Kwihlabathi lophuhliso lwe-Android, ukusebenzisa ngokupheleleyo izixhobo zanamhlanje kwenza umahluko phakathi kwe-app esebenzayo kunye neyona nto iphambili. Elinye lawona malungu aguquguqukayo sisikena esisekwe kwi-ML Kit, ekwaziyo ukuguqula indlela iapp enxibelelana ngayo nendawo yayo ngekhamera. Ukusuka kwiibhakhowudi zokufunda kunye neekhowudi ze-QR ukuya ekubhaqeni izinto okanye ukuqonda isicatshulwa kunye nobuso, izinto ezinokwenzeka azipheli ngendlela efanelekileyo.
Abaphuhlisi abaninzi banamathela kwiziseko, besilela ukusebenzisa ngokupheleleyo amandla eML Kit kwi-Android. Ukuyidibanisa ngokuchanekileyo akukhona nje ukubandakanya ukuxhomekeka kunye nemizekelo yokuvavanya; Into ephambili kukuphucula ukusebenza, ukuqwalasela ipharamitha nganye kwiimeko ezithile, kwaye uqonde zonke iinketho kunye namaqhinga anikezelwa ngamaxwebhu asemthethweni kunye namava abanye abaqeqeshi. Kweli nqaku, siya kudlula, inyathelo ngenyathelo kunye nobunzulu, yonke into oyifunayo ukuze ukwazi ukufumana okuninzi kwi-ML Kit scanner kwi-Android, ukusuka ekufakeni ukuya kwizicwangciso ezingaphezulu, kubandakanywa uhlalutyo lomfanekiso, ulawulo lokulungelelanisa, kunye neengcebiso ezibalulekileyo zokusebenza ngexesha langempela.
Yintoni i-ML Kit kwaye kutheni iguqule ukuskena kwi-Android
I-ML Kit yi-SDK ephuhliswe nguGoogle edibanisa itekhnoloji enamandla yengqondo eyenziweyo kwimisebenzi yombono wekhompyuter kwizixhobo eziphathwayo. Ayivumeli kuphela usetyenziso olulula lweendlela zokufunda koomatshini, kodwa ikwenza oko nakwisixhobo ngokwaso, ngaphandle kwesidingo soqhagamshelo lwe-intanethi. Oku kuyenza ibe sisixhobo esithembekileyo, esikhawulezayo, nesiluncedo kwizicelo ezifuna ukuskena ikhowudi, ukuqondwa kobuso, ukufunda okubhaliweyo, okanye ukuchongwa kwento.
Uyilo lwemodyuli yeML Kit lubonelela ngokuguquguquka kwabo bafuna ukubandakanya izinto eziyimfuneko okanye ukulungelelanisa ukuya kweyona nkcukacha incinci. Unokukhetha phakathi kweemodeli ezipakishweyo (ezinkulu, ezifumanekayo kwasekuqaleni) okanye iimodeli ezikhutshelwe ngamandla (ezigcina indawo kwi-app kodwa zifuna ukhuphelo lokuqala). Oku kubini kukuvumela ukuba ubeke phambili ubungakanani beapp okanye ukusetyenziswa kwangoko ngokuxhomekeke kwiimfuno zeprojekthi.
Ukudibanisa iKhithi yeML kunye neKhameraX: Udibaniso oluPhumelelayo
Ukuba ujonge ukudityaniswa okuqinileyo phakathi kwekhamera ye-Android kunye nobuchule beML Kit, iCameraX yeyona nto ilungileyo. Eli thala leencwadi liququzelela kakhulu ukufikelela kwi-hardware yekhamera kunye nolawulo lomjelo womfanekiso, kwaye likwavumela ukugquma izinto zojongano kunye neziphumo zokufunda komatshini ngaphezulu kokubonwa kwekhamera.
Ngeklasi ML Kit Analyzer, unokudibanisa imveliso yeCameraX kunye ne-detectors kunye nabahlalutyi be-ML Kit. Olu hlalutyi luphumeza ujongano Uhlalutyo Lomfanekiso.Uhlalutyi, ukulawula ngokufanelekileyo isisombululo somfanekiso, ukulungelelanisa ukuguqulwa, kunye nokuhanjiswa kweziphumo, okwenza lula kakhulu uphuhliso lwemisebenzi yokuskena ephezulu.
Ukusebenzisa iCameraController kunye nePreviewView yenza lula umboniso wojongano kunye nokwamkelwa kweziphumo. Xa udibanisa i-ML Kit Analyzer, kufuneka uchaze kuphela uhlobo lwe-detector oyifunayo (umzekelo, i-barcode) kunye nendlela ofuna ukufumana ngayo iziphumo. Ukongeza, i-CameraX ithatha unonophelo lweenkcukacha ezinje ngokujikeleza komfanekiso kunye ne-aspect ratio, ukuphepha iimpazamo eziqhelekileyo kuphuhliso lwezandla.
Ukubona kunye nokucacisa iikhowudi zebhakhowudi: ngaphaya komzekelo osisiseko
El ukuskena ibhakhowudi Yenye yeemeko zokusetyenziswa kweenkwenkwezi zeML Kit kwi-Android. Into enomdla kukuba i-API yayo ikuvumela ukuba uye kude ngaphaya kokubona ikhowudi ye-QR eqhelekileyo. Ukusuka kwi-GS1, EAN-13, PDF417 ukuya kwi-Data Matrix okanye i-Aztec, ukuguquguquka kuninzi, ukuvula umnyango wezicelo kwi-logistics, urhwebo, ukuchongwa kwemveliso, kunye nokunye okuninzi.
Xa uqwalasela i-detector, ungayinqanda kwiifomati ozifuna ngokwenene, ezikhokelela ekubeni isantya esongeziweyo kunye nokusetyenziswa okuphantsi kobutyebi. Ngokunciphisa uphendlo, ukusetyenzwa kusebenza kakuhle kwaye kuphendula ngcono ngexesha lokwenyani.
Phakathi kokhetho oluphawulweyo lu:
- enableAllPotentialBarcodes(): Ikuvumela ukuba ubone zonke iikhowudi ezikhoyo emfanekisweni, nokuba ezinye azikwazi ukuchazwa. Oku kuluncedo xa umsebenzisi efuna ukusondeza okanye ukugxila ngcono kwikhamera.
- setZoomSuggestionOptions(): Yenza isondezelo ngokuzenzekelayo, ke i-app inokucebisa ngokuzenzekelayo ukuba ikhamera kufuneka isondeze kangakanani ukuze yandise ukufundeka kwekhowudi echongiweyo.
Isindululo esizenzekelayo sokusondeza luphawu olutsha oluphambili okuphucula amava omsebenzisi, ngakumbi kwiimeko apho umgama okanye ubungakanani beekhowudi zihluka kakhulu. Unakho ukuphumeza eyakho buya utsale umnxeba ukulungisa iparameters zoom ngokweengcebiso ze ML Kit detector.
Ukulungisa ngokuchanekileyo imifanekiso yegalelo
Enye yezona zinto zibaluleke kakhulu ekusebenzeni kwe-ML Kit scanner ngumgangatho kunye nobukhulu bemifanekiso yegalelo. Amaxwebhu asemthethweni abeka ugxininiso olukhulu kwiipikseli ezaneleyo kunye nesisombululo, njengoko ukuchaneka kokufumanisa ngokuthe ngqo kuxhomekeke ekubonakalisweni kwedatha emfanekisweni.
Umzekelo, kwiikhowudi ze-EAN-13, imivalo kunye nezithuba kufuneka ubuncinane zibe 2 pixels ububanzi; Ikhowudi yonke kufuneka ingabi ngaphantsi kwe-190 pixels ububanzi. Kwiifomathi ezintsonkothileyo ngakumbi njengePDF417, imilinganiselo efunekayo inkulu ngakumbi, ifikelela kwiipikseli ezili-1156 ububanzi kumqolo omnye.
Ukugxila kunye nesisombululo kubalulekile. Ukuba umsebenzisi ufota umfanekiso ongacacanga okanye onesoso esisezantsi, iziphumo zisenokungangqinelani. Isenzo esihle kukucebisa imifanekiso enesisombululo se-1280 x 720 okanye i-1920 x 1080, nje kuphela ukuba ukusebenza kwesixhobo kuvumela. Ukuba i-latency ngumba, unokunciphisa isisombululo, kodwa qiniseka ukuba ikhowudi ithatha umfanekiso omninzi kangangoko kunokwenzeka.
ML Kit's own API ikuvumela ukuba uguqule imithombo yomfanekiso oyifumana kwikhamera ukuya InputImage. Ungakha le nto ukusuka:
- imidiya ethile yekhamera.Umfanekiso.
- Bitmap, ByteBuffer okanye ByteArray.
- Ifayile ye-URI (iluncedo ekulayisheni imifanekiso kwigalari).
Kwimeko yeCameraX, uninzi lwezi nguqulelo zisonjululwa ngokuzenzekelayo., ngakumbi malunga nokujikeleza komfanekiso kunye nokunikezelwa kwesakhelo sexesha langempela. Oku kuthatha ubunzima obuninzi kumphuhlisi kwaye iqinisekisa ukuba umfanekiso ocutshungulweyo uhlala ulungile.
Ukucubungula imifanekiso kunye nokulawula iziphumo
Emva kokuba umfanekiso wegalelo ulungisiwe, i-ML Kit detector iqhuba umfanekiso kwaye ibuyisela uluhlu lweziphumo. Ngeebhakhowudi, uya kufumana izinto yebhakhowudi ekuvumela ukuba ufikelele kwizilungelelanisi emfanekisweni, ixabiso elichaziweyo, uhlobo lwedatha (URL, umfowunelwa, umbhalo, njl.njl.), kunye neempawu ezongezelelweyo ukuba zikhona.
Ulawulo lwezi ziphumo lubhetyebhetye. Unokwaleka ulwazi kumfanekiso, unxibelelane nomsebenzisi (umzekelo, ngokuvula ikhonkco lewebhu ukuba yikhowudi ye-QR), okanye ugcine amaxabiso afunyenweyo kwisiseko sedatha. Enkosi ekusebenziseni abaphulaphuli, ungalawula zombini iimpumelelo kunye neempazamo okanye iziganeko zemifanekiso ezingaqatshelwanga.
Ngokudityaniswa neCameraX, kubalulekile ukuvala ngokufanelekileyo nganye Umfanekiso weProxy emva kokuyicubungula, ukunqanda ukuvuza kwememori kunye nokugcina ubumanzi bohlalutyo lwexesha lokwenyani.
Ukuphucula uhlalutyo lwexesha langempela
Amandla apheleleyo e-ML Kit akhululwa xa uqhuba imifanekiso ngexesha langempela, umzekelo, ukusuka kumjelo wevidiyo. Ukugcina amava agudileyo, asimahla, kukho iingcebiso ezininzi eziphambili:
- Musa ukusebenzisa esona sisombululo siphezulu sekhamera ngaphandle kokuba kuyimfuneko. Kwiimeko ezininzi, imifanekiso ye-2 megapixel ingaphezulu kunokwaneleyo kwaye iphucula isantya sokusebenza.
- Nciphisa isisombululo ukuba isantya siphambili, kodwa ukugcina ubuncinci obuyimfuneko kwikhowudi echanekileyo.
- Kwimijelo yevidiyo, umkhangeli unokunika iziphumo ezahlukeneyo phakathi kwezakhelo ezilandelelanayo. Kuyacetyiswa ukuba ulinde ubuncinci ufunyanwe olufanayo olulandelelanayo ngaphambi kokuba uqwalasele iziphumo njengezisebenzayo.
- Inciphisa inani leefowuni eziya kwisixhobo. Umzekelo, ukuba umfanekiso sele usenziwa, ungahoyi izakhelo ezilandelayo de ugqibe.
- Ukuba usebenzisa iCameraX, qiniseka ukuba usete isicwangciso Uhlalutyo Lomfanekiso.STRATEGY_KEEP_ONLY_LATEST ukuhlala uhambisa owona mfanekiso wamva nje kwaye ugcine usetyenziso lusebenza.
- Ukwaleka imizobo emfanekisweni (umzekelo, iibhokisi zokubopha), kuqala iqhubekisela phambili umfanekiso ize ke inike zombini umfanekiso kunye nesigqubuthelo kwipasi enye, ikhulisa inani lomsebenzi wegraphic kwisakhelo ngasinye.
Uhlobo lwefomathi yomfanekiso nalo lubalulekile. Ukuba usebenzisa iKhamera2, khetha YUV_420_888; kwi-APIs ezindala, NV21 iqinisekisa ukuhambelana kunye nesantya.
Ukufunyanwa kwento kunye nokulandelela kwi-ML Kit
Ubunakho be-ML Kit scanner ayiphelelanga kwiibhakhowudi: ukufunyanwa kwento kunye nokulandelwa yenye into ebalulekileyo. Le API ikuvumela ukuba uchonge ukuya kuthi ga kwizinto ezintlanu kumfanekiso okanye isakhelo, ufumane indawo yazo kwaye, ukongeza, ukwabela i-ID yokulandelela eyodwa komnye. Xa usebenza ngevidiyo yexesha lokwenyani, unokulandelela intshukumo yento phezu kwezakhelo ezilandelelanayo ngokuchaneka okupheleleyo.
Iisetingi zokubona into zikuvumela ukuba:
- Khetha phakathi kwendlela yothumelo (STREAM_MODE), yenzelwe i-latency ephantsi kunye nokulandela umkhondo, kunye nemo yomfanekiso omnye (SINGLE_IMAGE_MODE).
- Vula okanye uvale izinto zokuhlela ngokweendidi ezibanzi (ifashoni, ukutya, ikhaya, iindawo, izityalo, abantu ongabaziyo).
- Cwangcisa ukuba uyafuna na ukucubungula izinto ezininzi ngaxeshanye okanye eyona ibalulekileyo.
Izicelo zisusela ekuqatshelweni kwemveliso ezivenkileni, uhlalutyo lwezityalo okanye lwezilwanyana kwii-apps zemfundo, ukuya kwiinkqubo zoncedo oluphezulu kunye noncedo lwendawo yokugcina impahla.
Ukulungiswa komfanekiso kunye nolawulo lokujikeleza lulandela imigaqo efanayo ekhankanywe kwikhowudi yokuskena, kwaye kuyacetyiswa ukuba usebenzise InputImage yenziwe ngokuthe ngqo kwiifomati ezixhaswayo ukunyusa ukusebenza.
Ukuqaphela umbhalo kunye nokufunyanwa kobuso
Ukuqatshelwa kombhalo (OCR) kunye nohlalutyo lokwenene lobuso ngezinye iintsika ezimbini eziphambili zeML Kit. I-Text API iyakwazi ukufumana iibhloko, imigca, kunye nezinto ngaphakathi komfanekiso, zibuyisela indawo yazo kunye nomxholo owaziwayo, nokuba ngabalinganiswa abanzima okanye iilwimi ezahlukeneyo.
Ukuqondwa kobuso kongeza ukukwazi ukuchonga iikhonto, iimpawu ezithile (amehlo, umlomo, impumlo, njl. njl.), misela izinto ezinokwenzeka ezinxulumene nezijekulo (ezifana nokuncuma okanye ukuqhwanyaza), kwaye uzobe kumfanekiso ukujonga iziphumo. Ngokudibanisa iiklasi zesiko, unokwenza, umzekelo, ugqume iibhokisi okanye amachaphaza phezu kobuso obufunyenweyo kwaye ubonise ulwazi olwahlukeneyo kumsebenzisi.
Olu hlobo lophumezo lufuna ulawulo olongezelelweyo ngokwemizobo yokumelwa., kodwa ukudibanisa neCameraX kunye I-GraphicOverlay yenza inkqubo ibe lula. Abaphuhlisi banokwenza ngokwezifiso ngokupheleleyo ujongano kwaye bahlengahlengise iziphumo zohlalutyo oluzenzekelayo kwisitayile sabo esibonakalayo esifunwayo.
Iingcebiso zokufumana amava afanelekileyo omsebenzisi
Impumelelo yefitsha yokuskena ephucukileyo ayixhomekekanga kuphela kwii-algorithms; ikwangundoqo ekuthatheni unonophelo lwamava omsebenzisi. Nazi izikhokelo eziluncedo:
- Qinisekisa ukuba izinto ofuna ukuzibhaqa zivelele ngokwaneleyo kwaye zineenkcukacha ezibonakalayo. Izinto ezinomahluko ophantsi okanye iinkcukacha ezingacacanga zinokufuna ukuba umsebenzisi azisondeze kwikhamera.
- Xa usebenzisa ulwahlulo lwento, zilungiselele ukuphatha izinto ezingaziwayo okanye ezingaqondakaliyo, ukunika ingxelo ecacileyo kumsebenzisi.
- Bandakanya iimpawu ezibonakalayo okanye ezibhaliweyo kwi-interface ukunceda umsebenzisi ukuba agxininise ngokuchanekileyo, ukuphepha ukukhungatheka ngenxa yeziphumo eziphosakeleyo.
- Ukuba izinto ezininzi zichongiwe, qwalasela ukhetho lomsebenzisi ukukhetha isiphumo esifanelekileyo (umzekelo, kwii-apps zemfundo okanye ze-inventri).
- Khumbula ukulungelelanisa ujongano lwakho kunye nemiyalezo yeempazamo kwiimeko ezingasebenziyo, ngakumbi ukuba usebenzisa iimodeli ezikhutshelwe ngamandla.
Uhlalutyo lweziphumo kunye nemizekelo ebonakalayo
Ubunakho be ML Kit Zisivumela ukuba sinikezele ngolwazi olutyebileyo nolucwangcisiweyo kumsebenzisi wokugqibela. Kuzo zombini ubhaqo lwento kunye nokuskena ibhakhowudi, uneendawo, amaxabiso, iindidi, kunye namanqanaba okuzithemba. Oku kwenza kube nokwenzeka ukubonisa iitheyibhile, izishwankathelo, okanye ukwenza izenzo ezizenzekelayo (ezifana nokuvula amakhonkco okanye ukugcina ulwazi ngasemva).
Ngokomzekelo, kwimeko yokukhangela into, unokubonisa umsebenzisi kunye nodidi oluchongiweyo kunye nenqanaba lokuzithemba, i-ID yokulandelela ukuthelekisa into efanayo kwizakhelo ezininzi, kunye nemida yendawo efunyenweyo. Ke, kwi-app ye-inventri, kulula ukuqaqambisa eyona nto iphambili kwaye uyilandele isakhelo ngesakhelo ukusetyenzwa kweentshukumo okanye ukucupha ii-automations.
Udibaniso phakathi kweemodyuli ezahlukeneyo ze-ML Kit ikuvumela ukuba udibanise imisebenzi: Yiba nomfanekiso ngqondweni ufunda ibhakhowudi kwinto ebhaqiweyo, ufumane ixabiso layo kwaye uyihlele, konke ngexesha lokwenyani nakumfanekiso wekhamera. Ezi ntlobo ze-synergies zinokwenzeka kuphela ngoqwalaselo oluphezulu kunye nolawulo olupheleleyo lwazo zonke iinketho ezikhoyo.
Indlela yokugcina yonke into isexesheni kwaye uphephe iingxaki
Izinga lohlaziyo lwamathala eencwadi e-ML Kit liphezulu kwaye kubalulekile ukuba uhlale phezulu kukhupho olutsha ngalunye, njengoko ukusebenza, ukuhambelana, kunye nokuphuculwa kwezinto ezintsha ziziswa rhoqo.
Ngalo lonke ixesha khangela ukuxhomekeka kwifayile yakho yakha.gradle kwaye ujonge ukuba ubuncinci benqanaba le-SDK lichazwe ngokuchanekileyo (ngokuqhelekileyo i-API 21 okanye ngaphezulu). Kwakhona, qwalasela iindlela zokuqalisa, njengoko iinguqulelo zamva nje ziphucule ezinye iinkqubo kwaye zazisa iimpawu ezintsha, ezinjengokukhuphela imodeli ngokuzenzekelayo xa ufaka i-app.
Enye ingcebiso kukubonisana neesampulu zeapps kunye nemathiriyeli yokufundisa ekhoyo kuzo zombini iimpepha ezisemthethweni zikaGoogle kunye neendawo ezivulelekileyo zokugcina. Ezi zibonelelo zihlala ziquka Iimvavanyo zokusebenza, imizekelo yokudibanisa kunye nojongano oluphambili, kunye nezisombululo kwiimpazamo eziqhelekileyo.
Gcina ukuhambelana kunye nokwenza ngcono kuzo zonke izixhobo ezahlukeneyo
Omnye wemicelimngeni yokuphuhlisa usetyenziso lwe-Android luluhlu olukhulu lwezixhobo, izisombululo, kunye neekhamera ezikhoyo kwintengiso. I-ML Kit ilungiselelwe uluhlu olubanzi lwezixhobo, kodwa ihlala ingumbono olungileyo ukuvavanya i-app kwizixhobo ezininzi, zombini isiphelo esiphezulu kunye nesiphelo esisezantsi, ukuqinisekisa ukuba ukusetyenzwa kunye nomgangatho wokubonwayo ngokwaneleyo kuzo zonke iimeko.
Ukuba abaphulaphuli bakho ikakhulu basebenzisa izixhobo ezinamandla aphantsi, beka phambili ukusebenza kakuhle kunye nesantya, isisombululo sokuncama okanye uluhlu lokusebenza ukuba kuyimfuneko ukunqanda ukubeka esichengeni amava omsebenzisi.
Siphetha ngombono ohlaziyayo amanqaku aphambili okufumana okuninzi kwi-ML Kit scanner kwi-Android: khetha esona sicwangciso silungileyo sokuhlanganisa phakathi kweCameraX kunye neML Kit, lungisa imifanekiso yakho yegalelo kakuhle, lawula ngokufanelekileyo iifomathi kunye nezisombululo ngokusetyenziswa kwakho, kwaye wenze ngokwezifiso ujongano lwakho kunye nokuhamba komsebenzi ukuze unike awona mava anamandla kwaye aphezulu anokwenzeka kubo bobabini abasebenzisi kunye nabaphuhlisi.
