the   ab ove   stu dy   li sts   so me   of   t he   wor k   carried   out   on   sp eci es   ide ntifi cat ion   of   va rio us   medici nal   pl ants.   T he   c omparati ve   anal ysi s   of   va rio us   de ep   le arn in g   m odel s   in   medici nal   plant   sp e ci es   identific at io n   is   giv e n   in   Ta ble   1   an d   Fig ure   5.   Ta ble   1   s hows   t he   models   a nd   dat aset   us e d,   the   numb e r   of   e po chs,   the   acc ur a cy   ac hieve d,   a nd   the   re fer e nc es   use d   to   s ummari ze   the   anal ys is   of   var i ou s   m od el s .     Figure  6  s how s  the  grap hical   representat io n  of   the  c ompa r at ive  analysis  of   var i ou s  models  us e d  in   the  st udy.  F rom  this ,  it   is   e vi den t  t hat  A N N  a nd  De ns e N et   models   outp erform  oth e r   models  with   100%   a nd  97%  accu racy   resp ect ivel y.   T he  co mp a rati ve   analysis  of  t he  m od el s  us e d  f or   disease  detect ion   is   gi ven   i n  Table  2  w hich   co ntains   in f ormat ion  li ke   m od el s ,  dataset s,   num be r  of  e poch s,   te st  a nd  validat io n  acc ur ac y,   and  re fer e nces .   The   w ork   nee ds   t o  be  e xten de d  for  medici na l  plant  s pecie s  to  protect   the m  f rom  disea s es  that   are  bein g  at ta cked. Fi gure  7  s hows  t he gra ph ic al  r eprese ntat ion   of the  com par at ive  an al ysi s g ive n  i n  Ta bl e 2 .       Table   1.   C omp arati ve   a nalysi s   of   var i ou s  d e ep  le ar ning  me thods   in   plant   s pecies   ide ntific at ion   Mod el   u sed   Dataset   Epo ch s   Accuracy   Refer en ces   Res Net5 0   Ban g lad esh i   m ed icin al   p lan t   d ataset   10   72   [28 ]   Den seNet2 0 1   Ban g lad esh i   m ed icin al   p lan t   d ataset   10   97   [28 ]   VGG1 6   Ban g lad esh i   m ed icin al   p lan t   d ataset   10   96   [28 ]   Incep tio n V3   Ban g lad esh i   m ed icin al   p lan t   d ataset   10   95   [28 ]   Den seNet2 0 1   BDMed iLeaves   20   8 0 .69   [29 ]   Incep tio n Res NetV2   BDMed iLeaves   20   9 0 .09   [29 ]   ANN   Real   tim e   d ata   20   100   [30 ]   Ay u r - Plan tNet   Real   tim e   d ata   -   9 2 .27   [31 ]   Fed erate d   lea rnin g   IID   an d   NON   II D   5   9 4 .51   [32 ]   EE XR   Mend eley   d ataset   50   9 6 .71   [33 ]   Den seNet2 0 1   Medicin al   p lan t   d a taset   20   9 9 .64   [34 ]   YOLOv7   Cu sto m   d ataset   -   97   [35 ]   YOLOv5   Cu sto m   d ataset   -   9 3 .4   [35 ]   MT JNet   Ind ian   m ed icin al   l eaf   d ataset   -   9 9 .71   [36 ]   Evaluation Warning : The document was created with Spire.PDF for Python.
In t J  Elec  &  C omp E ng     IS S N:   20 88 - 8708         H ar ne ssin g de ep  le arni ng for  m e dicinal  pl ant rese ar c h:   a compre he ns iv e stu dy   ( Vidy a  Hu ll ekere  A na nda )   917       Figure  6. Com par is on of  dif f eren t  dee p  le ar ning a ppr oach e s for   pla nt sp ec ie s  identific at ion           Figure  7. Com par is on of  dif f eren t  dee p  le ar ning a ppr oach e s for   pla nt d ise ase detect io n       Table   2.   C omp arati ve   a nalysi s   of   var i ou s   de ep  le ar ning  me thods   in   plant   disease   detect ion   Mod el   u sed   Dataset   Epo ch s   Test   accur acy   Valid atio n   accurac y   Refere n ce   Incep tio n   V4   Plan tVillag e   30   9 8 .08   9 8 .02   [45 ]   VGG   n et   (16 )   Plan tVillag e   30   8 1 .83   8 1 .92   [45 ]   Res Net   (50 )   Plan tVillag e   30   9 9 .59   9 9 .67   [45 ]   Res Net   (10 1 )   Plan tVillag e   30   9 9 .66   9 9 .66   [45 ]   Res Net   (15 2 )   Plan tVillag e   30   9 9 .59   9 9 .68   [45 ]   Den seNet   (12 1 )   Plan tVillag e   30   9 9 .75   9 9 .76   [45 ]   Alex Net   Realtime   d ata   47   9 9 .06   -   [44 ]   Alex NetOWTBn   Realtime   d ata   46   9 9 .44   -   [44 ]   Go o g LeNet   Realtime   d ata   45   9 7 .27   -   [44 ]   Ov erf eat   Realtime   d ata   45   9 8 .96   -   [44 ]   VGG   Realtime   d ata   48   9 9 .48   -   [44 ]   Incep tio n   V4   Plan tv illag e   30   9 7 .59   8 8 .7   [40 ]   VGG - 16   Plan tVillag e   30   8 2 .75   9 0 .1   [40 ]   ReNet - 50   Plan tVillag e   30   9 8 .73   9 3 .5   [40 ]   Den seNet - 121   Plan tVillag e   30   9 9 .81   9 9 .8   [40 ]   Den seNet   b ased   m o d el   RGB   Dron e   i m ag e   d ata   100   0   97   [41 ]   Den seNet2 0 1   Brass ica   Nap u s   d at a   10   0   0 .98   [42 ]   Tri - CN N   a rchitect u re   Im ag e N et   d ataset   20   9 9 .39   9 8 .46   [43 ]     Evaluation Warning : The document was created with Spire.PDF for Python.