Co m pu t er   Science  a nd   I nfo r m a t io n T ec hn o lo g ies   Vo l.  7 ,   No .   2 J u ly   20 26 ,   p p .   1 67 ~ 1 78   I SS N:  2722 - 3 2 2 1 DOI 1 0 . 1 1 5 9 1 /csi t . v 7 i 2 . p 1 67 - 1 78           167       J o ur na l ho m ep a g e h ttp : //ia e s p r ime. co m/in d ex . p h p /csi t   Perf o rma nce   ev a lua tion   of   th e   dee p   lea rning   sy stem   f o r   weed   recog niza tion       Abd   Abra him   M o s s l a h 1 ,   Rey a dh   H a zim   M a hd i 2 ,   H a s s a n   K a s s im   Alba ha dil y 3   1 C o l l e g e   of   I sl a mi c   S c i e n c e ,   U n i v e r s i t y   of   A n b a r ,   A n b a r ,   I r a q   2 C o l l e g e   of   S c i e n c e ,   U n i v e r s i t y   of   M u st a n si r i y a h ,   B a g h d a d ,   I r a q   3 C o l l e g e   of   S c i e n c e ,   C o mp u t e r   S c i e n c e   D e p a r t m e n t   B a g h d a d ,   M u s t a n s i r i y a h   U n i v e r si t y ,   B a g h d a d ,   I r a q       Art icle   I nfo     AB S T RAC T   A r ticle   his to r y:   R ec eiv ed   No v   1 4 ,   2 0 2 5   R ev is ed   Ma r   3 ,   2 0 2 6   Acc ep ted   Ma y   1 6 ,   2 0 2 6       Nu m e ro u s   a p p r o a c h e s   b a se d   on   m a c h in e   lea rn in g   h a v e   e m e rg e d   in   re c e n t   y e a rs   to   e n h a n c e   c ro p   p ro tec ti o n   e fficie n c y .   On e   e x a m p le   is   th e   u ti li z a ti o n   of   d e e p   n e u ra l   n e two rk s   (DN Ns )   to   d iffere n ti a te   b e twe e n   v a ri o u s   we e d   ty p e s   in   a c tu a l   e v e n ts   sc e n a rio s.   Ne v e rth e les s,   th e se   m e th o d s   o ften   n e e d   su b sta n ti a l   in p u t   fr o m   e x p e rts   wh o   w o rk   it e ra ti v e ly   to   d e sig n   th e   r o b u st   d e e p   lea rn i n g   sy ste m .   To   sim p li f y   su c h   p r o c e ss   a n d   c o n se rv e   re so u rc e s,   re se a rc h e rs   h a v e   e x p lo re d   a   fre sh   m e th o d   k n o w n   as   a u to m a ted   de ep   lea rn in g   o ur   tec h n o lo g y s   re c o g n iza ti o n   of   we e d s   t h ro u g h   t h e   u se   of   m a c h in e   lea rn i n g   wa s   e v a lu a ted   u sin g   p la n t   se e d li n g s   a n d   we e d   c o ll e c ti o n s   fro m   p lan ts   d a tas e t   to   a d d re ss   a   issu e   of   we e d   re c o g n iza ti o n .   T h e   stu d y   c o m p a re d   v a rio u s   c o n fi g u ra ti o n s ,   in c lu d in g   p lan t   se g m e n tatio n ,   u s in g   a   c o ll e c ti o n   of   c las sifiers   in   p lac e   of   S o ftma x ,   a n d   trai n in g   wit h   d a tas e ts   th a t   c o n tain   n o ise .   Th e   fin d in g s   in d ica ted   e n su ri n g   p e rfo rm a n c e ,   with   F1 - sc o re s   of   9 3 . 1 %   a n d   9 0 . 2 %   b a se d   on   t h e   d a tas e t   u ti li se d .   T h e se   re su lt s   a li g n   to g e th e r   with   a u to m a ted   m a c h in e   lea rn in g   ( Au t o M L - li n k e d st u d ies ,   wh il e   fa ll   s h o rt   of   m a n u a ll y   f in e - tu n e d   d e e p - lea rn in g - b a se d   s y ste m s   c re a ted   th ro u g h   h u m a n   sp e c ialists.   To   c o n c lu d e ,   e x p lo ri n g   th e   p o te n ti a l   of   c o m b in i n g   m a n u a l   e x p e rt   wo rk   a n d   a u to m a ted   d e e p   lea rn i n g   c o u ld   be   a   p ro m isin g   d irec ti o n   fo r   e n h a n c in g   e fficie n c y   in   p lan t   d e fe n c e .   K ey w o r d s :   Au to ML     Deep   lear n in g   Hy p er p ar a m eter s   Sin g u lar   v alu e   d ec o m p o s itio n   W ee d s     T h is   is   an   o p e n   a c c e ss   a rticle   u n d e r   th e   CC   BY - SA   li c e n se .     C o r r e s p o nd ing   A uth o r :   Ab d   Ab r ah im   Mo s s lah   C o lleg e   of   I s lam ic   Scien ce ,   U n iv er s ity   of   An b ar   An b ar ,   Fallu jah ,   I r aq   E m ail:   aisl . ab id e@ u o an b ar . ed u . iq       1.   I NT RO D UCT I O N   In   r ec e n t   tim es,   t h e   n e g ativ e   i m p ac t   of   wee d s   h as   led   to   s ig n if ican t   g l o b al   c r o p   lo s s es,   an d   th is   tr e n d   is   ex p ec ted   to   co n tin u e   in   th e   f u tu r e   [ 1 ] .   W h ile   tr ad itio n al   m eth o d s   in v o lv e d   th e   u s e   of   p esti cid es   to   tack le   th is   is s u e,   t he   eu r o p ea n   u n io n   ( E U)   is   in cr ea s in g ly   ad o p tin g   a   p o licy   aim ed   at   d ec r ea s in g   th e   u s ag e   of   p la n t   p r o tectio n   p r o d u cts,   o win g   to   ap p r eh e n s io n s   r eg ar d i n g   ch e m ical   r esid u es   on   cr o p s ,   en v ir o n m en tal   p o llu tio n ,   an d   th e   p o ten tial   f o r   d r u g   [ 2 ] .   As   p ar t   of   th is   p o licy ,   i n   th e   co m in g   d ec ad e,   th e   EU   aim s   to   r ed u ce   p esti cid e   ap p licatio n   by   50%   [ 3 ] .   As   a   r esu lt,   au to m atic   weed   co n tr o l   is   b ein g   o b s er v ed   as   a   p r o m is in g   s o lu tio n   to   r ed u ce   th e   r elian ce   on   c h e m ical   h er b icid es   f o r   weed   m an ag em en t   [ 4 ] .   T h e   r ec e n t   p r o g r ess   in   im ag e   class if icatio n   m eth o d s   o f f er s   an   o p p o r tu n ity   to   en h an ce   au t o m atic   weed   co n tr o l.   W h ile   th e r e   h as   b ee n   a   p au s e   in   th e   a d o p tio n   of   t h ese   m eth o d s   in   th e   s ec to r   of   ag r icu ltu r e,   th eir   u s e   is   r a p id ly   g ain in g   m o m en tu m .   I m ag e   an aly s is   b ased   on   m ac h in e   lear n in g   o f f er s   a   s p ee d y ,   non - in v asiv e,   an d   n o n - d estru ctiv e   s o lu tio n   f o r   ad d r ess in g   weed   g r o wth .   L ea r n in g   p r o to co ls   h av e   b ee n   u s ed   in   th e   ag r icu ltu r al   f ield   to   id en tify   wee d s   an d   id en tify   d is ea s es   th at   af f ec t   p lan ts   [ 5 ] ,   [ 6 ] .   Am o n g   v ar i o u s   ap p r o ac h es,   co n v o lu tio n al   n eu r al   n etwo r k s   Evaluation Warning : The document was created with Spire.PDF for Python.
                      I SS N :   2 7 2 2 - 3 2 2 1   C o m p u t Sci  I n f   T ec h n o l Vo l.  7 ,   No .   2 J u ly   20 26 1 67 - 1 78   168   ( C NNs )   ar e   p r esen tly   th e   m o s t   k n o wn   due   to   th eir   ab ilit y   to   o v er co m e   ce r tain   ch allen g es,   th is   in clu d es   f ac to r s   s u ch   as   s im ilar ities   b e twee n   d if f er e n t   class es   with in   a   p lan t   f am ily ,   as   well   as   s ig n if ican t   v ar iatio n s   with in   a   class   due   to   b ac k g r o u n d ,   co lo r ,   o cc lu s io n ,   p o s e,   an d   li g h tin g   co n d itio n s .   B esid es   th eir   ex ce llen t   class if icatio n   p er f o r m an ce ,   s o m e   s tu d ies   h av e   h ig h lig h ted   t h e   p o ten tial   of   d ee p   n eu r al   n e two r k s   ( DNNs)   f o r   r ea l - tim e   weed   co n t r o l   in   a g r icu ltu r e,   b ased   on   th eir   in f er en ce   tim es   [ 7 ] .   E v en   wit h   v ar io u s   p r o p o s ed   tech n iq u es,   alth o u g h   d ee p   lea r n in g   m o d els   h a v e   b ee n   ap p lied   in   ag r icu ltu r e,   im p lem en ti n g   th ese   s o lu tio n s   f u lly   is   s till   ch allen g in g ,   p r im ar ily   due   to   th e   co m p lex ity   of   th e   ag r ic u ltu r al   en v ir o n m en t .   T h is   n ec ess itate s   th e   u s e   of   co m p lex   m ac h i n e   v is io n   alg o r ith m s   th at   r eq u ir e   iter ativ e   f in e - tu n in g   [ 8 ] .   B u ild in g   a   f itti n g   d ee p   lear n in g - b ased   s y s tem   in v o l v es   in teg r atin g   a   more   co m p o n en ts ,   as   f ea tu r e   d etec tio n ,   f ea tu r e   elicitatio n ,   a n d   class if ier .   T h e   task   n ec ess i tates   ex p er ien ce   in   s elec tin g   s u itab le   m o d el   ar c h itectu r es,   ex p er tis e   in   m ath em atics,   im ag e   an aly s is ,   an d   co d in g   [ 9 ] ,   [ 1 0 ] .   As   a   r esu lt,   ac h iev in g   th e   b e s t   p o s s ib le   s y s tem   p er f o r m an ce   r e q u ir es   co n s id er ab le   ex p er im e n tatio n   tim e,   an d   a   team   of   ex p er ts   is   n ee d ed   by   m an u ally   test in g   d if f er en t   m o d els   an d   co n f i g u r atio n .     B ased   on   th e   ab o v e,   th e   p la n t   n ee d s   to   be   r etr ai n ed   in   iter ativ e   p r o ce s s es,   as   th e   d if f er en ce s   in   co n d itio n s   ar e   clea r   b etwe en   p ests   an d   cr o p s   ac co r d in g   to   r eg io n s   an d   r e g io n s .   C o n s eq u en tly ,   th e   ab ilit y   to   p r o d u ce   au to m atica lly   a   Pr o s ag er   L ea r n i n g   tailo r ed   f o r   e v er y   u n iq u e   s itu ated ,   ev en   b ef o r e   in d iv id u als   with o u t   ex ten s iv e   ex p er tis e,   wo u ld   be   ex tr em ely   b en ef icial.   T h ese   s y s tem s   ar e   d ev is ed   to   a u to m atica lly   ev alu at e   m u ltip le   p ip elin e   co n f ig u r ati o n s   an d   en h an ce   p er f o r m an c e   iter ativ ely .   B u t   one   of   th e   b ig g est   is s u es   with   au to m ated   m ac h in e   lear n in g   ( Au to ML s y s tem s   is   th eir   h ig h   d em a n d   f o r   co m p u tin g   r eso u r ce s .   To   ad d r ess   th is   is s u e,   IT   co m p an ies   s u ch   as   Go o g le,   Mic r o s o f t,   an d   Ap p le   h av e   in tr o d u ce d   u s er - f r ie n d ly   Au to ML   clo u d   s o lu tio n s ,   s ev er al   co m m er cial   Au to ML   s o lu tio n s ,   co m m er cial   Au to ML   s o lu tio n s   o f f er ed   by   co m p a n ies   s u ch   as   Go o g le,   Ap p le   an d   Mic r o s o f t,   p r o v id e   s im p le   way s   to   u s e   an d   tr ain   m o d els   with   litt l e   n ee d   f o r   a r tific ial   in tellig en ce   k n o wled g e .   C o n v er s ely ,   o p en - s o u r ce   to o ls   Au to Sk lear n ,   Au to Ker as,   H2 O   Au to ML ,   Au to - W E KA,   au to x g b o o s t,   T POT,   an d   OB OE   h av e   em er g ed   to   in cr ea s e   k n o wled g e   of   A u to ML   p latf o r m s   b en ef its   an d   d r awb ac k s .   T a b le   1   p r o v id es   a   s u m m ar y   of   th ese   s y s tem s .       T ab le   1 .   Ov e r v iew   o f   v ar io u s   au to m ated   d ee p   lear n i n g   ( a u to ML )   s y s tem s   A u t o M L   s y st e m   Te c h n o l o g y   t y p e   R e f e r e n c e   G o o g l e   C l o u d   A u t o M L   C l o u d   s o l u t i o n   [ 1 1 ] ,   [ 1 2 ]   A u t o S k l e a r n   Li b r a r y   [ 1 3 ]   TPO T   Li b r a r y   [ 1 3 ]   A u t o K e r a s   Li b r a r y   [ 1 4 ]   O B O E   Li b r a r y   [ 1 4 ]       2.   RE L AT E D   WO RK S   In   r ec en t   y ea r s ,   Au to ML   h as   b ee n   ap p lied   in   th e   ag r icu lt u r a l   s ec to r   to   p r o ce s s   v ar io u s   ty p es   of   d ata,   in clu d in g   tim e   s er ies,   s atellite   an d   g r o u n d - b ased   p ictu r e s .   Fo r   in s tan ce Hay ash i   et  a l .   [ 1 1 ] ,   e m p lo y e d   Au to ML   to   id en tif y   p est   in s ec t   s p ec ies,   co n s tr u ctin g   m o d e ls   with   im ag es   of   th r ee   a p h id   s p ec ies   th at   wer e   tr ain ed   in   Go o g le   C lo u d   Au to ML   Vis io n .   W ith   400   im ag es   p er   class ,   th e   m o d el   ac h iev ed   a   co r r ec t   r ec o g n izatio n r ate   of   o v er   9 6 %,   d em o n s tr atin g   th e   p o te n tial   of   r ec o g n izatio n o f   p est   s p e cies   u s in g   A u to ML .   Similar ly ,   in   [ 1 2 ] ,   th e   au th o r   u tili ze d   A u to ML   to   class if y   f r u its ,   b u tter f lies ,   an d   lar v al   h o s t   p la n ts   an d   ac h iev ed   an   esti m ated   av e r a g e   ac cu r ac y   of   9 7 . 1 %.   In   A u to ML   was   in teg r ated   u s in g   class if y in g   n eu r al   n etwo r k   tech n iq u es   r ice   b last   d is ea s e   b ased   on   f iv e   y e ar s   of   c o n tin u o u s   clim ate   d ata,   ac h iev in g   an   8 9 %   ac cu r ac y   in   ex ac er b atio n   ca s es.   Ad d itio n ally ,   L ee   et  a l .   [ 1 3 ] ,   d em o n s tr ated   th e   ef f ec tiv en ess   of   Au to ML   in   cr ea tin g   m ap s   of   Par th en iu m   g r ass   u s in g   m o d els   b u ilt   with   s atellite   im ag es   f r o m   L an d s at   8   an d   Sen tin el - 2.   T h e   Au to ML   m o d el   attain e d   a   class if icatio n   ac cu r ac y   of   7 4 %   with   L an d s at   8   an d   8 8 . 1 5 %   with   Sen tin el - 2 . ,   h ig h lig h tin g   th e   u s ef u ln ess   of   Au to ML   in   cr ea tin g   weed   d is p er s al   m ap s   u s in g   s atellite   im ag er y .   Fin ally ,   Aco s ta - Gam b o a   et  a l.   [ 1 4 ] ,   co m p ar ed   Au t o Ker as   with   tr an s f er   lear n in g   m eth o d s   f o r   h ig h - th r o u g h p u t   p lan t   p h en o t y p in g   in   ass ess in g   wh ea t   lo d g in g   u s in g   d r o n e   im ag er y .   Alth o u g h   p r e v io u s   r esear ch   h as   ex am in ed   Au to ML ,   th er e   r em ain s   a   n ee d   to   ass ess   th e   t ec h n iq u es   ca p ac ity   f o r   g en er ality   u s in g   d iv er s e   p ictu r es   ca p tu r e d   u n d er   r ea l - wo r ld   c o n d itio n s .   To   e n h an ce   ac ce s s ib ilit y   an d   r ep r o d u cib ilit y ,   it   is   ess e n tial   to   em p lo y   u s e   o p en - s o u r ce   alter n ativ es   in s tead   of   clo u d - b ased   p r o p r ietar y   o n es.   T h is   s tu d y   ev alu ates   th e   ef f icac y   of   Au to ML   s y s tem s   b ased   on   o p en - s o u r ce   s o lu ti o n s   as   a   m ea n s   of   ac ce ler atin g   an d   s tr ea m lin in g   th e   u s e   of   v is io n   an d   m ac h in e   lear n in g   ap p licatio n s   in   ag r ic u lt u r e.   T h e   p r im a r y   o b jectiv e   of   th is   s tu d y   is   to   d eter m in e   wh et h er   Au t o ML   t ec h n iq u es   ca n   co m p ete   with   m an u ally - d esig n ed   ar ch itectu r es.   T h r ee   p r im ar y   co n tr ib u tio n s   ar e   p r esen ted   in   th is   p ap er :   i )   a   p r o ce d u r e   w ith   two   s tag es   th at   u tili ze s   Au to ML   to   d ee p   lear n in g   c o m p o n en t   ex tr ac tio n   an d   class if ier   en s em b les   f o r   p lan t   id en tific atio n ;   ii )   w e   ex clu s iv ely   u s ed   o p e n - s o u r ce   Au to ML   f r am ewo r k s   f o r   our   im p lem e n tatio n ,   alo n g   with   two   p u b licly   Evaluation Warning : The document was created with Spire.PDF for Python.
C o m p u t Sci  I n f   T ec h n o l     I SS N:   2722 - 3 2 2 1       P erfo r ma n ce   ev a lu a tio n   o f th e   d ee p   lea r n in g   s ystem  fo r   w ee d   r ec o g n iz a tio n   ( A b d   A b r a h im   Mo s s la h a )   169   ac ce s s ib le   d atasets ,   to   f ac ilit ate   tr an s p ar e n t   an d   r ep r o d u cib le   r esear ch ;   an d   iii )   th is   s tu d y   ai m s   to   e v alu ate   th e   r eliab ilit y   an d   s u s ce p tib ilit y   of   Au to ML   s y s tem s   to   o v er f itt in g   u s in g   n o is y   d ata   s am p les.   T h e   m eth o d o l o g y   an d   ex p er im en tal   s etu p   a r e   p r esen ted   in   s ec tio n   2,   f o llo wed   by   th e   r esu lts   in   s ec tio n   3.   T h e   im p licatio n s   of   th e   f in d in g s   an d   th e   s u itab ili ty   of   th e   m eth o d o lo g y   ar e   th e   p ap er   c o n clu d es   by   o u tlin in g   f u t u r e   r esear ch   d ir ec tio n s   in   s ec tio n   5,   as   d is cu s s ed   in   s ec tio n   4.   T ab le   1   p r o v id es   an   o v er v iew   of   v ar i o u s   au to m ated   d ee p   lear n in g   ( A u to ML )   s y s tem s ,   in clu d in g   th o s e   u s ed   in   ag r icu lt u r al   ap p licatio n s .         3.   M E T H O D   3 . 1 .     T he   pro po s ed  a pp ro a ch   T h e   p u r p o s e   of   th is   r esear ch   is   to   in v esti g ate   th e   ef f ec tiv e n ess   of   Au to ML   in   id e n tify in g   d if f er e n t   ty p es   of   wee d s .   T h e   r esear c h e r s   aim   to   d eter m i n e   wh eth e r   Au to ML   can   ac c u r ately   class if y   an d   d if f e r en tiate   am o n g   weed   s p ec ies   b ased   on   th eir   v is u al   ch ar ac ter is tics ,   s u ch   as   leaf   s h ap e,   tex tu r e ,   an d   co lo r .   T h e   r esear ch   is   im p o r tan t   b ec au s e   id e n tify in g   an d   c o n tr o llin g   wee d s   is   ess en tial   f o r   c r o p   m an a g em en t,   an d   tr ad itio n al   m an u al   m eth o d s   can   be   tim e - co n s u m in g   an d   co s tly .   If   Au to ML   p r o v es   to   be   an   e f f ec tiv e   to o l   f o r   weed   id en tific atio n ,   it   c o u ld   s ig n if ican tly   im p r o v e   weed   m a n ag em en t   p r ac tices   an d   in c r ea s e   cr o p   y ield s .   Ad d itio n ally ,   th e   s tu d y   m ay   co n tr ib u te   to   t h e   d ev el o p m en t   of   more   a d v an ce d   an d   au to m ated   s y s tem s   f o r   ag r icu ltu r al   ap p licatio n s .     3 . 2 .     T he   s ing ula v a lue dec o m po s it io ( SVD t heo ry   T h e   SVD   is   a   m ath em atica l   tech n iq u e   u s ed   to   b r ea k   d o wn   a   m atr ix   in to   its   co n s titu en t   p ar ts .   T h ese   p ar ts   in clu d e   a   s et   of   v ec to r s   t h at   ar e   p er p en d icu la r   to   each   o th er   an d   h av e   a   len g th   of   o n e ,   as   well   as   a   s et   of   s in g u lar   v alu es   th at   r ep r esen t   th e   s tr en g th   of   each   v ec to r .   T h e   lar g est   s in g u lar   v alu e   co r r e s p o n d s   to   th e   m o s t   im p o r tan t   v ec t o r ,   wh ile   th e   s m allest   s in g u lar   v alu e   r ep r esen t s   th e   least   im p o r tan t   v ec to r .   T h e   SVD   can   be   u s ed   f o r   v ar i o u s   ap p licatio n s ,   s u ch   as   r e d u cin g   th e   s ize   of   a   d ataset   wh ile   r etain in g   im p o r tan t   in f o r m atio n ,   id en tif y in g   r elev a n t   f ea tu r es   in   a   d a taset,   an d   f ilter in g   o u t   n o is e   f r o m   a   s ig n al.   T h is   tech n iq u e   is   v alu ab le   f o r   an al y zin g   an d   m an ip u latin g   m atr i ce s   in   m an y   d if f er e n t   co n tex t s .   It   is   im p o r tan t   to   u s e   th is   tech n iq u e   r esp o n s ib ly   an d   not   p r o m o te   it   as   a   to o l   f o r   s ca m   p r o g r am s   [ 1 5 ] .     3 . 3 .     T he   s o lutio n s   a rc hite ct ure   T h e   p ap e r   ev al u ates   a   m eth o d o lo g y   th at   c o m b in es   two   A u to ML   s tep s .   T h e   o b jectiv e   is   to   attain   co m p ar ab le   ef f icac y   to   co n v e n tio n al   m eth o d s ,   lik e   a   s o f t   m ax im u m   d ec o d e r   ato p   a   n eu r al - b ased   in f o r m atio n   ex tr ac to r ,   as   ev id e n ce d   by   ea r lier   r esear ch .   A   m eth o d o lo g y   co n s is ts   of   two   s tep s ,   wh er e   th e   f ir s t   s tep   u tili ze s   a   B ay esian   n eu r al   ar ch itectu r e   s ea r ch   ap p r o ac h   to   id en tif y   th e   m o s t   ef f ec tiv e   ex tr ac tio n   f ea tu r e   to o l   ca p ab l e   of   r e m o v in g   t h e   m o s t   s ig n if ic an t   f ea tu r es   f r o m   t h e   p ict u r es.   T h e   o u t p u t   of   th ese   s tep   is   a   DNN ,   au to m atica lly   f in e - tu n e d   th r o u g h   s ev e r al   co n v o lu tio n al   lay er s ,   T h e   d ef a u lt   ar ch itectu r e   s ea r ch   m et h o d   is   u s ed   to   o b tain   o p tim al   an d   r ele v an t   f ea tu r es   f r o m   th e   in p u t   im ag es.   T h is   ap p r o ac h   is   d ep icted   in   Fig u r e   1.   It   is   im p o r tan t   to   n o te   th at   th is   tech n iq u e   s h o u l d   not   be   p r o m o ted   as   a   to o l   f o r   s ca m   p r o g r a m s ,   b ec au s e   s u ch   an   ap p r o ac h   is   d esig n ed   to   in itiate   th e   a p p lica tio n   an d   d ev elo p m en t   of   im ag e   an aly s is   an d   m ac h in e   lear n in g   tech n iq u es.   T h e   f ir s t   s tep   of   th e   m eth o d o l o g y   in v o lv ed   e x tr ac tin g   th e   m o s t   ef f ec tiv e   ch ar ac ter is tics   of   th e   s o u r ce   p h o to s .   T h e   s ec o n d   s tep   co n c en tr ated   on   f i g u r in g   out   a   f u ll   p ip elin e   b ased   on   al g o r ith m ic   lear n in g   th at   m ig h t   p r o d u ce   th e   g r ea test   f in al   o u t co m es.   Stra teg ies   f o r   p ick in g   out   f ea tu r es,   r ed u cin g   s ize,   an d   ca teg o r izatio n   h av e   b ee n   test ed   in s id e   th is   p ip elin e.   Par ticu lar ly ,   s ev er al   m eth o d s   wer e   ass ess ed :     By   p r o jectin g   t h e   d ata   to   a   lo wer - d im en s io n al   s p ac e   u s in g   th e   SVD   of   t h e   d ata ,   th e   p r i n cip al   co m p o n en ted   m eth o d   [ 16 ]   p r o v id es   a   p r o ce d u r e   f o r   lin ea r   d e cr ea s in g   d im e n s io n ality   th at   l ess en s   th e   r is k   of   th e   ex ce s s iv e   f itti n g .   Un lik e   o th er   m eth o d s   of   d im en s io n ality   r ed u ctio n ,   p r io r   to   a d o p tin g   th e   SVD,   in p u t   f ea tu r es   wer e   ce n tr ed   t h o u g h   not   r esized .   It   is   im p o r tan t   to   u s e   th is   tech n iq u e   eth ically   an d   n o t   p r o m o te   it   as   a   to o l   f o r   s ca m   p r o g r am s .     An o th er   tech n iq u e   em p lo y e d   to   less en   o v er esti m atio n   an d   r ed u ce   d im en s io n s   is   T VD   [ 17 ] .   Ho wev er ,   u n lik e   th e   p r ev io u s   m eth o d ,   with   th is   m eth o d ,   th e   d ata   is   not   ce n tr ed   b ef o r e   th e   SVD   is   ca lcu lated .   It   is   im p o r tan t   to   n o te   th at   t h is   tech n iq u e   s h o u ld   n o t   be   p r o m o te d   as   a   to o l   f o r   s ca m   p r o g r am s .     C o m p ar ab le   to   th e   in itial   s tr ateg y ,   but   em p l o y in g   k er n els   f o r   not   lin ea r   r e d u ctio n   of   d im en s io n ality   in s tead   of   r eg u la r   d im i n u tio n   of   d im en s io n al   is   Ker n el   p r i n cip al  co m p o n e n an al y s is   ( PC A )   [ 18 ] .   It   i s   g o al   is   to   en h an ce   th e   class if ier s   ca p ac ity   f o r   g en e r aliza tio n   by   elim in atin g   r ed u n d an t   f ea tu r es .   It   is   im p o r tan t   to   u s e   th is   tech n iq u e   eth ically   an d   not   p r o m o te   it   as   a   to o l   f o r   s ca m   p r o g r am s .     Ad aBo o s tin g   is   an   en s em b le   lear n in g   ap p r o ac h   th at   can   be   o n ly   one   class if ier   is   em p lo y e d   or   as   one   of   th e   en s em b le   co m p o n en ts .   T h ey   m ak e   a   m eth o d   ca lled   b o o s tin g ,   in   wh ich   th e   ch o ice   tr ee s   ar e   r ep ea ted ly   tr ain ed ,   g iv i n g   m o r e   weig h t   to   th e   ex am p les   f o r   wh ich   th e   f o r ec ast   is   in co r r ec t.   T h is   te ch n iq u e   s h o u ld   not   be   p r o m o ted   as   a   to o l   f o r   s ca m   p r o g r am s .   Evaluation Warning : The document was created with Spire.PDF for Python.
                      I SS N :   2 7 2 2 - 3 2 2 1   C o m p u t Sci  I n f   T ec h n o l Vo l.  7 ,   No .   2 J u ly   20 26 1 67 - 1 78   170     E x tr a   tr ee s   is   a   m ac h in e   lear n in g   tech n i q u e   t h at   can   be   o n l y   one   ca teg o r y   is   e m p lo y e d   or   as   co m p o n e n t   in s id e   a   g r o u p .   T h e   r esem b la n ce   to   er r atic   f o r ests ,   but   u n lik e   er r atic   f o r ests ,   no   b o o ts tr ap   s am p lin g   is   u s ed .   As   a   r esu lt,   it   m ay   be   more   p r o n e   to   o v er f itti n g .   A n o th er   d if f er en ce   b etwe en   E x tr a   T r ee s   an d   r an d o m   f o r ests   is   th at   E x tr a   T r ee s   u s in g   an   ar b itra r y   cu t   to   cr ea te   n o d es   with in   th e   b r an c h ,   th at   can   h elp   to   r ed u ce   to o   tig h t.   It   is   im p o r tan t   to   u s e   th is   tech n iq u e   eth ically   an d   not   p r o m o te   it   as   a   to o l   f o r   s ca m   p r o g r a m s .           Fig u r 1 .   Dis p lay s   s am p les o f   im ag es f r o m   th e   b en c h m ar k   d atasets .   T h f ir s t - r o w   ex h ib its   p ictu r es f r o m   th ea r ly   cr o p   wee d s   d ataset,   wh ile  th s ec o n d - r o w   f ea tu r es p ictu r es f r o m   t h p lan t seed lin g s   d ataset       T h e   p u r p o s e   of   th is   s tu d y   is   to   d eter m in e   if   u s in g   p r e d icto r   b an d s   m a y   en h an ce   e f f icien c y   or   less en   v ar ian ce   in   th e   class if icatio n   task   r esu lts .   T h e   g r o u p   m eth o d   th ese   wo r d s   a   m ajo r ity   a p p r o v al   m eth o d ,   in   wh ich   th e   f in al   p r e d ictio n   is   t h e   p r o jecte d   ca teg o r y   th at   r ec eiv es   th e   m o s t   v o tes   f r o m   ea c h   class if ier .   Usi n g   B ay es   o p tim izatio n ,   th is   r o u te   was   au to m atica lly   s elec ted .   It   is   im p o r tan t   to   n o te   th at   o p en - s o u r ce   s o lu tio n s   wer e   u tili ze d   in   th e   d ev el o p m en t   of   th i s   p r o ce s s ,   wh ile   th e   f in is h ed   p ip e   m ay   be   im p o r ted   an d   put   to   u s e   f o r   a   s elf - s u f f icien t   weed   s u r v eillan ce   s y s tem .   W ith   r eg ar d   to   co m p u tin g   lim itatio n s   an d   d elay ,   th e   s y s tem   can   be   im p lem en ted   as   a   s ep ar ate   or   o n lin e   o p tio n .   It   is   cr u cial   to   u s e   th is   tech n o l o g y   eth ically   an d   av o id   p r o m o ti n g   it   as   p ar t   of   an y   s ca m   p r o g r am s .     3 . 4 .     M a k ing   ex perim ent a l   c ho ices   To   g ain   a   b etter   u n d e r s tan d in g   of   t h e   Au to ML   p r o ce s s   an d   id en tify   its   ad v an tag es   an d   li m itatio n s ,   ce r tain   ex p er im e n tal   lim itatio n s   wer e   estab lis h ed ,   d esp ite   th e   f ac t   th at   Au to ML   d o es   n o t   r eq u ir e   an y   s p ec ial   s etu p   to   o p er ate.   Du r i n g   th e   m eth o d o lo g y   e v alu atio n ,   ce r tain   th e   h y p e r p ar am ete r   s ettin g   of   t h e   Au t o ML   p r o ce s s   was   k ep t   co n s is ten t.   T ab le   2   p r esen ts   th e   s elec ted   h y p er p ar a m eter s   f o r   th e   Au t o ML   p ip elin e   b ased   on   th eir   p r o m is in g   p er f o r m an ce   a n d   a v ailab ilit y   of   co m p u tatio n al   r eso u r ce s .   T h e   B ay esian   o p tim izatio n   alg o r ith m   was   r u n   up   to   35   tim es   to   id en tify   th e   to p   ex tr ac to r   of   f ea tu r es,   with   an   in i tial   b atch   ca p ac ity   of   eig h t   an d   an y   d e p th   m o d el   test ed   f o r   an   ag g r eg ate   of   1 0 0   tim es.   R eg ar d in g   th e   s o r tin g   g r o u p ,   each   m o d el   was   tr ain e d   f o r   a   m ax im u m   of   2   m in u tes,   an d   all   m o d els   wer e   tr ain ed   f o r   a   to tal   of   20   m in u tes.   Data   au g m en tatio n   tech n iq u es   wer e   ap p lied   to   th e   im ag es   b ef o r e   f ea tu r e   ex tr ac tio n ,   in clu d in g   h o r iz o n t al   r o tatio n ,   cr o p p in g ,   s ca lin g ,   an d   m ir r o r in g .   To   im p r o v e   th e   Au t o ML   p r o g r a m s   ab ilit y   to   g en er alis e,   all   p h o to s   wer e   r ed u ce d   to   6 5 ×6 5   p ix els   in   o r d er   to   r em o v e   a n y   ass o ciatio n   b etwe en   b o th   p h o t o   s ize   an d   th e   ac tu al   s ize   of   th e   p la n ts .   In   th e   co n tr ar y ,   s ev er al   ty p e s   of   a r r an g em en ts   wer e   put   to   th e   test   ex p er im e n tally   in   o r d er   to   d eter m in e   w h ich   was   b est.   T h is   p r o ce s s   aim ed   at   test in g   th e   r o b u s tn ess   of   s p ec if ic   h y p er p ar am eter s   r e f er   to   T ab le   3   in   th e   Au to ML   p ip eli n e   d esig n   ag ain s t   o th e r   m o d if icatio n s .   T h e   b ac k g r o u n d s   p r esen ce   in   th e   p h o to   m ig h t   g r ea tly   s ig n if ican tly   af f ec t   th e   f ea tu r e   ex tr ac tio n   p r o ce s s .   Hen ce ,   th e   u s e   of   p lan t   s eg m en tatio n   was   ev alu ated .   H u e - s atu r atio n - v al u e   ( HSV)   co l o u r   s ch em e   was   u s ed   as   th e   th r esh o ld s   ap p r o ac h   f o r   s eg m e n ted   Evaluation Warning : The document was created with Spire.PDF for Python.
C o m p u t Sci  I n f   T ec h n o l     I SS N:   2722 - 3 2 2 1       P erfo r ma n ce   ev a lu a tio n   o f th e   d ee p   lea r n in g   s ystem  fo r   w ee d   r ec o g n iz a tio n   ( A b d   A b r a h im   Mo s s la h a )   171   im p lem en tatio n .   M o r eo v er ,   u s in g   n o is ier   d ata   in   th e   tr ain i n g   s tag e   was   ass ess ed   as   a   way   to   im p r o v e   t h e   au to n o m o u s   weed   r ec o g n izati o n s y s tem s   p er f o r m an ce .   T h e   f ea tu r e   ex tr ac to r   tr ai n in g   p r o ce s s   also   in clu d ed   ex p lo r in g   wh eth er   th e   So f tm ax   alg o r ith m   a n d   th e   lay e r   of   co n v o lu tio n   s h o u ld   be   c o u p led   to g eth er   in   a   n etwo r k .   Fin ally ,   o n ce   th e   f ea tu r es   wer e   ex tr ac ted ,   th e   in p u t   im ag e   c o u ld   be   class if ied   u s in g   a   s in g le   alg o r ith m ,   an   e n s em b le   of   clas s if ier s ,   or   a   So f tm ax   p r ed icto r .   All   of   t h ese   o p tio n s   wer e   ass ess ed .   T ab le   3   lis ts   th e   h y p er p ar am eter s   th at   w er e   ass ess ed   d u r in g   th e   ev alu atio n   p r o ce s s   to   id en tify   th e   m o s t   o p tim al   co n f ig u r atio n   f o r   th e   au to m l   p ip elin e.       T ab le   2 .   Fix ed   h y p e r p ar am ete r s   f o r   th e   e x p er im e n ts   F i x e d   h y p e r p a r a me t e r s   V a l u e   M a x i m u m   n u m b e r   of   t r i a l s   p e r   d e e p   mo d e l   35   Ep o c h s   1 0 0   I mp l e me n t a t i o n   of   i ma g e   mo d i f i c a t i o n   t e c h n i q u e s   b a s e d   on   g e o m e t r y   y e s   I mag e   si z e   65 × 65   S a mp l e   s i z e   8   M a x i m u m   t i me   a l l o c a t e d   f o r   f i t t i n g   e a c h   mo d e l   2   m i n   To t a l   t i me   t a k e n   to   f i n d   t h e   b e st   c l a ss i f i e r   20   mi n       T ab le  3 .   Hy p er p ar a m eter   co n f ig u r atio n s   ass ess ed   f o r   Au to M L   p ip elin ev alu atio n   V a r i a b l e s   of   t h e   h y p e r p a r a m e t e r s   e v a l u a t e d   F e a t u r e   e x t r a c t i o n   Emp l o y me n t   of   a   f u l l y - c o n n e c t e d   n e t w o r k   {Y e s,   N o }   Ev a l u a t i o n   of   t h e   i mp a c t   of   se g m e n t i n g   p l a n t   r e g i o n s   in   i m a g e s   on   f e a t u r e   e x t r a c t i o n   {Y e s,   N o }   N o i s y   t r a i n i n g   {Y e s,   N o }   Ty p e   of   c l a s si f i e r   {S i n g l e ,   S o f t m a x ,   E n se mb l e }       3 . 5 .     Appl ied   da t a s et s   T h is   s tu d y   u tili ze d   two   m ain   d atasets :   i )   th e   ea r ly   cr o p   weed   d ataset,   wh o s e   d ata   wer e   t ak en   f r o m   th e   r esear ch   [ 6 ] ,   co n s is ted   of   504   R GB   p h o to s   f ea tu r in g   f o u r   d is tin ct   s p ec ies   d u r in g   th eir   ea r ly   p h ases   of   g r o wth a n d   ii)   w ith   an   ac tu al   r eso lu tio n   of   r o u g h ly   10   p ix el s   ev er y   m illi m eter ,   th e   Plan t   S ee d lin g s   co llectio n   in clu d ed   R GB   p h o to s   of   r o u g h ly   9 6 0   d is tin ct   p lan ts   at   v a r io u s   s tag es   of   d ev elo p m en t   t h at   b elo n g e d   to   12   d if f er en t   s p ec iesAd d itio n al   d etails   r eg ar d in g   t h is   s et   of   d ata   can   be   f o u n d   in   [ 1 5 ] .   F i g u r e   2   s h o wca s es   ex am p les   of   p ictu r es   f r o m   b o th   k in d s   of   d ata,   a   f ew   of   wh ich   h av e   u n d er g o n e   s eg m e n tatio n   of   p lan ts .   T h e   f ir s t   d ataset   f ea tu r es   v ar ia b le   illu m in atio n   co n d itio n s ,   w h ich   ch allen g es   t h e   Au to ML   p r o g r am s   ca p ac ity   to   g en er alize   an d   d is r eg a r d   illu m in atio n   lev els   wh en   id en tify in g   cr o p s   an d   wee d s .   I m a g es   f r o m   in d o o r   p lan ts   cu ltiv ated   in   a   g r o w   r o o m   wit h   lig h tin g   t h at   is   ar tific ial   ad d ed   to   s u n lig h t   m ak e   up   th e   f o l lo win g   d ata.   Sin ce   th e   d ata   wer e   g ath er e d   in   a   l ab ,   it   i s   p o s s ib le   th at   s o m e   c h ar ac ter is tics   an d   m o r p h o lo g i ca l   tr aits   of   p la n ts   cu ltiv ated   o u ts id e   ar e   a b s en t   as   in   Fig u r 1 .           Fig u r 2 .   No is y   s am p les f r o m   b o th   d atasets .   Fro m   lef t to   r ig h t: p ep p er   b lu r r y   ch a r lo ck   a n d   f at  h en   with   s alt,   n o is to m ato   with   s alt  an d   p ep p er   n o is e,   b l u r r y   co tto n       3 . 6 .     Ana ly s is   T h e   Au to ML   s y s tem s   ex ec u tio n   was   an aly s ed   u s in g   th e   F1 - s co r e   ( 1 ) .   R ec all   is   a   r atio   of   ac cu r ate   ca teg o r ies   f o r   th e   in itial   in f o r m atio n   s et,   wh ile   ac cu r ac y   is   th e   f in al   r atio   of   t h e   r ig h t   la b els   in   th e   m o d el s   o u tp u t.   T h is   s tatis tic   is   f r eq u e n tly   u s ed   in   class if icatio n   p r o b lem s   [ 1 9 ] .   Sin ce   b o t h   d ataset s   in   th is   s tu d y   wer e   m u lti - class   p r o b lem s   with   a   class   im b alan ce ,   in   co m p ar is o n ,   we   ca lcu lated   th e   m icr o - av er ag in g   F1   g r a d e ,   wh ich   is   a   p r ef e r ab le   a g g r e g a tio n   m eth o d   o v er   th e   m ac r o - av er ag e.   To   co n d u ct   s tatis tical   co m p a r is o n s ,   we   Evaluation Warning : The document was created with Spire.PDF for Python.
                      I SS N :   2 7 2 2 - 3 2 2 1   C o m p u t Sci  I n f   T ec h n o l Vo l.  7 ,   No .   2 J u ly   20 26 1 67 - 1 78   172   u s ed   th e   r o b u s t,   p air ed   n o n - p ar am etr ic   s tatis tical   test s   [ 1 ] ,   [ 2 0 ] .   T h ese   test s   wer e   em p lo y ed   to   p r ev en t   d r awin g   to o   o p tim is tic   ass u m p tio n s .   T h e   in itial   test   was   co n d u cted   to   ass ess   co m p ar ab le   r esu lts   b etwe en   p ip elin e   s ets   s elec tio n   an d   co n tr o lled   f ea t u r es   ex tr ac tio n .   Me an wh ile,   th e   A   s ec o n d   test   was   r u n   to   co m p ar e   id en tical   p ip elin es   ca p ab ilit ie s   on   th e   f r esh   a n d   n o is y   in f o r m atio n   s ets.     1  = 2              +    (1 )     Av o id   o v er   f itti n g ,   it   was   ess en tial   to   m ea s u r e   th e   v ar iatio n   in   F1 - s co r e   b etwe en   th e   tes t   an d   tr ain   in f o r m atio n   s ets,   g iv en   th at   Au to ML   h as   th e   p o te n tial   to   lead   to   o v e r   f itti n g .   T h is   allo wed   us   to   d eter m in e   wh ile   th e   p ip elin e   f o r   Au to M L   p r o d u ce d   a   ca teg o r izer   ca p ab le   m eth o d   id en tify i n g   f r esh   s am p les   of   wee d s   in s tead   if   th at   m er ely   f it   t h e   tr ain in g   s et   an d   was   th u s   u n s u itab le   f o r   r ea l - wo r l d   ap p licatio n s   [ 21 ] .   Ad d itio n ally ,   t h e   e v alu atio n   of   th e   r o b u s tn ess   was   co n d u cted   u n d er   more   ch allen g in g   s ce n ar io s ,   s u c h   as   I m ag es   ar e   lo u d ,   h az y ,   an d   s p r in k led   with   s alt.   T h e   F1   r a n k in g   m icr o - av er a g in g   was   ad d i tio n ally   d eter m in e d   u s in g   co m p lete   n o is ier   in f o r m atio n   s ets,   s u ch   as   d ep icted   in   Fig u r e   2 .   T h e   p r o b lem s   a s s o ciate d   with   d ee p   lear n in g - b ased   tech n o l o g ies   r esil ien ce   was   ex ten s iv ely   ex p lo r ed   in   [ 22 ] .     3 . 7 .     So f t wa re   a nd   ha rdwa re   T h is   wo r k   u tili ze d   two   p r im ar y   s o f twar e   p ac k ag es:   Au to - Sk lear n   0 . 1 0 . 0   an d   Au to Ke r as   1 . 0 . 8 .   Au to Ker as   is   a   s o f twar e   to o l   th at   u tili ze s   B ay es   o p tim iz atio n - g u id e d   n etwo r k s   of   n e u r o n s   m o r p h ic   to   o p tim ize   b o th   ar ch itectu r e   a n d   h y p e r p ar am ete r s   f o r   th e   s elec tio n   of   th e   m o s t   au s p icio u s   p r o ce d u r es   at   ev er y   lev el.   Ker as   2 . 4 . 3   a n d   T en s o r f lo w   2 . 3 . 8   b ac k e n d s   ar e   u s ed   in   it   i s   o p er atio n   [ 23 ] .   T h e   s ec o n d   b u n d le   u s es   non - d ee p   lear n in g   tec h n iq u es   to   ac co m p lis h   Au t o ML ;   it   is   a   lib r ar y   t h at   is   o p e n - s o u r ce   ca lled   Au to - Sk lear n .   Fo r   tr an s f o r m in g   d ata   an d   au t o m ated   lear n in g ,   th is   b u n d le   u s es   th e   Scik it - L ea r n   au to m at ed   lear n in g   e n g in e   ( v er s io n   0 . 2 2 . 2 ) .   Au to - Sk lear n   u tili ze s   a   B ay es   Op tim is a ti o n   s ea r ch   tech n i q u e,   lik e   Au t o - Ker as,   to   q u ic k ly   id en tify   th e   b est   m o d el   p ip elin e   f o r   a   g iv en   c o llectio n   of   c h ar ac ter is tics .   Op en C V   3 . 4 . 2   was   u s ed   as   th e   im ag e   p r ep r o ce s s in g   lib r ar y ,   an d   all   th e   ex p er im en ts   wer e   co n d u ct ed   u s in g   Ub u n tu   1 8 . 0 4   as   th e   o p er atin g   s y s tem ,   alo n g   with   a   GeFo r ce   R T X   2 0 8 0 T i   GPU.     3 . 8 .     Alg o rit hm s   Her e   is   a   d escr ip tio n   of   th e   alg o r ith m   f o r   th e   Su p er   L ea r n er   m o d el   s tack in g   m et h o d :     Sp lit   th e   tr ain in g   d ata   i n to   K   e q u ally   s ized   f o l d s .     Fo r   each   f o l d   k,   t r ain   N   d iv e r s e   m ac h in e   lear n i n g   m o d els   on   th e   r em ain in g   K - 1   f o ld s .     Use   each   of   th e   N   m o d els   to   p r ed ict   th e   o u tc o m e   f o r   th e   k - th   f o ld .     C o m b in e   th e   p r ed ictio n s   f r o m   all   N   m o d els   f o r   th e   k - th   f o ld   to   f o r m   a   n ew   K - f o ld   d ata   s et.     T r ain   a   m eta - lear n er   on   th e   n e w   K - f o ld   d ata   s et.     Use   th e   m eta - lear n er   to   p r ed ic t   th e   o u tco m e   f o r   th e   test   d ata.   It   is   im p o r tan t   to   n o te   th at   in   s tep   2,   th e   N   m o d els   s h o u ld   be   d iv er s e   an d   u n co r r elate d   with   each   o th er ,   as   th is   wo u ld   lead   to   b etter   p er f o r m an ce .   In   s tep   5,   th e   m eta - lear n er   can   be   an y   m ac h in e   lear n in g   alg o r ith m ,   s u ch   as   lo g is tic   r eg r ess io n   or   a   n eu r al   n etwo r k .   T h e   s u p er   lear n er   m eth o d   is   u s ef u l   b ec au s e   it   can   ad ap t   to   d if f er en t   ty p es   of   d ata   an d   lear n   to   co m b in e   th e   s tr en g th s   of   d if f er en t   m ac h in e   lear n in g   alg o r ith m s .   T h e   m eth o d o lo g y   em p lo y ed   by   th e   alg o r ith m   in   th is   ar ticle   can   be   s u m m ar ized   as   f o llo ws:     Data   p r ep ar atio n :   two   d if f e r en t   b en ch m a r k   d atasets   co n tain in g   cr o p s ,   s ee d lin g s ,   an d   wee d s   wer e   p r ep r o ce s s ed ,   au g m e n ted ,   an d   d iv id ed   i n to   tr ain in g   an d   test in g   s ets.     I n teg r atio n   of   Au t o ML   s y s tem s :   two   d is tin ct   Au to ML   s y s tem s   wer e   in teg r ated   to   ev alu ate   th e   m eth o d o l o g y   of   weed   id e n tific atio n .   T h ese   s y s tem s   u s ed   d if f er e n t   alg o r ith m s   to   g e n er ate   m ac h in e   lear n in g   m o d els.     Per f o r m an ce   ev alu atio n :   th F1   s co r e   was   u s ed   to   id en ti f y   th e   b est   Au to ML   co n f ig u r atio n s ,   wh ich   m ea s u r es   th e   ac cu r ac y   a n d   p r ec is io n   of   th e   s y s tem s   p r ed ictio n s .   T h e   F1   s co r es   of   th e   s y s tem s   wer e   ev alu ated   on   th e   test in g   s ets.     Fu tu r e   wo r k :   th s tu d y   p r o p o s ed   p o ten tial   f u tu r e   wo r k   to   en h an ce   th e   p er f o r m an ce   a n d   r o b u s tn ess   of   th e   Au to ML   s y s tem s .   T h is   in clu d ed   test in g   th e   s y s tem s   with   n ew   d atasets   an d   n o is y   s am p les.     Su p er   lear n er   alg o r ith m :   th s tu d y   in tr o d u ce d   th e   Su p er   L ea r n er   al g o r ith m   as   a   m eth o d   to   c o m b in e   th e   s tr en g th s   of   v ar io u s   m ac h in e   lear n in g   alg o r ith m s .   T h e   alg o r ith m   was   ex p lain ed   in   d eta il,   in clu d in g   its   ad v an tag es   o v er   o t h er   m eth o d s .   In   c o n clu s io n ,   th e   m eth o d o lo g y   in clu d ed   ev al u atin g   th e   p er f o r m a n ce   of   Au to ML   s y s te m s   f o r   id en tif y in g   wee d s   an d   p r o p o s in g   f u tu r e   r esear ch   th at   co u ld   en h an ce   t h eir   ac cu r ac y   an d   ef f ec tiv e n e s s   with   a   p o s itiv e   im p ac t   on   in c r ea s in g   p r o d u cti o n   of   f ield   cr o p s .     Evaluation Warning : The document was created with Spire.PDF for Python.
C o m p u t Sci  I n f   T ec h n o l     I SS N:   2722 - 3 2 2 1       P erfo r ma n ce   ev a lu a tio n   o f th e   d ee p   lea r n in g   s ystem  fo r   w ee d   r ec o g n iz a tio n   ( A b d   A b r a h im   Mo s s la h a )   173   4.   RE SU L T S   AND   D I SCU SS I O N   T h is   s ec tio n   p r esen ts   th e   r esu lts   of   th e   ex p er im e n ts   to   d eter m in e   wh ich   Au to ML   p ip elin e   wo r k s   b est   with   ev er y   s et.   E ac h   p r o ce s s   c o n f ig   was   test ed   ten   tim es   u s i n g   v ar io u s   r an d o m   s ee d s   in   o r d er   to   in c r ea s e   th e   r eliab ilit y   of   th e   r esu lts .   T h e   m ed ian   F1 - s co r e   f o r   ev er y   co n f ig   on   ev er y   s et   of   d ata   ( o r ig in al/clea n   an d   n o is e   o n es)   was   th en   p r o v i d ed .   T h e   d ata   was   s p lit   u s in g   s tr atif ied   s p litt in g ,   wh er ein   25%   of   th e   s am p les   wer e   allo ca ted   f o r   v alid atio n ,   2 5 %   f o r   test in g ,   an d   50%   f o r   tr ain i n g .   W h en   a   s et   of   d ata   is   d esc r ib ed   as   h av in g   an   u n eq u al   v e r s io n ,   it   in d icate s   th at   n o is e   h as   b ee n   in clu d e d   in   25%   of   th e   tr ain in g   d ataset   s am p les   ( s o d iu m   an d   p ep p er )   an d   50%   of   t h e   s et   of   s am p les   o v er all.   T h e   o u tco m e   of   th e   tr ain in g   s et   an d   th e   o u t co m e   of   th e   test in g   s et   d if f er ,   as   s h o wn   by   th e   O v er f itti n g   lin e.   F1 - Sco r e   co lu m n   s h o ws   th e   p e r f o r m an ce   on   th e   test   s et.     4 . 1 .     Da t a s et   f o r   ea rly   cr o p   wee ds   Usi n g   th e   o r ig in al   d ataset   f o r   tr ain in g ,   T ab le   4   s h o ws   th e   to p   10   p er f o r m in g   Au to M L   p ip elin es   b ased   on   th eir   F1   s co r es   f o r   th e   ea r ly   cr o p   wee d s   d ataset.   To   en s u r e   co n s is ten cy ,   10   test s   wer e   co n d u cte d   f o r   ev er y   p ip elin e   d esig n   u s in g   v ar io u s   r an d o m   s ee d lin g s ,   an d   th e   r esu lts   g iv en   F1   s co r e   is   th e   av er ag e   s co r e   ac r o s s   all   r u n s .   T h e   p i p elin e s   wer e   tr ain ed   with   a   s tr atif ied   s p lit   of   50%   d ata   f o r   25%   v alid atio n ,   25%   ev alu atio n ,   an d   25%   lea r n in g .   Fo r   lo u d   d atasets ,   50%   of   th e   tr ain in g   d ata   was   c o n tam in a ted   with   eith er   s alt   an d   p ep p er   n o is e   or   h az in ess .   Fo llo win g   th e   co m p letio n   of   t h e   Frie d m an   ev al u atio n   at   a   0 . 1   co n f id en ce   lev el,   to   co m p a r e   a   d if f er e n t   d is tr ib u tio n ,   wh e r e   p lan t   s eg m en tati o n   was   u s ed ,   t h er e   was   a   n o ti ce ab le   d if f er e n ce   in   co m p ar in g   it   with   th e   alter n ati v e.   Ho wev e r ,   th e   So f tm ax   class if ier   s h o wed   h i g h er   v ar ia n ce   in   t h eir   r esu lts   an d   co u ld   be   co n s id er ed   less   r o b u s t.   So m e   p ip elin es   ac h iev ed   n ea r ly   100%   p er f o r m a n ce   at   th e   ex er cise   g r o u n d ,   in d icatin g   p o ten tial   o v er f itti n g .   As   an   illu s tr atio n ,   a   t o p - r an k ed   p ip elin e   h ad   an   a v er ag e   tr ain i n g   s et   p er f o r m an ce   of   9 9 . 9 8 %.   W h en   ev alu ated   on   a   n o is y   test   d ataset,   p er f o r m a n ce   d ec r e a s ed   in   m o s t   ca s es,   p h o to s   th at   ar e   h az y   ar e   more   d if f icu lt   to   ca teg o r is e   c o r r ec tl y .   So m e   p ip elin es   m ai n tain ed   ex ce llen t   o u t p u t   f o r   th e   n o is e   of   s alt   an d   p ep p er ,   w h ile   o th er s   s h o wed   a   g lar in g   lack   of   r esil ien ce   to   th ese   k in d s   of   n o is es.   U s i n g   a   v e r s i o n   of   t h e   d a t a s e t   f o r   t r a i n i n g   w i t h   a d d e d   n o i s e ,   in   T a b l e   5,   we   p r e s e n t   t h e   i m p a c t   of   t r a i n i n g   A u t o M L   p i p e l i n e s   w i t h   n o i s y   s a m p l e s .   S p e c i f i c a l l y ,   we   t r a i n e d   t h e   m o d e l s   on   a   d a t a s e t   t h a t   w a s   c o m p o s e d   of   5 0 %   c l e a n   s a m p l e s ,   2 5 %   s a l t   a n d   p e p p e r   s a m p l e s ,   a n d   2 5 %   b l u r r y   s a m p l e s .   U s i n g   a   0 . 1   p r o b a b i l i t y   t h r e s h o l d ,   we   r a n   a   F r i e d m a n   e x a m   a n d   d i s c o v e r e d   t h a t   t h e   i n i t i a l   f o u r   l i n e s   h a d   s i m i l a r   p e r f o r m a n c e s   b a s e d   on   t h e   F1   s c o r e   c o l u m n ,   but   t h e y   o u t p e r f o r m e d   t h e   o t h e r   p i p e l i n e s .   I n t e r e s t i n g l y ,   it   w a s   d i s c o v e r e d   t h a t   u t i l i z i n g   b o t h   p l a n t   r e c o g n i t i o n   a n d   a   f u l l y - c o n n e c t e d   p r e d i c t o r   t o g e t h e r   l e d   to   b e t t e r   p e r f o r m a n c e   l o u d   s u r r o u n d i n g s ,   w h i c h   is   in   c o n t r a s t   to   t h e   r e s u l t s   p r e s e n t e d   in   T a b l e   4.   T h e   F1   s c o r i n g   l i n e   r e s u l t s   w e r e   c o n s t a n t   w i t h   t h o s e   p r e s e n t e d   in   T a b l e   4,   j u s t i f y   it   t r a i n i n g   p o t e n t i a l l y   r e s u l t   in   p i p e l i n e s   p e r f o r m i n g   s i m i l a r l y   to   t h o s e   t r a i n e d   on   c o m p l e t e l y   c l e a n   d a t a s e t s   w h e n   u s e d   w i t h   n o i s y   d a t a .   P a r t i c u l a r   p i p e l i n e s   v o l a t i l i t y   a n d   p e r f o r m a n c e ,   h o w e v e r ,   m i g h t   d e c l i n e .   B e c a u s e   i n s t a n c e ,   t h e   t o p - p e r f o r m i n g   s y s t e m   in   T a b l e   4,   p r e s e n t e d   in   t h e   f i r s t   r o w ,   s a w   a   d r o p   of   2 . 3 4 %   ( 9 0 . 9 3 %   to   8 8 . 8 % )   on   t h e   n o i s y   s a l t   a n d   p e p p e r   t e s t   s e t   a n d   5 . 2 3 %   ( 9 3 . 7 %   to   8 8 . 8 % )   on   t h e   p u r e   t e s t   s e t   s e e   T a b l e   5;   r o w   3.   H o w e v e r ,   it   p e r f o r m e d   b e t t e r   ( 6 9 . 0 7 %   to   8 9 . 1 7 % )   on   t h e   h a z y   d a t a s e t .   T h a t   i m p l i e s   t h i s   t r a i n i n g   w i t h   e r r a t i c   s a m p l e s   m a y   r e s u l t   in   b e t t e r   p e r f o r m a n c e   on   c l e a n   d a t a s e t s   s e e   T a b l e   4;   r o w   4   a n d   T a b l e   5;   r o w   1,   but   a l s o   in   d e c r e a s e d   p e r f o r m a n c e   s e e   T a b l e   4;   ro w   1   a n d   T a b l e   5;   r o w   8.   N o t a b l y ,   we   o b s e r v e d   a   s i g n i f i c a n t   i m p r o v e m e n t   (P - v a l u e   f o r   W i l c o x o n   < 0 . 0 5 )   in   p e r f o r m a n c e   w h e n   e v a l u a t i n g   w i t h   b l u r r y   d a t a s e t s ,   w i t h   a l l   c a s e s   s h o w i n g   a   n o t a b l e   i n c r e a s e   in   p e r f o r m a n c e .   T ab le   4   s h o ws   th e   h i g h est   p er f o r m in g   au to m l   co n f ig u r ati o n s   f o r   t h e   ea r ly   c r o p   wee d s   d ataset,   p r esen ted   as   m ea n   an d   s tan d ar d   d ev iatio n   v alu es.   ( No te :   ps   r ef er s   to   p lan t   s eg m en tatio n   a n d   fc   r e f er s   to   f u lly - co n n ec ted ) .   T ab le  5   p r esen ts   th h ig h est  p er f o r m in g   au t o m co n f ig u r atio n s   with   th eir   m ea n   an d   s tan d ar d   d ev iatio n   f o r   th p lan t seed lin g s   d ataset.   T h ab b r ev iatio n s   u s ed   in   th tab le  ar p s   f o r   p la n t seg m en tatio n   an d   f f o r   f u lly - c o n n ec ted .       T ab le  4 .   T o p - p e r f o r m in g   A u to ML   co n f ig u r atio n s   f o r   th ea r ly   cr o p   wee d s   da taset    PS   FC   C l a s si f i e r   F 1   S c o r e   O v e r f i t t i n g   S a l t   F 1   B l u r   F 1   Y e s   No   En se mb l e   9 3 . 7 ± 1 . 1 3   6 . 2 8 ± 1 . 0 6   9 0 . 9 3 ± 5 . 5 6   6 9 . 0 7 ± 1 1 . 5 4   Y e s   No   S i n g l e   9 3 . 6 ± 1 . 6   6 . 4 ± 1 . 6   8 8 . 4 ± 6   7 0 . 8 ± 2   No   Y e s   En se mb l e   9 2 . 1 5 ± 2 . 4   7 . 8 2 ± 2 . 5 8   6 0 . 8 ± 8   4 9 . 2 ± 1 4 . 8   Y e s   No   S o f t ma x   9 1 . 9 ± 6 . 3 6   7 . 8 4 ± 6 . 6 7   8 7 . 2 ± 9 . 2 4   6 7 . 6 ± 9 . 8 1   Y e s   Y e s   En se mb l e   9 1 . 3 1 ± 2 . 9 4   8 . 5 8 ± 3 . 0 1   8 8 . 8 ± 3 . 4 2   6 2 . 6 3 ± 1 0 . 8       T ab le  5 .   T o p - p e r f o r m in g   A u to ML   co n f ig u r atio n s   f o r   th p lan t seed lin g s   d ataset   PS   FC   C l a s si f i e r   F 1   S c o r e   O v e r f i t t i n g   S a l t   F 1   B l u r   F 1   Y e s   Y e s   En se mb l e   9 3 . 8 ± 3 . 4 4   6 . 0 9 ± 3 . 4   9 3 ± 4 . 4 1   9 3 . 4 ± 4 . 2 5   Y e s   Y e s   S i n g l e   9 2 . 9 1 ± 4 . 7 1   6 . 9 3 ± 4 . 6 4   9 2 . 5 7 ± 4 . 8 7   9 2 . 1 1 ± 4 . 3 7   Y e s   Y e s   S o f t ma x   9 2 . 0 9 ± 4 . 7 3   7 . 4 6 ± 4 . 8 6   9 2 . 2 7 ± 4 . 8   9 0 . 8 4 ± 4 . 6 5   No   No   S i n g l e   9 1 . 6 ± 4 . 8 8   8 . 0 4 ± 4 . 9 6   8 8 ± 4 . 3 8   9 0 . 2 ± 5 . 7   Y e s   No   En se mb l e   8 8 . 8 ± 1 . 9 6   1 1 . 0 2 ± 2 . 1   8 8 . 8 ± 2 . 8 5   8 9 . 1 7 ± 3 . 2 2   Evaluation Warning : The document was created with Spire.PDF for Python.
                      I SS N :   2 7 2 2 - 3 2 2 1   C o m p u t Sci  I n f   T ec h n o l Vo l.  7 ,   No .   2 J u ly   20 26 1 67 - 1 78   174   4 . 2 .     Seedlin g s   of   pla nts   da t a s et   Usi n g   th e   o r ig in al  d ataset  f o r   tr ain in g ,   T ab le   6   s h o wca s es   th e   to p - p er f o r m i n g   Au to M L   p ip elin es   b ased   on   th e   F1   s co r e   m etr ic   f o r   th e   in f o r m atio n   s et   f o r   p lan t   s ee d lin g s .   T h e   s etu p s   with   th e   h ig h est   ef f icien cy   ar e   r ep r esen te d   by   th e   in itial   two   r o ws,   ac co r d in g   to   th e   Frie d m an   ex am ,   wh i ch   was   ca r r ied   out   with   a   lev el   of   tr u s t   of   0 . 0 1 .   T h ese   co n f ig u r atio n s   s h ar e   th e   u s e   of   v eg etativ e   s eg m en ts   an d   s teer   clea r   of   f u lly   in ter co n n ec ted   n etwo r k s   wh e n   co llectin g   f ea tu r es.   T h e   g r ea test   r esu lts   wer e   o b tain e d   wh en   So f tm a x   was   s u b s titu ted   with   a   n ew   p r ed ict o r   ( b o th   g r o u p   a n d   s in g le;   9 0 . 7 4 ± 0 . 8   an d   9 0 . 1 6 ± 0 . 6 7 ,   co r r e s p o n d in g l y ) ,   wh ich   is   co n s is ten t   with   th e   f in d in g s   in   T ab le   4.   T h e r ef o r e ,   th ese   co n f ig u r atio n s   can   be   co n s id er ed   as   a   s o lid   s tar tin g   p o in t   f o r   a d d itio n al   r esear ch   on   d atasets   th at   ar e   f r ee   f r o m   n o is e.   T h e   p lan t   s eg m en t   is   a   u s ef u l   ch o ice   a m o n g   th e   ex tr em e   p ar a m eter s   to   g et   th e   b est   o u tco m es.   Fu r th er m o r e,   in s tead   of   em p lo y in g   th e   f ea tu r e   ex tr ac to r   to   tr ain   ad d itio n al   class if ier s ,   S o f tm ax ,   to   b eh a v e   well   o v er a ll,   as   th is   s tr ateg y   was   u s ed   by   8   of   th e   to p   10   s y s tem s .   Ho wev er ,   wh e n   test ed   on   n o is y   in f o r m ati o n   s ets,   ev er y   s y s tem   s aw   a   s h ar p   d ec lin e   in   ef f icien cy   ( W ilco x o n   p - v alu e< 0 . 0 1 ) ,   s im ilar   to   wh at   h ap p en ed   with   th e   E ar lier   C r o p   W ee d s   d ata.   H o wev er ,   it s   cr u cial   to   r em em b er   th at   ce r tain   s tr u c tu r es   wer e   more   r esis tan t   to   p ar ticu lar   k in d s   of   n o is e   th an   o th er s .   In   g en er al,   t h e   v alu es   in   th e   Ov er f itti n g   b o x   ar e   g r ea ter   th a n   th o s e   f o u n d   in   T ab le   4,   p a r ticu lar ly   f o r   th e   f ir s t   p air   of   p ip elin es   ac r o s s   b o th   lis ts   ( W i lco x o n   p - v alu e< 0 . 0 1 )   [ 16 ] .   An   u n f o cu s ed   v er s io n   f o r   t h e   tr ain   d ataset ,   T ab le   7   s h o ws   wh eth er   Au to ML   p r o ce s s o r s   p er f o r m ed   b etter   or   wo r s e   wh e n   tau g h t   with   n o is y   d ata   ( i.e . ,   50%   clea n ,   25%   p ep p er   a n d   s alt,   an d   2 5 %   b lu r r in g ) .   Fo llo win g   th e   Frie d m an   ev al u atio n   at   a   lev el   of   tr u s t   of   0 . 0 5 ,   th e   p ip es   with   th e   b est   p er f o r m an ce   ar e   d is p lay ed   in   th e   t o p   4   r o ws   of   th e   tab le.   T h e   two   m o s t   o f ten   u s ed   ex tr em e   p ar am eter s   in   th ese   s tr u ctu r es   wer e   av o id in g   a   f u lly   co n n ec ted   lin k   an d   u s in g   cr o p   d iv is io n .   As   an ticip ated ,   th e   r esu lt   of   t h e   e v alu atio n   with   d ata   th at   was   n o is y   s h o wed   an   o v er all   im p r o v em en t.   No n eth e less ,   T ab le   5 s   Ov er f itti n g   co lu m n   r ev ea led   g en er ally   wo r s e   o u tco m es   ( W ilco x o n   p - v alu e< 0 . 1 ) .   T h is   im p lies   th at   th e   p ip elin es   m ig h t   s tr u g g le   with   ac cu r ate   ca teg o r is atio n .   B ased   o n   th an aly s is   o f   th r esu lts   p r esen ted   ea r lier ,   it  is   ev id en th at  ce r tain   h y p er p ar am eter s   h ad   g r ea ter   im p ac t o n   th last   p er f o r m an ce .   I n   Fig u r 3 ,   it is   co n s o lid atin g   th e   o u tco m es  f r o m   th two   s ets s u p p o r th id ea   th at  ap p ly in g   v eg etatio n   s eg m en tatio n   as a   p r elim in ar y   p r o ce s s in g   m eth o d   im p r o v ed   p er f o r m an ce as  s h o wn   in   Fig u r 3 ( a) .   T h b est  attr ib u tes  f r o m   th o r ig in al  p h o to s   wer ex tr ac ted   b y   th is   n eu r al  n etwo r k .   Fin d in g   co m p r eh en s iv m ac h in lear n in g - b ased   p ip elin ca p ab le  o f   p r o d u cin g   th o p tim al  o u tco m was  th s ec o n d   s tag af ter   th ch ar ac ter is tics   wer ex tr ac ted .   I n   co n tr ast,  th u s o f   f u lly - co n n ec ted   n etwo r k s   h ad   r elativ ely   lim ited   im p ac o n   th o v er all  p er f o r m an ce ,   as  illu s tr ated   in   Fig u r 3 ( b ) .   T h d is tr ib u tio n s   o f   th p er f o r m an ce   m etr ics  in d icate   th at  en ab lin g   o r   d is ab lin g   th f u lly - co n n ec ted   lay er   d id   n o lead   to   s tatis tically   s ig n if ican t d if f er en ce s .   I n   s im ilar   v ein ,   th ty p o f   class if ier   was im p o r tan t.  Fig u r 3 ( c)   s h o ws th at  wh ile  th Sin g le  an d   So f tm ax   class if ier s   m ay   also   p r o d u ce   g o o d   r esu lts ,   th E n s em b le  tech n iq u h ad   th h ig h est  m ed ian   p er f o r m an ce .   T h v ar ian ce   f o r   th Sin g le  class if ier   was  lo wer   th an   th at  o f   th So f tm ax   ap p r o ac h ,   s u g g esti n g   th at  it  was  m o r r eliab le  class if ier .   T h W ilco x o n   ex am   was  u s ed   to   co m p ar th v ar io u s   d is tr ib u tio n s ,   an d   ju s th s eg m en tatio n   o f   p lan ts   was  f o u n d   to   b s ig n if ic an tly   d if f er en f r o m   d if f er en o n (p - v alu e< 0 . 0 1 ) .   T h o u tco m es  s h o th at  class if ier s   ca n   b g en er ated   v ia  Au to ML   with   F1   s co r es  ab o v 9 0 %,  wich   is   co n s is ten with   o th er   r elate d   s tu d ies  [ 22 ] ,   [ 23 ] .   Ho wev er ,   o u r   p r ev io u s   wo r k   ac h iev ed   h ig h er   p er f o r m an ce   [ 16 ] ,   [ 17 ] ,   b u at  th co s o f   s ig n if ican tim an d   ef f o r f r o m   d ee p   lear n in g   ex p er ts   to   f in e - tu n th n etwo r k s .   Au to ML   o f f er s   s o lu tio n   to   s h o r ten   o r   av o id   th is   p r o ce s s .   Ou r   m ain   f o cu s   in   th is   s tu d y   was  to   p r o v id tr u s two r th y   ex p er im en tal  co n f ig u r atio n   f o r   ass ess in g   Au to ML   q u ality   in   d if f er en s ce n ar io s ,   as  o p p o s ed   to   au to m atica lly   d eter m in in g   th o p tim al  ML   p r o ce s s es u s in g   B ay es o p tim izatio n .     I t is wo r th   n o tin g   th at  t h er wer r eso u r ce   lim itatio n s   f o r   th lin es th is   s tu d y   ex am in ed   r ef er   to   T ab le  2 .   T h u s ,   g iv en   m o r tim o r   it  with   ex p er im en ts ,   th o u tco m es  m ig h p o ten tially   en h an ce   in   ter m s   o f   d u r ab ilit y   v er s u s   th o v er f itti n g   an d   F1   s co r e.   Po ten tial  to p ics  f o r   co n s id er atio n   will  in clu d s tr ik in g   th co r r ec b alan ce   b etwe en   Au to ML ,   m an u al  ex p er m ac h in lear n in g   tu n in g ,   an d   th b est p o s s ib le  p er f o r m an ce .   I n   r elatio n   to   th p r ev io u s   in q u ir y   q u esti o n ,   it  is   wo r th   d is cu s s in g   if   th p r ed ictiv m o d elin g   p r o ce s s es  co n s tr u cted   o n   b ase  o f   th n eu r al - b ased   ex tr ac tio n   o f   ch ar ac ter is tics   s h o wed   ce r tain   tr en d .   B a s ed   o n   th r esu lts ,   it  ap p ea r s   th at  th d if f er en ap p r o ac h es  wer tak en   b y   th B ay es  o p tim izatio n   m eth o d ,   lead in g   to   v ar io u s   ar r an g em en ts   f o r   p r ed icto r   tu n in g ,   d en s ity   r ed u ctio n ,   an d   ch o ice  o f   f ea tu r es.  T h s u g g ests   th at  ev en   m in o r   v ar iatio n s   with in   th d a taset  m ig h lead   to   s ig n if ican tly   d is tin ct  p ip elin es.  T h class if ier   m ig h b an   ar b itra r y   f o r est  o r   d ec is io n   tr ee ,   f o r   in s tan ce ,   with   n o   clea r   ad v an tag f o r   eith er   o n e.   T h is   p h en o m en o n   is   clo s ely   co n n ec ted   to   th th eo r em   o f   n o - f r ee   m ea l,  wh ich   is   esp ec i ally   r elev an in   th co n tex o f   Au to ML .   Fu r th er m o r e,   o v er f itti n g   h as  b ee n   r ed u ce d   with   th u s o f   tr ee   g r o u p s ,   th tab les  in d icate   th at  s o m o v er f itti n g   o cc u r r ed ,   wh ich   lim its   th s ce n ar io s   wh er th ese  s y s tem s   ca n   b s af ely   ap p lied .   I n   co n clu s io n ,   Au to ML   a d d s   an   ad d itio n al  lay er   o f   co m p lex ity ,   an d   ac h iev in g   b alan ce   b etwe en   in ter p r etab ilit y   an d   p er f o r m an ce   is   cr u cial  an d   d ep en d s   o n   th s p ec if ic  ap p licatio n   an d   it  is   ass o ciate d   r is k s .   I n   co n clu s io n ,   Au to ML   h as  th p o ten tial  to   ass is th ag r o tech n o lo g y   co m m u n ity   in   test in g   m ac h in e - lear n in g - b ased   Evaluation Warning : The document was created with Spire.PDF for Python.
C o m p u t Sci  I n f   T ec h n o l     I SS N:   2722 - 3 2 2 1       P erfo r ma n ce   ev a lu a tio n   o f th e   d ee p   lea r n in g   s ystem  fo r   w ee d   r ec o g n iz a tio n   ( A b d   A b r a h im   Mo s s la h a )   175   s o lu tio n s   with   r ed u ce d   im p lem en tatio n   r eso u r ce s ,   en ab lin g   m o r r eso u r ce s   to   b f o cu s ed   o n   th d o m ain - s p ec if ic  p ar o f   th p r o b lem .   Fu r th er m o r e,   th Au to ML   wo r k f lo d em o n s tr ated   with in   th is   s tu d y   p o s s ib ly   r ap id ly   g en e r ate  f r esh   m o d el  u s in g   f r esh   d ata,   f ac ilit atin g   th im p lem en tatio n   o f   ex ce llen tech n o lo g ies  in   r esp o n s to   th ev er - ch an g in g   n atu r o f   ag r icu ltu r [ 18 ] ,   [ 2 4 ]   as   s h o wn   in   Fig u r 3 .           ( a)   ( b )         ( c)       Fig u r 3 .   Dis p lay s   th s tatis tic al  an aly s is   r esu lts   o f   th r ee   ex p er im en tal  v ar iab les:   ( a)   t h u s o f   p lan s eg m en tatio n ,   ( b )   th u s o f   f u lly - co n n ec ted   n etwo r k s ,   an d   ( c)   th ty p o f   class if ier       T ab le   6   s h o ws   th e   m ea n ± s tan d ar d   d ev iatio n   of   th e   h ig h est   p er f o r m in g   b est   au to m l   co n f i g u r atio n s   in   th e   p lan t   s ee d lin g s   d ataset .   T ab le   7   s h o ws   th e   h ig h est   p er f o r m in g   b est   au to m l   co n f ig u r atio n s   in   th e   p lan t   s ee d lin g s   d ataset,   with   r esu lts   p r esen ted   as   th e   m ea n   v al u e   pl us   th e   s tan d ar d   d ev iatio n .   T ab le  8   s u m m ar izes  th d if f er e n ce s   b etwe en   th Au to ML   ap p r o ac h es  u s ed   in   th is   s tu d y   an d   r ec e n ad v a n ce m en ts   in   au to m ated   an d   d ee p   lear n in g   s y s tem s .       T ab le  6 .   Hig h est - p e r f o r m in g   Au to ML   co n f i g u r atio n s   f o r   th p lan t seed lin g s   d ataset  u n d e r   n o is y   im ag co n d itio n s   PS   FC   C l a s si f i e r   F 1   S c o r e   O v e r f i t t i n g   S a l t   F 1   B l u r   F 1   Y e s   Y e s   S o f t ma x   8 6 . 9 8 ± 1 . 9 2   1 2 . 2 6 ± 2 . 1 5   8 4 . 5 5 ± 1 . 7 9   8 7 . 3 4 ± 2 . 1 1   Y e s   No   En se mb l e   8 5 . 9 7 ± 4 . 2 4   1 3 . 8 8 ± 4 . 0 7   8 5 . 0 1 ± 3 . 9 1   8 5 . 8 7 ± 4 . 5 1   Y e s   No   S i n g l e   8 5 . 2 9 ± 5 . 2 9   1 4 . 4 ± 4 . 9   8 4 . 2 8 ± 4 . 5 1   8 4 . 8 1 ± 5 . 0 3   No   No   En se mb l e   8 3 . 7 8 ± 3 . 9   1 5 . 4 9 ± 4 . 4 2   8 1 . 7 6 ± 3 . 2 5   8 3 . 1 3 ± 3 . 8 6   No   No   S o f t ma x   8 0 . 4 5 ± 4 . 1 3   1 7 . 6 1 ± 6 . 9 4   7 8 . 7 2 ± 3 . 2 2   8 0 . 0 9 ± 4   N o t e :   ps   r e f e r s   to   p l a n t   s e g me n t a t i o n ,   w h i l e   fc   st a n d s   f o r   f u l l y - c o n n e c t e d   Evaluation Warning : The document was created with Spire.PDF for Python.
                      I SS N :   2 7 2 2 - 3 2 2 1   C o m p u t Sci  I n f   T ec h n o l Vo l.  7 ,   No .   2 J u ly   20 26 1 67 - 1 78   176   T ab le  7 .   Hig h est - p e r f o r m in g   Au to ML   co n f i g u r atio n s   f o r   th ea r ly   cr o p   wee d s   d ataset  u n d er   n o is y   im ag e   co n d itio n s   PS   FC   C l a s si f i e r   F 1   S c o r e   O v e r f i t t i n g   S a l t   F 1   B l u r   F 1   Y e s   No   En se mb l e   9 0 . 7 4 ± 0 . 8   8 . 5 1 ± 1 . 2 5   7 2 . 8 9 ± 7 . 5 2   7 2 . 0 6 ± 1 7 . 8 3   Y e s   No   S i n g l e   9 0 . 1 6 ± 0 . 6 7   9 . 3 ± 0 . 8 3   7 5 . 4 3 ± 1 . 2 2   8 0 . 9 4 ± 3 . 6 5   Y e s   Y e s   S i n g l e   8 8 . 6 4 ± 0 . 6 6   1 1 . 0 4 ± 0 . 4 2   8 0 . 9 6 ± 1 . 5 6   8 6 . 6 2 ± 1 . 0 9   No   No   En se mb l e   8 8 . 6 3 ± 1 . 2 6   8 . 1 6 ± 0 . 3   6 0 . 9 4 ± 1 7 . 3 3   8 3 . 5 7 ± 0 . 8 3   Y e s   No   S o f t ma x   8 7 . 1 7 ± 2 . 6 3   6 . 3 7 ± 1 . 0 2   6 7 . 7 4 ± 6 . 1 8   7 1 . 5 9 ± 1 7 . 2 4   N o t e :   ps   r e f e r s   to   p l a n t   s e g me n t a t i o n ,   w h i l e   fc   st a n d s   f o r   f u l l y - c o n n e c t e d .       T ab le  8 .   C o m p a r is o n   b etwe en   th Au to ML   ap p r o ac h es u s ed   in   th is   s tu d y   an d   m o d e r n   s tate - of - th e - ar tech n o lo g ies   C r i t e r i a   Te c h n o l o g i e s   u se d   i n   t h e   st u d y   M o d e r n   s t a t e - of - t h e - a r t   t e c h n o l o g i e ( 2 0 2 3 2 0 2 4 )   M o d e l   t y p e   A u t o M L   ( A u t o K e r a s,   A u t o - S k l e a r n )   A d v a n c e d   A u t o M L+ l a r g e   l a n g u a g e   m o d e l s   ( LL M s)   +   h y b r i d   d e e p   l e a r n i n g   S p e e d   a n d   p e r f o r ma n c e   G o o d   p e r f o r ma n c e   (F1 - sc o r e   90 9 3 % ) ,   b u t   r e l a t i v e l y   l o n g   t r a i n i n g   t i m e   F a st e r   t r a i n i n g   w i t h   d i s t r i b u t e d   t r a i n i n g   a n d   a d v a n c e d   n e u r a l   a r c h i t e c t u r e   se a r c h   (NAS)   C o s t   a n d   c o m p u t a t i o n   Lo c a l   c o mp u t a t i o n   (GPU   R TX   2 0 8 0 T i )   A d v a n c e d   c l o u d   c o m p u t i n g   ( e . g . ,   G o o g l e   TPU ,   A W S   I n f e r e n t i a )   w i t h   o p t i m i z e d   e n e r g y   c o n su m p t i o n   Ea se   of   u se   R e l a t i v e l y   e a s y   f o r   d e v e l o p e r s,   b u t   r e q u i r e s   se t u p   R e a d y - to - u s e   c l o u d   i n t e r f a c e s   ( e . g . ,   G o o g l e   V e r t e x   A I ,   A z u r e   A u t o M L)   w i t h   No - C o d e / L o w - C o d e   s u p p o r t   S c a l a b i l i t y   Li mi t e d   by   l o c a l   h a r d w a r e   a n d   d a t a   s i z e   H o r i z o n t a l l y   a n d   v e r t i c a l l y   sc a l a b l e   w i t h   b i g   d a t a   s u p p o r t   N o i se   h a n d l i n g   a b i l i t y   Te st e d   w i t h   n o i s y   d a t a   ( sal t   a n d   p e p p e r ,   b l u r )   N o i se - r e si s t a n t   mo d e l s   u si n g   c o n t r a s t i v e   l e a r n i n g   a n d   a d v a n c e d   d a t a   a u g me n t a t i o n   I n t e r p r e t a b i l i t y   Li mi t e d ,   e s p e c i a l l y   w i t h   e n s e mb l e   m o d e l s   A d v a n c e d   i n t e r p r e t a b i l i t y   t o o l s:   S H A P ,   LI M E,   e x p l a i n a b l e   AI   ( X A I )   C o mm u n i t y   su p p o r t   a n d   u p d a t e s   O p e n - so u r c e   c o mm u n i ty   ( A u t o K e r a s ,   S c i k i t - l e a r n )   La r g e   c o mm u n i t y   su p p o r t   a n d   c o n t i n u o u s   u p d a t e s   ( e . g . ,   H u g g i n g   F a c e ,   P y To r c h   Li g h t n i n g )   A d v a n c e d   a g r i c u l t u r a l   a p p l i c a t i o n s   W e e d   a n d   p l a n t   r e c o g n i t i o n   I n t e g r a t e d   s y st e ms:   S mart   r o b o t s,   d r o n e - b a se d   a e r i a l   sca n n i n g ,   smar t   i r r i g a t i o n   s y st e ms   C o n t i n u o u s   l e a r n i n g   su p p o r t   N o t   i n h e r e n t l y   s u p p o r t e d   A d a p t i v e   m o d e l s   w i t h   o n l i n e   l e a r n i n g   a n d   c o n t i n u a l   l e a r n i n g   c a p a b i l i t i e s       5.   CO NCLU SI O N   T h is   s tu d y   ev alu ated   a   m eth o d o lo g y   f o r   id e n tify in g   wee d s   by   in teg r atin g   Au to ML   tech n o lo g ies   an d   b en ch m ar k ed   it   ag ain s t   on   t wo   s ets   of   d ata   with   f o u r   a n d   th ir teen   class if icatio n s   of   wee d s ,   s ee d lin g s ,   an d   cr o p s ,   r esp ec tiv ely .   T h e   p r o p o s ed   m eth o d o lo g y   ac h iev ed   p r o m is in g   F1   s co r es   of   90%   to   9 3 %   o b tain ed   by   th e   s u g g ested   m eth o d s ,   ac c o r d in g   to   th e   i n f o r m atio n   at   h a n d   an d   th e   ex is ten ce   of   n o is e   o b s er v atio n s .   Fu t u r e   wo r k   will   ex p lo r e   m o r e   co s tl y   m eth o d s ,   lik e   g r o win g   t h e   b atch   s ize   wh ile   r ec eiv in g   in s tr u ctio n ,   an d   u s in g   f r esh   d atab ases ,   lik e   Dee p W ee d s ,   to   g ain   f u r th er   in s ig h ts   in to   th e   g en er aliza tio n   ab ilit y   of   Au to ML .   Ad d itio n ally ,   th e   s tu d y   will   ex ten d   ex p er im en ts   to   test   th e   r o b u s tn ess   of   th e   s y s tem   with   n o is y   s am p les,   in clu d in g   s m ea r in g   n o is e   r elat ed   to   v eh icle   m o v em en t   an d   e v alu atin g   on   test   s ets   with   m a n y   k i n d s   of   n o is e.   S in ce   u tili s in g   d ec is io n   tr ee   e n s em b les   don t   h av e   co m p lete ly   ev a d ed   o v er f itti n g ,   th e   s tu d y   will   ex p lo r e   n ew   m ac h in e   lear n in g   p ip elin es,   i n clu d in g   r aisi n g   th e   en s em b l e s   d ec is io n   tr ee   ( DT )   co u n t   or   u s in g   a   Su p er   L ea r n er   m o d el   s tack in g   m eth o d .   T h e   s tu d y   will   also   ev alu ate   class if ier s   s ep ar ately   an d   co n s tr ain   th e   B ay esian   ap p r o ac h   to   a   m o r e   lim ited   s u b s et   of   im p r o v e   th e   in ter p r etab ilit y   of   th e   o u tco m es.   T h ese   f in d in g s   s u g g est   th at   Au to ML   tech n o lo g y   can   ai d   th e   ag r o tech n o lo g y   co m m u n ity   by   p r o v id in g   h ig h - p er f o r m in g   s o lu tio n s   with   f ewe r   r eso u r c es   r eq u ir e d   f o r   im p lem en tati o n ,   f ac ilit atin g   th e   cr ea tio n   of   n ew   m o d els   f o r   d y n am ic   a g r icu ltu r al   e n v ir o n m en ts .       ACK NO WL E DG E M E NT   T h e   au th o r s   w o u l d   lik e   to   t h an k   th e   An b ar   of   Un iv er s it y   an d   Mu s tan s ir iy ah   Un iv er s ity   f o r   th e   s u p p o r t   in   t h e   p r e v en tio n   wo r k .       CO NF L I C T   O F   I N T E R E S T   ST A T E M E NT   T h e   au th o r ( s )   d ec lar e ( s )   th at   t h er e   is   no   c o n f lict   of   in ter est   r eg ar d in g   th e   p u b licatio n   of   th i s   p ap er .           Evaluation Warning : The document was created with Spire.PDF for Python.