Itekhnoloji kunye neemodeli zenkqubo yethu yombono wekhompyuter yexesha elizayo zenziwa kwaye zaphuculwa ngokuthe ngcembe nakwiiprojekthi ezahlukeneyo zenkampani yethu - kwi-Mail, Cloud, Search. Bakhula njengetshizi elungileyo okanye i-cognac. Ngenye imini siye sabona ukuba uthungelwano lwethu lwe-neural lubonisa iziphumo ezigqwesileyo ekuqaphelisweni, kwaye sagqiba ekubeni sizidibanise zibe yimveliso enye ye-b2b-Umbono-esiyisebenzisayo ngoku kwaye sinikezela ukuba uyisebenzise.
Namhlanje, iteknoloji yethu yombono wekhompyutheni kwi-platform ye-Mail.Ru Cloud Solutions isebenza ngempumelelo kwaye ixazulula iingxaki ezinzima kakhulu. Isekwe kwinani leenethiwekhi ze-neural eziqeqeshwe kwiiseti zethu zedatha kwaye zisebenza ngokukodwa ekusombululeni iingxaki ezisetyenzisiweyo. Zonke iinkonzo ziqhutywa kwizibonelelo zethu zeseva. Unokudibanisa i-API ye-Vision yoluntu kwizicelo zakho, apho zonke izakhono zenkonzo zikhoyo. I-API iyakhawuleza-enkosi kwi-GPU yeseva, ixesha lokuphendula eliphakathi kwinethiwekhi yethu yi-100 ms.
Yiya ekati, kukho ibali elineenkcukacha kunye nemizekelo emininzi yomsebenzi weVision.
Umzekelo wenkonzo apho thina ngokwethu sisebenzisa itekhnoloji yokuqaphela ubuso ekhankanyiweyo . Elinye lamacandelo ayo yiVision photo stands, esiyifaka kwiinkomfa ezahlukeneyo. Ukuba usondela kwindawo yokuma yeefoto, thatha ifoto kunye nekhamera eyakhelweyo kwaye ufake i-imeyile yakho, inkqubo iya kufumana ngokukhawuleza phakathi kweefoto ezininzi apho ubanjwe ngabafoti benkomfa, kwaye, ukuba unqwenela, izakuthumela iifoto ezifunyenweyo kuwe nge-imeyile. Kwaye asithethi ngokufota okuhleliweyo-Umbono uyakuqaphela nasemva kanye kwinginginya yeendwendwe. Ewe kunjalo, ayizizo iifoto ezizimela ngokwazo, ezi ziitafile nje ezikwindawo entle ezithatha iifoto zeendwendwe kunye neekhamera ezakhelwe ngaphakathi kwaye zigqithise ulwazi kwiiseva, apho kwenzeka khona wonke umlingo wokuqonda. Kwaye siye sabona ngaphezu kwesihlandlo esinye ukuba iyamangalisa indlela yokusebenza kwetekhnoloji naphakathi kweengcali zokuqaphela umfanekiso. Ngezantsi siza kuthetha ngemizekelo ethile.
1. IModeli yethu yokuNakana ubuso
1.1. Inethiwekhi ye-Neural kunye nesantya sokucubungula
Ukuqaphela, sisebenzisa ukuguqulwa kwemodeli yenethiwekhi ye-neural ye-ResNet 101. Umyinge we-Pooling ekupheleni uthatyathelwa indawo ngumaleko oqhagamshelwe ngokupheleleyo, kufana nendlela okwenziwa ngayo kwi-ArcFace. Nangona kunjalo, ubungakanani bokubonakaliswa kwe-vector yi-128, kungekhona i-512. Iseti yethu yoqeqesho iqulethe malunga ne-10 yezigidi zeefoto zabantu abangama-273.
Imodeli ibaleka ngokukhawuleza ngenxa yolwakhiwo olukhethwe ngononophelo lweseva kunye nekhompyuter ye-GPU. Kuthatha ukusuka kwi-100 ms ukufumana impendulo kwi-API kwiinethiwekhi zethu zangaphakathi - oku kubandakanya ukukhangela ubuso (ukubona ubuso kwisithombe), ukuqonda nokubuyisela i-PersonID kwimpendulo ye-API. Ngomthamo omkhulu wedatha engenayo - iifoto kunye neevidiyo - kuya kuthatha ixesha elide ukudlulisa idatha kwinkonzo kunye nokufumana impendulo.
1.2. Ukuvavanya ukusebenza komzekelo
Kodwa ukufumanisa ukusebenza kakuhle kothungelwano lwe-neural ngumsebenzi onzima kakhulu. Umgangatho womsebenzi wabo uxhomekeke ekubeni zeziphi iisethi zedatha iimodeli eziqeqeshwe kuzo kunye nokuba zilungiselelwe ukusebenza ngedatha ethile.
Siqale ukuvavanya ukuchaneka kwemodeli yethu kunye novavanyo lokuqinisekiswa kweLFW oludumileyo, kodwa luncinci kwaye lulula. Emva kokufikelela ku-99,8% ukuchaneka, akusekho luncedo. Kukho ukhuphiswano olulungileyo lokuvavanya iimodeli zokuqaphela - iMegaface, apho sathi ngokuthe ngcembe safikelela kwi-82% kwinqanaba le-1. Uvavanyo lweMegaface luneefoto ezisisigidi - iziphazamisi - kwaye imodeli kufuneka ikwazi ukwahlula kakuhle amawaka aliqela eefoto zabantu abadumileyo kwiFacescrub. isethi yedatha evela kwiziphazamisi. Nangona kunjalo, emva kokucima uvavanyo lweMegaface yeempazamo, safumanisa ukuba ngoguqulelo olucinyiweyo sifumana ukuchaneka kwe-98% yenqanaba 1 (iifoto zabantu abadumileyo zihlala zichanekile). Ke ngoko, benze uvavanyo lokuchonga olwahlukileyo, olufana neMegaface, kodwa ngeefoto zabantu "abaqhelekileyo". Emva koko siye saphucula ukuchaneka kolwazi kwiiseti zethu zedatha kwaye saya phambili. Ukongeza, sisebenzisa uvavanyo lomgangatho wokudibanisa oluqulathe amawaka aliqela eefoto; ilinganisa ukuthegiswa kobuso kwilifu lomsebenzisi. Kule meko, amaqela ngamaqela abantu abafanayo, iqela elinye kumntu ngamnye owaziwayo. Sihlolisise umgangatho womsebenzi kumaqela okwenene (oyinyani).
Ngokuqinisekileyo, iimpazamo zokuqaphela zenzeka kuyo nayiphi na imodeli. Kodwa iimeko ezinjalo zihlala zixazululwa ngokulungelelanisa imiqobo yeemeko ezithile (kuzo zonke iinkomfa sisebenzisa imilinganiselo efanayo, kodwa, umzekelo, kwiinkqubo zolawulo lokufikelela kufuneka sandise kakhulu imiqobo ukwenzela ukuba kubekho izinto ezimbalwa zobuxoki). Uninzi lwabakhenkethi benkomfa baqatshelwa ngokuchanekileyo ziindawo zethu zeefoto zeVision. Ngamanye amaxesha umntu wayejonga isibonisi esisikiweyo athi, "Inkqubo yakho yenze impazamo, yayingendim." Saye sayivula yonke loo foto, kwathi gqi ukuba ngenene kukho lo ndwendwe kulo mfanekiso, qha besingamfoti, kodwa omnye umntu, lo mntu usuke wasemva kwi-blur zone. Ngaphezu koko, inethiwekhi ye-neural ihlala iqonda ngokuchanekileyo naxa inxalenye yobuso ingabonakali, okanye umntu emi kwiprofayili, okanye ujike ngesiqingatha. Inkqubo inokubona umntu nokuba ubuso bukwindawo yokuphazamiseka kwe-optical, yithi, xa udubula nge-lens ebanzi.
1.3. Imizekelo yovavanyo kwiimeko ezinzima
Apha ngezantsi kukho imizekelo yendlela inethiwekhi yethu ye-neural esebenza ngayo. Iifoto zingeniswa kwigalelo, ekufuneka lileyibhelishe usebenzisa i-PersonID - isazisi esikhethekileyo somntu. Ukuba imifanekiso emibini okanye ngaphezulu ine-ID efanayo, ngoko, ngokweemodeli, ezi zithombe zibonisa umntu omnye.
Masiqaphele ngokukhawuleza ukuba xa sivavanya, sinokufikelela kwiiparitha ezahlukeneyo kunye neemodeli zemodeli esinokuyiqwalasela ukufezekisa umphumo othile. I-API yoluntu ilungiselelwe ukuchaneka okuphezulu kwiimeko eziqhelekileyo.
Masiqale ngeyona nto ilula, ngokuqaphela ubuso obujonge phambili.
Ewe, loo nto yayilula kakhulu. Masiwuxakekise umsebenzi, yongeza iindevu kunye nesandla seminyaka.
Abanye baya kuthi oku kwakungenzima kakhulu, kuba kuzo zombini iimeko ubuso bonke bubonakala, kwaye ulwazi oluninzi malunga nobuso luyafumaneka kwi-algorithm. Kulungile, masiguqule uTom Hardy kwiprofayile. Le ngxaki intsonkothe ngakumbi, kwaye sichithe umzamo omkhulu ekuyisombululeni ngempumelelo ngelixa sigcina inqanaba lempazamo elisezantsi: ukukhetha iseti yoqeqesho, ukucinga ngoyilo lwenethiwekhi ye-neural, ukuqinisa imisebenzi yelahleko kunye nokuphucula ukusetyenzwa kwangaphambili. iifoto.
Masimnxibe intloko:
Ngendlela, lo ngumzekelo wemeko enzima kakhulu, ekubeni ubuso bufihliwe kakhulu, kwaye kwisithombe esisezantsi kukho isithunzi esinzulu esifihla amehlo. Kubomi bokwenyani, abantu bahlala betshintsha imbonakalo yabo ngoncedo lweeglasi ezimnyama. Masenze okufanayo noTom.
Kulungile, makhe sizame ukuphosa iifoto ukusuka kwiminyaka eyahlukeneyo, kwaye ngeli xesha siza kuzama umdlali owahlukileyo. Makhe sithathe umzekelo onzima kakhulu, apho utshintsho olunxulumene neminyaka lubonakaliswa ngokukodwa. Imeko ayikho kude; kwenzeka rhoqo xa kufuneka uthelekise ifoto kwipaspoti kunye nobuso bomntu ophetheyo. Emva kwayo yonke loo nto, ifoto yokuqala yongezwa kwipasipoti xa umniniyo eneminyaka engama-20 ubudala, kwaye eneminyaka engama-45 umntu unokutshintsha kakhulu:
Ngaba ucinga ukuba ingcali ephambili kwimisebenzi engenakwenzeka ayitshintshanga kakhulu ngeminyaka? Ndicinga ukuba nokuba abantu abambalwa baya kudibanisa iifoto eziphezulu kunye nezantsi, inkwenkwe ishintshile kakhulu kwiminyaka.
Iinethiwekhi zeNeural zidibana notshintsho kwinkangeleko rhoqo. Ngokomzekelo, ngamanye amaxesha abafazi banokutshintsha kakhulu umfanekiso wabo ngoncedo lwezithambiso:
Ngoku masenze nzima ngakumbi umsebenzi: masithi iindawo ezahlukeneyo zobuso zigqunywe kwiifoto ezahlukeneyo. Kwiimeko ezinjalo, i-algorithm ayikwazi ukuthelekisa iisampuli ezipheleleyo. Nangona kunjalo, uMbono uzisingatha kakuhle iimeko ezinje.
Kakade ke, kusenokubakho ubuso obuninzi kwifoto; ngokomzekelo, abantu abangaphezu kwe-100 banokungena kwifoto eqhelekileyo yeholo. Le yimeko enzima kuthungelwano lwe-neural, kuba ubuso obuninzi bunokukhanya ngokwahlukileyo, obunye ngaphandle kokugxila. Nangona kunjalo, ukuba ifoto ithathwe ngesisombululo esaneleyo kunye nomgangatho (ubuncinci iipikseli ezingama-75 kwisikwere esigqume ubuso), uMbono uya kukwazi ukuyibona kwaye uyiqaphele.
Ubungangamsha beefoto zengxelo kunye nemifanekiso ephuma kwiikhamera zokucupha kukuba abantu bahlala bemfiliba ngenxa yokuba bebengajoliswanga okanye bebehamba ngelo xesha:
Kwakhona, ubukhulu bokukhanya bunokwahluka kakhulu ukusuka kumfanekiso ukuya kumfanekiso. Oku, nako, kuhlala kuba sisikhubekiso; Ii-algorithms ezininzi zinobunzima obukhulu bokwenza imifanekiso emnyama kakhulu kwaye ilula kakhulu, singasathethi ke ngokuyithelekisa ngokuchanekileyo. Makhe ndikukhumbuze ukuba ukuphumeza esi siphumo kufuneka uqwalasele imida ngendlela ethile; olu phawu alukafumaneki esidlangalaleni. Sisebenzisa inethiwekhi efanayo ye-neural kubo bonke abaxumi; ineentsika ezilungele uninzi lwemisebenzi ebonakalayo.
Kutshanje sikhuphe inguqulelo entsha yemodeli ebona ubuso baseAsia ngokuchaneka okuphezulu. Oku kwakukade kuyingxaki enkulu, eyayide yabizwa ngokuba “kukufunda ngomatshini” (okanye “uthungelwano lwemithambo-luvo”) ubuhlanga. Uthungelwano lwe-neural lwaseYurophu naseMelika lwaqaphela ubuso beCaucasian kakuhle, kodwa ngobuso be-Mongoloid kunye ne-Negroid imeko yayimbi kakhulu. Mhlawumbi, eTshayina imeko yayichasene ngokupheleleyo. Konke kumalunga neeseti zedatha zoqeqesho ezibonisa iindidi eziphambili zabantu kwilizwe elithile. Nangona kunjalo, imeko iyatshintsha; namhlanje le ngxaki ayikho nzima kangako. Umbono awunangxaki nabantu beentlanga ezahlukeneyo.
Ukuqondwa kobuso sesinye sezicelo ezininzi zetekhnoloji yethu; Umbono unokuqeqeshwa ukuqaphela nantoni na. Ngokomzekelo, iipleyiti zelayisensi, kubandakanywa kwiimeko ezinzima kwii-algorithms: kwii-angles ezibukhali, ezingcolileyo kwaye kunzima ukufunda iipleyiti zelayisensi.
2. Iimeko zokusetyenziswa okusebenzayo
2.1. Ulawulo lokufikelela ngokwasemzimbeni: xa abantu ababini besebenzisa ipasi enye
Ngoncedo loMbono, unokuphumeza iinkqubo zokurekhoda ukufika kunye nokuhamba kwabasebenzi. Inkqubo yendabuko esekelwe kwiipaseji ze-elektroniki ineziphene ezicacileyo, umzekelo, unokudlula abantu ababini usebenzisa ibheji enye. Ukuba inkqubo yokulawula ukufikelela (ACS) yongezwa kunye noMbono, iya kurekhoda ngokunyanisekileyo ukuba ngubani oweza / oshiye kwaye nini.
2.2. Ukulandelela ixesha
Le meko yokusetyenziswa kweVision inxulumene ngokusondeleyo nangaphambili. Ukuba uncedisa inkqubo yokufikelela kunye nenkonzo yethu yokuqaphela ubuso, ayiyi kukwazi ukubona kuphela ukuphulwa kolawulo lokufikelela, kodwa nokubhalisa ubukho bokwenene babasebenzi kwisakhiwo okanye kwiziko. Ngamanye amazwi, Vision iya kukunceda ngokunyanisekileyo ingqalelo ngubani beze emsebenzini kwaye wemka ngaliphi ixesha, yaye ngubani utsibe umsebenzi ngokupheleleyo, nokuba oogxa bakhe zigubungele kuye phambi kwabaphathi bakhe.
2.3. Uhlahlelo lweVidiyo: Ukulandelwa kwabantu kunye noKhuseleko
Ngokulandelela abantu usebenzisa uMbono, unokuvavanya ngokuchanekileyo itrafikhi yokwenyani yeendawo zokuthenga, izikhululo zikaloliwe, iipaseji, izitrato kunye nezinye iindawo zikawonke-wonke. Ukulandelela kwethu kwakhona kunokuba luncedo olukhulu ekulawuleni ukufikelela, umzekelo, kwindawo yokugcina impahla okanye kwezinye iindawo ezibalulekileyo zeofisi. Kwaye kunjalo, ukulandelela abantu kunye nobuso kunceda ukusombulula iingxaki zokhuseleko. Ngaba ubambe umntu ebe evenkileni yakho? Yongeza i-PersonID yakhe, ebuyiselwe nguMbono, kuluhlu olumnyama lwesoftware yakho yohlalutyo lwevidiyo, kwaye ngexesha elizayo inkqubo iya kwazisa ngokukhawuleza ukhuseleko ukuba olu hlobo luvela kwakhona.
2.4. Kurhwebo
Ukuthengisa kunye namashishini eenkonzo ezahlukeneyo anomdla ekuqapheliseni umgca. Ngoncedo loMbono, unokuqonda ukuba ayisiyiyo isihlwele sabantu, kodwa ngumgca, kwaye umisele ubude bayo. Kwaye ke inkqubo yazisa abo baphetheyo malunga nomgca ukuze bakwazi ukuyiqonda imeko: mhlawumbi kukho ukunyuka kweendwendwe kunye nabasebenzi abongezelelweyo kufuneka babizwe, okanye umntu uyancipha kwimisebenzi yabo yomsebenzi.
Omnye umsebenzi onomdla kukwahlula abasebenzi benkampani kwiholo kwiindwendwe. Ngokuqhelekileyo, inkqubo iqeqeshelwe ukwahlula izinto ezinxibe iimpahla ezithile (ikhowudi yesinxibo) okanye kunye nenye into eyahlukileyo (isikhafu esinophawu, ibheji esifubeni, njalo njalo). Oku kunceda ukuvavanya ngokuchanekileyo ngakumbi ukubakho emsebenzini (ukuze abasebenzi "bangakhuphisi" izibalo zabantu eholweni ngokubakho nje kwabo).
Usebenzisa ukubonwa kobuso, unokuvavanya abaphulaphuli bakho: yintoni ukunyaniseka kwabatyeleli, oko kukuthi, bangaphi abantu ababuyela kwindawo yakho kunye nokuba kuphi na rhoqo. Bala ukuba zingaphi iindwendwe ezizodwa eziza kuwe ngenyanga. Ukwandisa iindleko zomtsalane kunye nokugcinwa, unokufumana kwakhona utshintsho kwizithuthi ngokuxhomekeke kusuku lweveki kunye nexesha losuku.
IiFranchisor kunye neenkampani zamakhonkco zinokuodola uvavanyo olusekwe kwiifoto zomgangatho wophawu lwentengiso ezahlukeneyo: ubukho beelogo, iimpawu, iipowusta, iibhanela, njalo njalo.
2.5. Ngezothutho
Omnye umzekelo wokuqinisekisa ukhuseleko usebenzisa uhlalutyo lwevidiyo kukuchonga izinto ezilahliweyo kwiiholo zezikhululo zeenqwelomoya okanye kwizikhululo zikaloliwe. Umbono unokuqeqeshwa ukuqaphela izinto zamakhulu eeklasi: iingcezu zefenitshala, iibhegi, iisutikheyisi, iiambrela, iintlobo ezahlukeneyo zempahla, iibhotile, njalo njalo. Ukuba inkqubo yakho yohlalutyo lwevidiyo ibona into engenamniniyo kwaye iyayibona isebenzisa uMbono, ithumela umqondiso kwinkonzo yokhuseleko. Umsebenzi ofanayo uhambelana nokufunyanwa ngokuzenzekelayo kweemeko ezingaqhelekanga kwiindawo zoluntu: umntu uziva egula, okanye umntu utshaya kwindawo engafanelekanga, okanye umntu uwela kwiireyile, njalo njalo - zonke ezi patheni zinokuqatshelwa ngeenkqubo zokuhlalutya ividiyo. ngokusebenzisa Vision API.
2.6. Ukuhamba koxwebhu
Olunye usetyenziso olunomdla lwekamva loMbono esiwuphuhlisayo ngoku kukunakana kwamaxwebhu kunye nokwahlulwa kwawo okuzenzekelayo koovimba beenkcukacha. Endaweni yokungena ngesandla (okanye okubi ngakumbi, ukungena) uthotho olungapheliyo, amanani, imihla yokukhutshwa, iinombolo zeakhawunti, iinkcukacha zebhanki, imihla kunye neendawo zokuzalwa kunye nezinye iinkcukacha ezininzi ezisemthethweni, unokuskena amaxwebhu kwaye uwathumele ngokuzenzekelayo kwisitishi esikhuselekileyo nge I-API kwifu, apho inkqubo iya kuqonda la maxwebhu kwi-fly, ihlalutye kwaye ibuyisele impendulo ngedatha kwifomathi efunekayo yokungena ngokuzenzekelayo kwisiseko sedatha. Namhlanje Umbono sele uyayazi indlela yokuhlela amaxwebhu (kubandakanywa nePDF) - iyahlula phakathi kweepaspoti, i-SNILS, i-TIN, izatifikethi zokuzalwa, izatifikethi zomtshato kunye nabanye.
Ngokuqinisekileyo, inethiwekhi ye-neural ayikwazi ukusingatha zonke ezi meko ngaphandle kwebhokisi. Kwimeko nganye, imodeli entsha yakhelwe umthengi othile, izinto ezininzi, ama-nuances kunye neemfuno zithathelwa ingqalelo, iiseti zedatha zikhethiwe, kunye nokuphindaphinda koqeqesho, ukuvavanywa kunye noqwalaselo lwenziwa.
3. Iskimu sokusebenza kwe-API
"Isango lokungena" lombono kubasebenzisi yi-REST API. Inokufumana iifoto, iifayile zevidiyo kunye nosasazo kwiikhamera zenethiwekhi (imijelo ye-RTSP) njengegalelo.
Ukusebenzisa uMbono, kufuneka kwinkonzo ye-Mail.ru Cloud Solutions kwaye ufumane amathokheni okufikelela (client_id + client_secret). Uqinisekiso lomsebenzisi lwenziwa kusetyenziswa i-OAuth protocol. Idatha yomthombo kwimizimba yezicelo ze-POST ithunyelwa kwi-API. Kwaye ekuphenduleni, umxhasi ufumana kwi-API isiphumo sokuqaphela kwifomathi ye-JSON, kwaye impendulo yakhiwe: iqulethe ulwazi malunga nezinto ezifunyenweyo kunye nokulungelelanisa kwazo.
Umzekelo wempendulo
{
"status":200,
"body":{
"objects":[
{
"status":0,
"name":"file_0"
},
{
"status":0,
"name":"file_2",
"persons":[
{
"tag":"person9"
"coord":[149,60,234,181],
"confidence":0.9999,
"awesomeness":0.45
},
{
"tag":"person10"
"coord":[159,70,224,171],
"confidence":0.9998,
"awesomeness":0.32
}
]
}
{
"status":0,
"name":"file_3",
"persons":[
{
"tag":"person11",
"coord":[157,60,232,111],
"aliases":["person12", "person13"]
"confidence":0.9998,
"awesomeness":0.32
}
]
},
{
"status":0,
"name":"file_4",
"persons":[
{
"tag":"undefined"
"coord":[147,50,222,121],
"confidence":0.9997,
"awesomeness":0.26
}
]
}
],
"aliases_changed":false
},
"htmlencoded":false,
"last_modified":0
}Impendulo iqulethe ipharamitha enomdla wobuhle - oku kuxhomekeke "ukuphola" kobuso kwifoto, ngoncedo lwayo, sikhetha olona mfanekiso ulungileyo wobuso ngokulandelelana. Siqeqeshe inethiwekhi ye-neural ukuqikelela ukuba ifoto iya kuthandwa kwiintanethi zentlalo. Okukhona umgangatho wefoto ungcono kwaye ngakumbi ubuso obuncumayo, kokukhona ukulunga okukhulu.
Umbono we-API usebenzisa ingqiqo ebizwa ngokuba yindawo. Esi sisixhobo sokwenza iiseti ezahlukeneyo zobuso. Imizekelo izithuba uluhlu olumnyama nomhlophe, uluhlu abatyeleli, abasebenzi, abathengi, etc. Kuba umqondiso ngamnye Vision, uyakwazi ukwenza ukuya 10 izithuba, indawo nganye unokuba ukuya ku 50 amawaka PersonIDs, oko kukuthi, ukuya ku 500 amawaka. ngomqondiso . Ngaphezu koko, inani lamathokheni kwi-akhawunti ngalinye alikhawulelwanga.
Namhlanje i-API ixhasa ezi ndlela zilandelayo zokubona kunye nokuqaphela:
- Ukuqaphela/ukuseta-ukubona nokuqondwa kobuso. Ukwabela ngokuzenzekelayo i-PersonID kumntu ngamnye owahlukileyo, ibuyisela i-PersonID kunye nolungelelwaniso lwabantu abafunyenweyo.
- Cima - ukucima i-PersonID ethile kwiziko ledatha lomntu.
- I-Truncate - icoca yonke indawo esuka kwi-PersonID, iluncedo ukuba isetyenziswe njengendawo yokuvavanya kwaye kufuneka ubuyisele i-database yokuvelisa.
- Khangela -ukubona izinto, imiboniso, iipleyiti zelayisensi, iindawo ezibonisa indawo, imigca, njl njl. Ibuyisela udidi lwezinto ezifunyenweyo kunye nezilungelelanisi zazo
- Ukukhangela amaxwebhu - ufumanisa iintlobo ezithile zamaxwebhu e-Russian Federation (ukwahlula ipasipoti, i-SNILS, inombolo yesazisi serhafu, njl.).
Kwakhona kungekudala siza kugqiba umsebenzi kwiindlela ze-OCR, ukugqiba isini, ubudala kunye neemvakalelo, kunye nokusombulula iingxaki zokuthengisa, oko kukuthi, ukulawula ngokuzenzekelayo ukuboniswa kwempahla kwiivenkile. Ungafumana amaxwebhu apheleleyo e-API apha:
4. Isiphelo
Ngoku, nge-API kawonke-wonke, unokufikelela kulwazi lobuso kwiifoto kunye neevidiyo; ukuchongwa kwezinto ezahlukeneyo, iipleyiti zelayisensi, iimpawu zomhlaba, amaxwebhu kunye nemiboniso yonke iyaxhaswa. Iimeko zesicelo - ulwandle. Yiza, uvavanye inkonzo yethu, uyibeke eyona misebenzi inzima. Iintengiselwano zokuqala ezingama-5000 zisimahla. Mhlawumbi iya kuba "sisithako esilahlekileyo" kwiiprojekthi zakho.
Ukufikelela kwi-API kunokufumaneka ngokukhawuleza xa kubhaliswa kunye nokudibanisa . Bonke abasebenzisi be-Habra bafumana ikhowudi yokuphromotha kwiintengiselwano ezongezelelweyo. Nceda undibhalele idilesi ye-imeyile oyisebenzise xa ubhalisa iakhawunti yakho!
umthombo: www.habr.com