Uhlalutyo Lomfanekiso: Yintoni na kwaye Uyisebenzisa njani

ngo kim Ukuhlaziywa: Juni 29, 2022

Uhlalutyo lomfanekiso sisixhobo esinamandla sokukhupha ulwazi kwimifanekiso.

Oku kungasetyenziselwa iinjongo ezahlukeneyo, ukusuka ekulandeleni utshintsho kwimo engqongileyo ukuya ekuqaphelisweni kobuso kuhlalutyo lwemifanekiso yonyango.

Kweli nqaku, siza kukunika intshayelelo emfutshane yohlalutyo lomfanekiso kunye nendlela ongayisebenzisa ngayo. Siza kugubungela inkcazo yohlalutyo lomfanekiso, iindlela ezahlukeneyo kunye nezixhobo ezikhoyo, kunye nokusetyenziswa okunokwenzeka kohlalutyo lomfanekiso.

Inkcazo yohlalutyo lomfanekiso

Uhlalutyo lomfanekiso sisixhobo esisetyenziselwa ukufumana ingqiqo yedijithali imifanekiso. Isetyenziswa ngokuqhelekileyo kwiinkalo ezahlukeneyo ezifana neyeza, uphando lwezenzululwazi, ukucupha, ukusetyenzwa kwemifanekiso, kunye nezinye ezininzi.

Ngokwenene, yinkqubo yokuqokelela idatha kumfanekiso kunye nokuhlalutya idatha ukufumana ulwazi malunga nomxholo womfanekiso. Uhlalutyo lomfanekiso lunceda ukunika ulwazi malunga nomgangatho okanye iimpawu zomfanekiso ezifana nobukhulu bawo, imilo, ubunzulu/ukukhanya, umbala ukwakheka okanye isicwangciso somgangatho wayo.

Kukho iindlela ezininzi ezisetyenziselwa uhlalutyo lwemifanekiso ezifana neendlela zokuguqula indawo ezifana nokutsalwa komfanekiso ongaphantsi kunye nokulungiswa kwebhloko; bonisa ubuchule bokukhupha obufumana amanqaku okanye imimandla enomdla usebenzisa i-edge detectors okanye i-algorithms yokulandela umda; ubuchule bokuhlela ezahlula izinto ngokweendidi; ubuchule bokwahlulahlula izinto ukusuka ngasemva; kunye neealgorithms zokuqaphela ipateni eziqaphela izinto zisebenzisa iindlela zokubala.

Ngokusebenzisa izixhobo zokuhlalutya umfanekiso kwizicelo ezinjengokubona into kunye nokuqondwa kobuso, inokuphucula amava omsebenzisi xa usebenzisa imidiya yedijithali. Kukwayenzeka ukuba kusetyenziswe uhlalutyo kuphando lwezenzululwazi ukuze kuhlolwe ulungelelwaniso phakathi kwemida echaziweyo ngaphakathi kwemifanekiso enye okanye iqela lemifanekiso. Okokugqibela, obu buchwephesha bukwasetyenziswa kakhulu ngoochwephesha bezonyango kuzo zonke iingcali ezibandakanya iradiology kunye ne-pathology ukufumanisa iimeko ezahlukeneyo ngokusekwe kwimifanekiso eskeniweyo ethathwe kwizigulana.

Iindidi zokuhlalutya umfanekiso

Uhlalutyo lwemifanekiso kukutsalwa kolwazi olunentsingiselo kwimifanekiso yedijithali. Ligama elibanzi kakhulu elibandakanya iindidi ezininzi ezahlukeneyo zemisebenzi enxulumene nomfanekiso, kubandakanywa ukuqondwa kwezinto kunye nokuchongwa, umlinganiselo weemilo, ukuqaphela ipateni, uhlalutyo lwendawo kunye nokuhlelwa.

Kukho iintlobo ezininzi zeendlela zokuhlalutya umfanekiso; ezinye zezona zixhaphakileyo ziquka:
-pixelUhlalutyo olusekelwe / lwe-Spatial - Olu hlobo lokuhlalutya lubandakanya ukubala ngesandla iipikseli zomntu okanye amaqoqo eepikseli ukulinganisa iindawo (ubukhulu, imilo) kunye nobuninzi (ukusabalalisa).
-Uhlalutyo lweNkcazo - Le ndlela isebenzisa iindlela zokubala ukuvavanya iimpawu ezinxulumene nemimandla ngaphakathi komfanekiso onjengokuchasa okanye ukukhanya.
-Ukufunda ngomatshini / ubukrelekrele boBuqhetseba - I-Artificial intelligence (AI) i-algorithms yenza imisebenzi enamandla yokucwangcisa imifanekiso ezenzekelayo njengokubona into okanye ukwahlula kwe-semantic usebenzisa i-convolutional neural networks (CNNs).
-Uhlahlelo lweTexture - Le ndlela yokulinganisa indlela izinto ezingaphakathi komfanekiso ezisasazwa ngayo ngokumalunga nomnye ngokwendlela yokuthungwa-ukuhluka kumanqanaba angwevu okanye imibala kwindawo enkulu.
-Uhlalutyo lwe-Histogram - Uhlalutyo lwe-Histogram lubandakanya ukulinganisa ukuhanjiswa kwamanani ngaphakathi komfanekiso ofana nokwahlukileyo, ukukhanya, i-hue kunye ne-saturation.
-Iindlela zokucoca -Iindlela zokuhluza zibandakanya ukusebenza kwezibalo kuluhlu olupheleleyo olucutshungulwayo oluguqula iindawo ezithile zoluhlu ngelixa ugcina ezinye. Izihluzi zihlala zisetyenziselwa ukuphucula imiphetho okanye ukususa ingxolo kumfanekiso ngokugudisa ezi ndawo zingafunwayo.
-IiNdlela zoTshintsho eziNne – Le ndlela isebenzisa iinguqu ezine-Fourier ezibolisa imifanekiso ibe ngamacandelo amaninzi ngokusekelwe kuluhlu lwamaza kunye/okanye ulungelelwaniso olukhoyo kwisakhelo songeniso. Oku kuvumela ukwahlula iinxalenye kunye nezinto ezinokuthi zibe neempawu ezithile ezinxulumene namanye amacandelo ukubonakala ngokubanzi kwisakhelo - njengokuthungwa okanye ukungafani kobume - okwenza kube lula ukufunyanwa kwezinto ngexesha leenkqubo zokwahlula.

Izibonelelo zokuhlalutya umfanekiso

Uhlalutyo lomfanekiso sisixhobo esixabisekileyo sokufumana ulwazi kwimifanekiso yedijithali. Isetyenziselwa ukuqaphela iipateni, ukuchonga izinto, kunye nokukhupha ulwazi olunentsingiselo kumajelo eendaba. Ukususela kuphando lwezonyango ukuya ekubonweni kobuso, uhlalutyo lwemifanekiso lunokuzuza uluhlu olubanzi lwamashishini. Apha siza kuphonononga iingenelo zokuhlalutya umfanekiso kunye nendlela enokusetyenziswa ngayo ukuphucula iinkalo ezahlukeneyo.

Ukuphucula amava abathengi

Uhlalutyo lwemifanekiso lunokunceda ukuphucula amava abathengi, ukuvumela iinkampani ukuba ziqonde izinto ezikhethwa ngabathengi ngokukhawuleza nangokuchanekileyo. Ngokusebenzisa iindlela zokuhlalutya umfanekiso, amashishini anokusebenzisa idatha evela kwimifanekiso efana neendawo, iindawo, kunye neempendulo zeemvakalelo ukuze zijolise ngcono kwiintengiso zabo okanye umxholo wobuqu kubathengi. Ukusebenzisa le datha kunokukhokelela ekuphuculweni kothethathethwano nabathengi kunye nokwanda kweentengiso zenkampani.

Uhlalutyo lwemifanekiso lukwanceda amashishini aqonde ngcono umxholo wemifanekiso yawo. Umzekelo, itekhnoloji yokuqaphela imibala ivumela iinkampani ukuba zichonge imibala emfanekisweni kwaye iqinisekise ukuba inxulunyaniswa njani neemveliso okanye iinkonzo. Olu lwazi lunokunceda amashishini afumane ukuqonda kwindlela yokuziphatha yabathengi malunga nenkonzo ethile okanye imveliso. Ukongeza, amashishini anokusebenzisa itekhnoloji yokuqaphela into ukuchonga izinto ezikumfanekiso kwaye aqonde ngcono ukhetho lwabathengi ngokusekelwe kwizinto abazithandayo.

Ekugqibeleni, uhlalutyo lwemifanekiso lwenza ukuba amashishini aqokelele idatha ngokukhawuleza ngelixa egcina ukuchaneka kwinkqubo. Iinkampani azisafuni ukuphonononga ngamakhulu emifanekiso ngesandla xa zizama ukumisela izinto ezikhethwa ngabathengi; endaweni yoko bayakwazi ukuzenzela le nkqubo besebenzisa iindlela zokuhlalutya umfanekiso eziphucula ukusebenza kakuhle kakhulu ngelixa befumana iziphumo ezichanekileyo kuhlalutyo lwayo lwemifanekiso. Oku kuthetha ukuba abathengi bafumana ukufikelela kwiintengiso ezenzelwe wena ngokukhawuleza kunanini na ngaphambili, okwenza iinkampani zibonelele ngamava angcono omsebenzisi kubathengi bazo.

Ukonyuka kokuchaneka kunye nokusebenza kakuhle

Uhlalutyo lwemifanekiso lubonelela ngeenzuzo ezininzi kumashishini, kubandakanya ukuchaneka okwandisiweyo kunye nokusebenza kakuhle. Ngokuhlalutya imifanekiso endaweni yokuxhomekeka kwiindlela ezisetyenziswayo okanye ezisekelwe kwisicatshulwa, ukuhlalutya kwemifanekiso kunokunciphisa kakhulu iimpazamo zokungena kwedatha kunye nokukhawulezisa ixesha lokucubungula umsebenzi onikiweyo.

Uhlalutyo lwemifanekiso lukwasusa isidingo sokuhluza ubuninzi bolwazi ngesandla. Kunoko, i-algorithm inokukhawuleza ifumane idatha efanelekileyo kwaye ichonge iipateni ukuze zitolikwe ngokulula ngabantu. Oku kukhawulezisa inkqubo yokwenziwa kwezigqibo kwaye kunciphisa iimpazamo ezinokubiza iindleko ezinokwenzeka ngenxa yempazamo okanye ukongamela komntu.

Ukongeza, uhlalutyo lwemifanekiso lwenza kube lula ukuchonga iipateni ngokukhawuleza kwiiseti zedatha ezinkulu ngokumangalisayo ezinokuthatha amawaka eeyure zendoda ukuhluza ngesandla. Oku kunciphisa kakhulu iindleko zabasebenzi ngelixa kwandisa ukuchaneka nokusebenza kakuhle ngaxeshanye. Uhlalutyo lwemifanekiso lukwavumela imibutho ukuba ilinganise iintsingiselo ezinokuthi ziphoswe ngeendlela zemveli, zibanceda benze izigqibo ezingcono ngokukhawuleza kunangaphambili.
Ngokusebenzisa iteknoloji yokuhlalutya umfanekiso, amashishini anokunciphisa iimpazamo zokungena kwedatha kunye nokunciphisa iindleko zabasebenzi ngelixa ukwandisa ukuchaneka nokusebenza kakuhle kwimisebenzi yabo. Oku kukhokelela ekuphuculweni kwesakhono sokwenza izigqibo kunye neziphumo ezichaneke ngakumbi ezivela kwiiseti zedatha ezinkulu kakhulu kunokuba zinokufezekiswa ngokusebenzisa iindlela zemveli zodwa.

Ukuphuculwa kokwenziwa kwezigqibo

Ukusetyenziswa kohlalutyo lwemifanekiso kumashishini ukuncedisa ekwenzeni izigqibo kunokukhokelela ekuphuculeni ukusebenza kakuhle kunye nokunceda imibutho ilawule ngcono izixhobo zabo. Ngokusebenzisa iindlela ezahlukeneyo zobugcisa kunye nobuchwepheshe obubandakanyekayo, umbutho unamandla okwenza ngokukhawuleza kunye nokuhlalutya imifanekiso evela kwimithombo eyahlukeneyo. Oku kubanceda benze izigqibo ezivakalayo ngephanyazo.

Uhlalutyo lwemifanekiso lunokunceda imibutho ihlele, ibone, kwaye ivavanye idatha ukuze ibeke phambili imisebenzi okanye iinjongo zobuchule. Ikwavumela uhlalutyo olukhawulezayo lweepateni ezintsonkothileyo ngaphakathi kwemifanekiso, egcina ixesha elixabisekileyo kwimisebenzi yokuthatha izigqibo ezikwinqanaba eliphezulu. Ukongeza, uhlalutyo lwemifanekiso lunokunceda iinkampani zifumanise iindlela okanye imiba eqhubekayo enokuthi ingakhange iqatshelwe.

Imibutho ikhulisa amandla obuchwepheshe bokuhlalutya imifanekiso efana nobukrelekrele bokwenziwa (AI) kunye nokufunda koomatshini (ML) izicelo zokudala iimodeli ezixela kwangaphambili ezibanika ingqiqo malunga noko kuya kwenzeka malunga nokuziphatha kwabathengi kunye neemeko zentengiso kwixesha elizayo. Ngeli xesha, amaqela ohlalutyo lweemarike ahlala ehlola ngemifanekiso abayiqokelele kwimithombo efana newebhusayithi ekhuphisanayo kunye namajelo edijithali. Ngezixhobo ezifanelekileyo, la maqela anokukhawuleza agqibe ukuba abathengi banxibelelana njani neemveliso okanye iinkonzo zombutho ukuze bakhe iimbono ezibambekayo ezibonelela ngomda wokukhuphisana kwabanye kushishino.

Uhlalutyo lwemifanekiso ludlala indima ebalulekileyo kwimisebenzi yezoshishino namhlanje ngokubonelela ngemibutho ngokufikelela kwiinkcukacha ezibalulekileyo eziqhutywa yidatha ezikhuthaza ukukhula kunye nenkqubela phambili ngokukhawuleza kunanini ngaphambili.

Indlela yokusebenzisa uHlahlelo lweMifanekiso

Uhlalutyo lwemifanekiso yindlela esebenzayo yokucubungula inani elikhulu ledatha yesithombe kunye nokukhupha ulwazi oluluncedo kuyo. Uhlalutyo lwemifanekiso lubandakanya ukusetyenziswa kwee-algorithms zekhompyuter ukumisela intsingiselo yomfanekiso. Ingasetyenziselwa imisebenzi eyahlukeneyo efana nokuqondwa kwento, ukuqondwa kobuso, kunye nokwahlulwa kwemifanekiso. Kweli candelo, siza kuxubusha indlela uhlalutyo lomfanekiso olunokusetyenziswa ngayo kunye nezinye zezicelo zayo.

Ukwahlulwa kwemifanekiso

Ukwahlula komfanekiso lisebe lokwenziwa komfanekiso apho umfanekiso wahlulwe ube ngamacandelo amaninzi, okanye iingingqi, ezithi kunye zenze umfanekiso uphela. Icandelo ngalinye limele indawo enomdla ngaphakathi komfanekiso, kwaye inokucingelwa njengento ecacileyo yodwa. Injongo yokwahlula kukwenza lula kunye/okanye ukuguqula ukumelwa komfanekiso ube yinto enentsingiselo nelula ukuyihlalutya. Iindlela zokwahlulahlula zisetyenziswa kwimifanekiso yezonyango kwimisebenzi eyahlukeneyo, ukusuka ekuchongeni izakhiwo ze-anatomical ukujonga ukuqhubela phambili kwesifo. Ukongeza, i-algorithms yokwahlula ngamacandelo aphambili kwiinkqubo zokucupha ngokuzenzekelayo kwaye ingasetyenziselwa imisebenzi yokukhangela i-robotic kunye nokufumanisa into.

I-algorithms yokwahlula ngokubanzi ixhomekeke kwiindlela ezisekelwe kwi-pixel okanye iindlela ezisekelwe kuhlelo. Iindlela ezisekwe kwiPixel zisebenzisa izalathiso ezifana nombala okanye ukuthungwa ukwahlula umfanekiso ube ngamacandelo ahlukeneyo. Kungenjalo, iindlela ezisekelwe kuhlelo zidibanisa iindidi ezisekelwe kwiimpawu ezichazayo ezifana nokumila okanye ukuthungwa ukuze kuqokelelwe ngokuzenzekelayo iipikseli zibe ngamacandelo/izinto ezifanelekileyo.

Ulwahlulo olusekwe kwiPixel lujolise ekwahluleni izinto ngokokuqina kwazo okanye amaxabiso ochasaniso ngokubala izibalo zehlabathi (umzekelo, ixabiso eliqhelekileyo) kwindawo yonke. Iindlela ezisekelwe kuhlelo zihlala zifuna igalelo lezandla kunye nokuleyibheli ngaphambi kokuba i-algorithm iqalise ukusebenza; nangona kunjalo ezi ndlela zinamandla kakhulu xa ujongana neeseti zedatha ezinkulu okanye iintlobo zedatha ezingaziwayo ezingenako ukwahlulwa ngempumelelo kusetyenziswa amaxabiso epixel kuphela.

Ukuhlelwa komfanekiso

Ukuhlelwa komfanekiso yinkqubo yokusebenzisa i-algorithms yekhompyutha ukuqikelela kunye neeklasi zeleyibheli kumfanekiso onikiweyo. Kule nkqubo, kuqhelekile ukuba ube "nengqokelela" yemifanekiso yoqeqesho ebhaliweyo esebenza njengegalelo kwi-algorithm yokufunda. Ngokusekelwe kule mifanekiso yoqeqesho, iimpawu ezifana nobukhulu, imilo kunye nombala zicatshulwa kumfanekiso kwaye zisetyenziselwa ukuhlelwa kwemifanekiso. Obu buchule buvumela iziphumo ezichaneke ngakumbi xa kuthelekiswa nokuleyibheli ngesandla, apho abantu banokwenza iimpazamo ngenxa yokungahlelwa kakuhle okanye iimpazamo zokuleyibheli.

Kunikwe umfanekiso ongaziwayo, umntu unokusebenzisa umdidi okhoyo othi ke anike eli galelo likhethekileyo udidi lwemveliso ngokwenani elichazwe kwangaphambili leeklasi ezahlukileyo. Eminye imizekelo yokusetyenziswa kolu luhlu ukusuka kwiisistim zokuqhuba imoto ezizimeleyo zinika amandla okuqaphela into ukuya ekubhaqweni kobuso kwizicelo zebhayometriki. Ukongeza, ukuba nedatha enelebhile echanekileyo ngokuhlelwa kwemifanekiso kunokuboleka ukuqonda ngakumbi kokusingqongileyo ngokuxhasa iiseti zedatha zikarhulumente ezinkulu ukuze zihlalutywe ngakumbi - ngakumbi kunxibelelwano phakathi kweentlobo zezinto kunye nezicelo zeshishini ezinjengohlalutyo lokuthengisa okanye uhlalutyo lokuzonwabisa.

Ukusuka ekuphumezeni uthungelwano lwe-neural kunye neendlela zokuphucula i-GPU ukuze kufumaneke ukusetyenzwa ngokuhambelanayo kude kube ziindlela ezahlukeneyo zokufunda koomatshini - kukho iindlela ezingenakubalwa umntu anokuqeqesha iimodeli ze-AI okanye asebenzise ubuchule bokufunda obunzulu kwinkalo yoHlelo loMfanekiso; nangona ukufunda okungajongwanga kusenemiqobo ethile egcina abaphandi ukuba bayisebenzise ngokupheleleyo ngeyona ndlela iphezulu, ezi ndlela zisaphononongwa kwaye zisaphandwa. Ke ngoko ukuqonda indlela yokuyiphumeza kufuna ukuqonda okubanzi kwiialgorithms zombono wekhompyuter kunye nolwazi olomeleleyo lobuchwephesha olubandakanya izenzo zobunjineli zasemva ezinje ngolwimi lokubhala olunje ngePython ngeenjongo zokuphunyezwa kunye nokuba nemvelaphi esondeleyo malunga nobuchwepheshe obuvelayo obufana ne-blockchain okanye i-server-less infrastructure. ukuphunyezwa

Ukufunyanwa kwento

Ukufunyanwa kwento bubuchwepheshe obusekwe kwikhompyuter kunye nobuchule bokuqinisekisa ukuchaneka kuhlalutyo lwedatha yemifanekiso. Le teknoloji isebenzisa ii-algorithms ezahlukeneyo eziyinkimbinkimbi ukuchonga izinto kunye neempawu zazo kwimifanekiso echongiweyo ngokulula. Idlala indima enkulu kwiindawo ezininzi, ezifana ne-forensics, isayensi yezonyango, i-automation yemizi-mveliso, ukuchonga ubuso kunye neenkqubo zokuhlola ezizenzekelayo.

Ukufunyanwa kwento kubandakanya ukuhlalutya okanye ukuqokelela idatha ukumisela ubungakanani, imilo kunye nezinye iimpawu zezinto ezithile kumfanekiso. Olu hlobo lohlalutyo lwemifanekiso lunokubandakanya ukuchonga izinto ezisekelwe kumbala wazo, ukuthungwa okanye iimilo ezizenzayo xa zidibene. Ukusebenzisa iimodeli zemathematika kunye neendlela zobalo kwenza umsebenzi ube lula kakhulu kwiikhomputha kule mihla.

Eyona njongo iphambili apha kukuchonga ngokuchanekileyo into nganye evela kwiseti yedatha yomfanekiso ngokusebenzisa iilebhile zophawu oluyabelweyo. I-algorithm idinga ukuqeqeshwa kwangaphambili ngedatha eninzi, kuquka imizekelo 'elungileyo' yoqikelelo oluchanekileyo malunga neempawu zento ekujoliswe kuyo phambi kokuba uqeqesho luqale. Emva kokuba uqeqesho lugqityiwe kwaye uqikelelo luyimpumelelo, luya kulandelwa zizigaba ezinjengokuqikelelwa apho iziphumo ezilindelekileyo ezivela kwigalelo elinikiweyo zisebenza kunye neziphumo ezifunyenwe kwizifundo zangaphambili.

Itekhnoloji efana nokufunda nzulu (DL) algorithms inceda ngokuvumela iileya ezininzi ezithi kunye zisebenze ukufikelela kwiziphumo ezichanekileyo ngexesha elifutshane kuneendlela zemveli ezinokubonelela ngaphambi kokuba zifumaneke ngaphakathi kolu hlobo lokuseta izixhobo ezisetyenziswa namhlanje. Ekugqibeleni oku kwenza iindlela ezincedwa yi-AI ezinje ngokuFunda okuNzulu zithandeke ngakumbi kuneendlela eziqhelekileyo njengoko zinciphisa kakhulu iimpazamo ngelixa zibonelela ngeemetriki ezichanekileyo zokubona into enokuvelisa imveliso ngokulula kwiimeko zexesha lokwenyani ngokunjalo.

Izixhobo zokuhlalutya umfanekiso

Uhlalutyo lomfanekiso yinkqubo ebandakanya ukukhupha ulwazi olunentsingiselo kwimifanekiso ngeendlela zayo zokubala. Le nkqubo isetyenziswa kwiinkalo ngeenkalo, ezifana ne-medical imaging, intelligence artificial, kunye ne-remote sensing. Ngethamsanqa, kukho izixhobo ezahlukeneyo ezikhoyo zokuhlalutya umfanekiso, okwenza inkqubo ibe lula kwaye isebenze ngakumbi. Makhe sijonge ezinye zezixhobo ezikhoyo zokuhlalutya umfanekiso.

I-OpenCV

I-OpenCV lithala leencwadi lemisebenzi yenkqubo ejolise ikakhulu kwimbono yekhompyuter yexesha lokwenyani. Ixhasa iifestile ezahlukeneyo, iLinux, iqonga kunye namanye amaqonga esoftware kwaye ke isoloko isetyenziselwa uhlalutyo lomfanekiso kunye nemisebenzi yokwenziwa komfanekiso. Imimandla yesicelo se-OpenCV ibandakanya: umbono wekhompyutha, ukuqonda ubuso, ukuchongwa kwezinto, ukulandelela intshukumo, ukwahlula kunye nokuqondwa.

I-OpenCV inikezela ngoluhlu olubanzi lwezixhobo ezinamandla ezinokusetyenziswa ukwenza imisebenzi entsonkothileyo njengokubhaqa into, ukuqondwa kobuso, ukutsalwa kwesici ngokuzenzekelayo kunye nokunye. Inoluntu olusebenzayo olubonelela ngezisombululo ezisebenziseka lula kwiingxaki eziqhelekileyo. Ithala leencwadi lenzelwe ukudibanisa neelwimi ezininzi zokucwangcisa ezifana neC ++, iJava okanye iPython. Ithala leencwadi lenza abaphuhlisi basebenze ngemifanekiso kwiilwimi zabo abayifunayo ngaloo ndlela benciphisa ixesha lophuhliso kakhulu. I-OpenCV ixhasa naluphi na uhlobo lwedatha (efana nemifanekiso kwi-2D kunye nokwakheka kwe-3D) eyenza ibe sesona sixhobo siguquguqukayo phakathi kwazo zonke iilayibrari zokucwangcisa imifanekiso ezikhoyo namhlanje.

Izixhobo ezibandakanyiweyo kwi-OpenCV zezi:
-Ukuhluzwa komfanekiso
-Ukwahlulwa kwento
-Utsalo lweempawu
-Uthelekiso nolwahlulo
-Ukulandelwa kwento kunye nohlalutyo lwentshukumo
-Ukwakhiwa ngokutsha kwe-3D kwizinto ezifundwayo

Izixhobo zayo zisetyenziswa ngokubanzi ukuhlola okubonakalayo okuzenzekelayo kwiisethingi zemizi-mveliso ngenxa yeempawu zayo eziluncedo ezifana nokusetyenziswa kwememori ephantsi, ukuphunyezwa ngokukhawuleza kumaqonga amaninzi (ngokukodwa afakwe ngaphakathi), ukuphatheka phakathi kweenkqubo ezisebenzisa iilayibrari eziqhelekileyo ezifana ne-OpenMPI/MPI4Py. Ilayisenisi yayo yomthombo ovulekileyo iyenza ibe nomtsalane ngakumbi kubasebenzisi abafuna ukuyisebenzisa ngaphandle kweentlawulo zobukhosi okanye iilayisensi ezivela kwiinkampani okanye imibutho efana nezibalo njl.

TensorFlow

I-TensorFlow lithala leencwadi lesoftware elivulelekileyo elisetyenziselwa ukuphuhlisa nokuqeqesha imifuziselo yokufunda enzulu. Iinkqubo zokufunda ezinzulu, ezinje ngeemodeli zokuhlalutya imifanekiso, zakhiwe ngamaleko amaninzi ee-neuron ezenza idatha yegalelo, njengomfanekiso. I-TensorFlow yenza ukuba abaphuhlisi baphuhlise ngokulula kwaye basebenzise iimodeli zokufunda ezinzulu ezinokuthi zisetyenziswe kumbono, ukusetyenzwa kolwimi lwendalo kunye nezinye iindawo.

Inzuzo enkulu yokusebenzisa i-TensorFlow kukuba inika indlela elula yokudala kunye nokusebenzisa iimodeli zokufunda koomatshini abanamandla ngokukhawuleza nangempumelelo. I-TensorFlow ikwaxhasa uluhlu olubanzi lwezicelo kuyo yonke iMacOS, iWindows, iLinux kunye namaqonga aphathwayo. Kwaye ngenxa yokuba ingumthombo ovulekileyo kwaye inenkxaso yoluntu etyebileyo, ungasebenzisa izixhobo ezikhoyo kwinkqubo yakho yophuhliso ngaphandle kokubuyisela amavili okanye uchithe ixesha elongezelelweyo kwikhowudi yokulungisa ingxaki ukusuka ekuqaleni.

Ukongeza ekuqeqesheni iimodeli ezintsha ukusuka ekuqaleni, iTensorFlow ikuvumela ukuba uphinde usebenzise iimodeli eziqeqeshwe kwangaphambili zokudlulisela ukufunda okanye ukulungiswa kakuhle kwiiseti zengxaki ezithile. Oku kwenza kube lula ukwakha izisombululo zesiko ezilungiselelwe iingxaki ezithile ngelixa uthatha ithuba lolwazi olukhoyo olwakhiwe kuthungelwano olukhoyo. Ngolu bhetyebhetye lusesandleni, abaphuhlisi banokuqala ukuzama iindlela ezahlukeneyo zokwakha ngokukhawuleza kunanini na ngaphambili - ukubenza baphakame kwaye baqhube ngeziphumo ezichaneke kakhulu ngexesha lonke.

Google Cloud Vision

Umbono weLifu likaGoogle sisixhobo esinamandla sokwenza imifanekiso eyilelwe ukunceda amashishini ahlalutye imifanekiso kwaye afumane ulwazi oluxabisekileyo. Isekwe kwitekhnoloji efanayo esetyenziswa ziiFoto zikaGoogle, iyenza ithembekile kwaye kulula ukuyisebenzisa. Esi sixhobo esisekwe kwilifu sivumela amashishini ukuba asebenze inani elikhulu lemifanekiso ngokukhawuleza, ukubona ubuso, ukuchonga izinto, kunye nokukhupha itekisi yonke ngelixa uphepha umsebenzi wezandla.

Enye into ebalulekileyo ye-Google Cloud Vision kukukwazi ukuchonga izinto ezivela kuluhlu olubanzi lweendidi ezifana nezilwanyana okanye iimveliso. Ngenqaku elithi "ukufumanisa uphawu", abasebenzisi banokunxibelelana neempawu ezifana ne-Facebook okanye i-Instagram ukuze bafumane iinkcukacha zokulandela umkhondo kunye nohlalutyo. Inqaku elithi "uphawu lophawu" lunceda abasebenzisi babone iimpawu zomhlaba ezibangqongileyo kwaye baqonde ngcono inkcubeko yabo yasekhaya ngephanyazo.

Ngaphandle kwezakhono zokuqaphela into, iGoogle Cloud Vision ikwabonelela abasebenzisi ngokuhlalutya kweemvakalelo kunye nesixhobo esizenzekelayo se-captioning yemifanekiso - ilungele uhlalutyo lwangoko! Ukongeza, esi sixhobo sinikezela ngezakhono zokuqondwa kobuso ezilungele iinjongo zokuthengisa kuba zibona iimpawu ezinje ngobudala kunye nesini ngokuchaneka okuphezulu. Ekugqibeleni, amandla okukhutshwa kwegama elingundoqo avumela abasebenzisi ukuba bahlolisise ngokukhawuleza imifanekiso yokufaka ngelixa befumana amagama abalulekileyo anokuthi asetyenziswe kwiinkqubo zamva okanye izicelo.

Ukuqukumbela, Umbono weLifu likaGoogle lukhetho olukhulu xa ufuna isixhobo sokwenza umfanekiso ukunceda amashishini ahlalutye imifanekiso yawo ngokufanelekileyo kunangaphambili!

isiphelo

Uhlalutyo lwemifanekiso sisixhobo esinamandla sokuqokelela ulwazi olusuka kwimifanekiso yedijithali. Ingasetyenziswa kwiinkalo ezininzi ezahlukeneyo ezifana neyeza, iimoto, kunye nokufota. Iingenelo zokusebenzisa uhlalutyo lwemifanekiso zininzi kwaye zinokugcina ixesha, imali kunye nomgudu. Kule nqaku, siye sajonga iziseko zokuhlalutya umfanekiso, ukuba yintoni, kunye nendlela enokukunceda ngayo. Siye saxubusha kwakhona ngeentlobo ezahlukeneyo zokuhlalutya kunye nezenzo ezilungileyo zohlalutyo oluyimpumelelo lwemifanekiso. Ekugqibeleni, kuyacaca ukuba uhlalutyo lwemifanekiso sisixhobo esinamandla sokuqokelela ulwazi kunye nokwenza izigqibo ezisekelwe kwidatha enolwazi.

isishwankathelo

Ukukhetha iirosti zekofu yakho kuxhomekeke kwincasa kunye namandla owakhethayo kunye nezinto ozikhethayo zengingqi. Ii-roasters ezahlukeneyo zibonelela ngoluhlu olubanzi lwamagama kunye neendidi zemibala, kodwa ngokubanzi, iirosti ezikhanyayo ziya kuba nencasa ethambileyo kunye nomxholo wecaffeine ophakamileyo ngelixa iirosti ezimnyama zikrakra kakhulu zineasidi encinci. Iindidi ezidumileyo zesiXeko ngasinye zibandakanya isiXeko esiKhanyayo, isiXeko esisisiqingatha, iCinnamon, isiXeko, iMelika, isidlo sakusasa, isiXeko esiPheleleyo, iNew Orleans yaseYurophu Espresso yaseViennese yesi-Italian French. Nokuba ungumntu othanda iimbotyi zekofu ezikhanyayo okanye ezimnyama-okanye kwenye indawo phakathi-qiniseka ukuba ubuze ngaphambi kokuba uthenge ukuze ufumane kanye le nto uyikhangelayo!

U kufunda ngakumbi

Ukusetyenziswa kokuhlalutya komfanekiso ukufumana ingqiqo kunye nokwenza izigqibo kwimifanekiso kuya kuthandwa kakhulu kwiindawo ezininzi zamashishini, ukusuka kukhathalelo lwempilo ukuya kukhuseleko. Ukuba ungathanda ukufunda ngakumbi malunga nobu buchwepheshe, qwalasela ezi zixhobo zilandelayo:

Amanqaku kunye tutorials
-"Uhlalutyo lomfanekiso: Yintoni na kunye nendlela yokuyisebenzisa" ngu-Oracle
- "Intshayelelo yoHlalutyo lweMifanekiso kubantu abangazi nto malunga noHlalutyo lweMifanekiso" nguKent Woodard
-"Amanyathelo e-8 ekuqondeni uHlalutyo lweMifanekiso kunye nokuCwangciswa kokuFunda koMatshini" nguVictor Charpenay
Iincwadi kunye noPapasho
-Ukwenziwa koMfanekiso: Imigaqo kunye nezicelo nguMilan Sonka et al.
-Umbono weKhompyutha kunye noomatshini-Ithiyori, ii-algorithms, ii-Practicalities (4th Ed.) ngu-ER Davies
-Ukwenziwa kweMifanekiso yeDijithali nge-OpenCV-iBibliography (Ver 4.1) Ihlanganiswe nguDavid Dardas Webinars kunye neePodcasts
-Umbono weKhompyutha Ngokufunda ngoMatshini weWebinar ephethwe nguTreehouse AI
-Ii-Machine Learning Fundamentals I-Podcast ebanjwe yi-Google Cloud Platform Kukho nezinye iincwadi ezininzi, amanqaku, ii-webinars, ii-podcasts, iindibano zocweyo kunye nezinye izixhobo ezikhoyo ezinokukunceda uqonde ngcono indlela yokuhlalutya umfanekiso kunye nendlela yokuyisebenzisa kwizicelo ezithile.

Molo, ndinguKim, umama kunye nomthandi we-stop-motion onemvelaphi kwindalo yemidiya kunye nophuhliso lwewebhu. Ndinothando olukhulu lokuzoba kunye noopopayi, kwaye ngoku ndintywila kuqala kwihlabathi le-stop-motion. Ngebhlog yam, ndabelana nawe ngokufunda kwam.