Ukadaulo ndi zitsanzo zamakompyuta athu am'tsogolo zidapangidwa ndikusinthidwa pang'onopang'ono komanso m'mapulojekiti osiyanasiyana akampani yathu - mu Mail, Cloud, Search. Iwo anakhwima ngati tchizi wabwino kapena cognac. Tsiku lina tidazindikira kuti ma neural network athu amawonetsa zotsatira zabwino kwambiri pakuzindikirika, ndipo tidaganiza zowaphatikiza kukhala chinthu chimodzi cha b2b - Vision - yomwe timagwiritsa ntchito tokha ndikukupatsirani kuti mugwiritse ntchito.

Masiku ano, teknoloji yathu yowonetsera makompyuta pa nsanja ya Mail.Ru Cloud Solutions ikugwira ntchito bwino ndikuthetsa mavuto ovuta kwambiri. Zimatengera ma neural network angapo omwe amaphunzitsidwa pama seti athu a data ndipo amakhala okhazikika pakuthana ndi zovuta zomwe zimagwiritsidwa ntchito. Ntchito zonse zimagwira ntchito pa seva yathu. Mutha kuphatikiza Vision API yapagulu pamapulogalamu anu, momwe kuthekera konse kwautumiki kulipo. API ndi yachangu - chifukwa cha ma GPU a seva, nthawi yoyankhira pakati pa netiweki yathu ndi 100 ms.

Pitani ku mphaka, pali nkhani yatsatanetsatane ndi zitsanzo zambiri za ntchito ya Vision.

Monga chitsanzo cha ntchito yomwe ife tokha timagwiritsa ntchito matekinoloje ozindikiritsa nkhope omwe tatchulidwawa, titha kutchulapo. Events. Chimodzi mwa zigawo zake ndi Vision photo stands, zomwe timayika pamisonkhano yosiyanasiyana. Mukayandikira chithunzi chotere, tengani chithunzi ndi kamera yomangidwa ndikulowetsa imelo yanu, dongosololi lidzapeza nthawi yomweyo pakati pa zithunzi zomwe mudajambulidwa ndi ojambula ogwira ntchito pamsonkhanowo, ndipo, ngati mukufuna, adzakutumizirani zithunzi zomwe mwapeza ndi imelo. Ndipo sitikulankhula za zithunzi zojambulidwa pasiteji — Masomphenya amakuzindikirani ngakhale kumbuyo komwe kuli pakati pa alendo. Zoonadi, sizithunzi zomwe zimadziwikiratu zomwe zimadziwika, awa ndi mapiritsi okha omwe ali pamalo okongola omwe amangojambula zithunzi za alendo ndi makamera awo omangidwa ndikutumiza zambiri ku maseva, kumene matsenga onse ozindikira amapezeka. Ndipo tawona kangapo momwe ukadaulo waukadaulo ulili wodabwitsa ngakhale pakati pa akatswiri ozindikira zithunzi. Pansipa tikambirana zitsanzo zina.

1. Mtundu Wathu Wozindikira Nkhope

1.1. Neural network ndi liwiro la processing

Kuti tizindikire, timagwiritsa ntchito kusinthidwa kwa chitsanzo cha neural network cha ResNet 101. Avereji Pooling pamapeto pake amalowetsedwa ndi wosanjikiza wogwirizana, mofanana ndi momwe amachitira ku ArcFace. Komabe, kukula kwa mavekitala ndi 128, osati 512. Gulu lathu la maphunziro lili ndi zithunzi pafupifupi 10 miliyoni za anthu 273.

Mtunduwu umayenda mwachangu kwambiri chifukwa cha kasinthidwe ka seva kosankhidwa bwino ndi makompyuta a GPU. Zimatengera kuchokera ku 100 ms kuti mulandire yankho kuchokera ku API pamanetiweki athu amkati - izi zimaphatikizapo kuzindikira nkhope (kuzindikira nkhope pa chithunzi), kuzindikira ndi kubwezera PersonID mu yankho la API. Ndi kuchuluka kwa data yomwe ikubwera - zithunzi ndi makanema - zidzatenga nthawi yochulukirapo kusamutsa deta ku ntchito ndikulandila yankho.

1.2. Kuwunika momwe chitsanzocho chimagwirira ntchito

Koma kudziwa momwe ma neural network amathandizira ndizovuta kwambiri. Ubwino wa ntchito yawo umatengera ma data omwe zitsanzozo zidaphunzitsidwa komanso ngati zidakonzedwa kuti zigwire ntchito ndi deta inayake.

Tinayamba kuwunika kulondola kwachitsanzo chathu ndi mayeso odziwika bwino a LFW, koma ndi ochepa komanso osavuta. Pambuyo pofika ku 99,8% kulondola, sikuthandizanso. Pali mpikisano wabwino pakuwunika zitsanzo zozindikirika - Megaface, pomwe tidafika pang'onopang'ono 82% 1. Mayeso a Megaface ali ndi zithunzi miliyoni - zosokoneza - ndipo chitsanzocho chiyenera kusiyanitsa bwino zithunzi zikwi zingapo za anthu otchuka kuchokera ku Facescrub. deta kuchokera ku distractors. Komabe, titachotsa zolakwika zoyeserera za Megaface, tidapeza kuti ndi mtundu woyeretsedwa timakwaniritsa zolondola za 98% 1 (zithunzi za anthu otchuka nthawi zambiri zimakhala zenizeni). Chifukwa chake, adapanga chizindikiritso chosiyana, chofanana ndi Megaface, koma ndi zithunzi za anthu "wamba". Kenako tidawongolera kulondola kwa kuzindikira pamaseti athu ndikupita patsogolo. Kuphatikiza apo, timagwiritsa ntchito kuyesa kwamagulu komwe kumakhala ndi zithunzi masauzande angapo; imatsanzira zizindikiro za nkhope mumtambo wa wogwiritsa ntchito. Pamenepa, masango ndi magulu a anthu ofanana, gulu limodzi la munthu aliyense wodziwika. Tinayang'ana ubwino wa ntchito pamagulu enieni (zoona).

Zachidziwikire, zolakwika zozindikirika zimachitika ndi mtundu uliwonse. Koma mikhalidwe yotereyi nthawi zambiri imathetsedwa mwa kukonza bwino zitseko za zikhalidwe zinazake (pamisonkhano yonse timagwiritsa ntchito malire omwewo, koma, mwachitsanzo, pa machitidwe olamulira olowera tikuyenera kuonjezera kwambiri kuti pakhale zolakwika zochepa). Alendo ambiri amisonkhano adadziwika bwino ndi malo athu owonetsera zithunzi za Vision. Nthawi zina wina amayang'ana chithunzithunzi chodulidwa ndikuti, "Dongosolo lanu lalakwitsa, sindinali ine." Kenako tinatsegula chithunzi chonsecho, ndipo zidapezeka kuti panalidi mlendo uyu pachithunzipa, kungoti sitinali kumujambula, koma munthu wina, munthuyo adangokhala kumbuyo komweko. Komanso, neural network nthawi zambiri imazindikira molondola ngakhale mbali ya nkhopeyo sikuwoneka, kapena munthuyo atayima pa mbiri, kapena kutembenuka. Dongosolo limatha kuzindikira munthu ngakhale nkhope ili m'dera la kupotoza kwa kuwala, tinene, powombera ndi lens lalikulu.

1.3. Zitsanzo zoyesera pazovuta

Pansipa pali zitsanzo za momwe neural network yathu imagwirira ntchito. Zithunzi zimaperekedwa ku zomwe alowetsa, zomwe ayenera kuzilemba pogwiritsa ntchito PersonID - chizindikiritso chapadera cha munthu. Ngati zithunzi ziwiri kapena zingapo zili ndi ID yomweyo, ndiye, malinga ndi zitsanzo, zithunzizi zikuwonetsa munthu yemweyo.

Tiyeni tiwone nthawi yomweyo kuti tikamayesa, timakhala ndi mwayi wopeza magawo osiyanasiyana ndi mazenera omwe tingathe kukonza kuti tikwaniritse zotsatira zina. API ya anthu onse imakonzedwa kuti ikhale yolondola kwambiri pazochitika wamba.

Tiyeni tiyambe ndi chinthu chophweka, ndi kuzindikira nkhope ya kutsogolo.

Chabwino, izo zinali zophweka kwambiri. Tiyeni tiwunikire ntchitoyi, onjezerani ndevu ndi zaka zingapo.

Ena anganene kuti izi sizinali zovuta kwambiri, chifukwa muzochitika zonsezi nkhope yonse ikuwoneka, ndipo zambiri zokhudza nkhope zimapezeka ku algorithm. Chabwino, tiyeni tisinthe Tom Hardy kukhala mbiri. Vutoli ndizovuta kwambiri, ndipo tidachita khama kwambiri kuti tithane nalo ndikusunga zolakwika zochepa: tidasankha gulu lophunzitsira, kuganiza kudzera mu kamangidwe ka neural network, kulemekeza ntchito zotayika ndikuwongolera kukonzanso kusanachitike. za zithunzi.

Tiyeni timuveke mutu:

Mwa njira, ichi ndi chitsanzo cha zovuta kwambiri, popeza nkhope imabisika kwambiri, ndipo pa chithunzi chapansi palinso mthunzi wakuya wobisa maso. M'moyo weniweni, anthu nthawi zambiri amasintha maonekedwe awo mothandizidwa ndi magalasi akuda. Tichitenso chimodzimodzi ndi Tom.

Chabwino, tiyeni tiyese kuponya zithunzi za mibadwo yosiyana, ndipo nthawi ino tiyesa ndi wosewera wina. Tiyeni titenge chitsanzo chovuta kwambiri, pomwe kusintha kokhudzana ndi zaka kumawonekera makamaka. Zinthu sizili kutali; zimachitika nthawi zambiri pamene muyenera kufananiza chithunzi cha pasipoti ndi nkhope ya wonyamulayo. Pambuyo pake, chithunzi choyamba chikuwonjezeredwa ku pasipoti pamene mwiniwake ali ndi zaka 20, ndipo pofika zaka 45 munthu akhoza kusintha kwambiri:

Kodi mukuganiza kuti katswiri wamkulu pa ntchito zosatheka sanasinthe kwambiri ndi zaka? Ndikuganiza kuti ngakhale anthu ochepa angaphatikize zithunzi zapamwamba ndi zapansi, mnyamatayo wasintha kwambiri pazaka zambiri.

Neural network imakumana ndi kusintha kwamawonekedwe nthawi zambiri. Mwachitsanzo, nthawi zina amayi amatha kusintha kwambiri chithunzi chawo pogwiritsa ntchito zodzoladzola:

Tsopano tiyeni tiwunikire ntchitoyi kwambiri: tiyerekeze kuti mbali zosiyanasiyana za nkhope zaphimbidwa ndi zithunzi zosiyanasiyana. Zikatero, algorithm sangathe kufananiza zitsanzo zonse. Komabe, Vision imayendetsa bwino zinthu ngati izi.

Mwa njira, pakhoza kukhala nkhope zambiri pachithunzi, mwachitsanzo, anthu opitilira 100 amatha kukhala pachithunzi cha holo. Izi ndizovuta kwa ma neural network, popeza nkhope zambiri zimatha kuyatsidwa mosiyana, zina zosayang'ana. Komabe, ngati chithunzicho chitengedwa ndi kusamvana kokwanira komanso mtundu (osachepera ma pixel 75 pa lalikulu lalikulu kuphimba nkhope), Masomphenya azitha kuchizindikira ndikuchizindikira.

Chodabwitsa cha zithunzi ndi zithunzi zochokera ku makamera owunikira ndikuti anthu nthawi zambiri sawoneka bwino chifukwa anali osayang'ana kapena akusuntha panthawiyo:

Komanso, mphamvu yowunikira imatha kusiyana kwambiri kuchokera ku chithunzi kupita ku chithunzi. Izi, nazonso, nthawi zambiri zimakhala chopunthwitsa; ma aligorivimu ambiri amakhala ndi vuto lokonza bwino zithunzi zomwe zili zakuda kwambiri komanso zopepuka kwambiri, osatchulanso zofananira nazo. Ndiroleni ndikukumbutseni kuti kuti mukwaniritse izi muyenera kukonza zolowera mwanjira inayake; izi sizikupezeka pagulu. Timagwiritsa ntchito neural network yomweyo kwa makasitomala onse; ili ndi malire omwe ali oyenera ntchito zambiri zothandiza.

Posachedwa tatulutsa mtundu watsopano wamtunduwu womwe umazindikira nkhope zaku Asia molondola kwambiri. Ili kale linali vuto lalikulu, lomwe linkatchedwanso "kuphunzira pamakina" (kapena "neural network") kusankhana mitundu. European ndi American neural maukonde anazindikira Caucasian nkhope bwino, koma ndi Mongoloid ndi Negroid nkhope zinthu zinali zoipitsitsa. Mwinamwake, ku China zinthu zinali zosiyana ndendende. Zonse ndi zophunzitsa ma seti a data omwe amawonetsa mitundu yayikulu ya anthu m'dziko linalake. Komabe, zinthu zikusintha, ndipo masiku ano vutoli si lalikulu kwambiri. Masomphenya alibe vuto ndi anthu amitundu yosiyanasiyana.

Kuzindikira nkhope ndi chimodzi mwazinthu zambiri zaukadaulo wathu; Masomphenya amatha kuphunzitsidwa kuzindikira chilichonse. Mwachitsanzo, ziphaso zamalayisensi, kuphatikizira m'mikhalidwe yovuta ku ma aligorivimu: pamakona akuthwa, zonyansa komanso zovuta kuwerenga ma laisensi.

2. Milandu yothandiza

2.1. Kuwongolera mwakuthupi: pamene anthu awiri amagwiritsa ntchito chiphaso chimodzi

Mothandizidwa ndi Vision, mutha kugwiritsa ntchito machitidwe ojambulira kufika ndi kunyamuka kwa antchito. Dongosolo lachikhalidwe lochokera pamagetsi amagetsi lili ndi zovuta zoonekeratu, mwachitsanzo, mutha kudutsa anthu awiri pogwiritsa ntchito baji imodzi. Ngati njira yowongolera (ACS) ikuphatikizidwa ndi Vision, idzalemba moona mtima yemwe adabwera / adachoka komanso liti.

2.2. Kutsata nthawi

Mlanduwu wa Vision wogwiritsa ntchito umagwirizana kwambiri ndi wam'mbuyomu. Ngati muwonjezera njira yolumikizirana ndi ntchito yathu yozindikiritsa nkhope, sichitha kungozindikira kuphwanya kuwongolera, komanso kulembetsa kukhalapo kwenikweni kwa ogwira ntchito mnyumbayo kapena malo. Mwanjira ina, Masomphenya akuthandizani moona mtima omwe adabwera kuntchito ndikuchoka nthawi yanji, ndi amene adalumpha ntchito kwathunthu, ngakhale anzake atamuphimba pamaso pa akuluakulu ake.

2.3. Video Analytics: Kutsata Anthu ndi Chitetezo

Potsata anthu pogwiritsa ntchito Vision, mutha kuwunika molondola kuchuluka kwa magalimoto m'malo ogula, masitima apamtunda, misewu, misewu ndi malo ena ambiri. Kutsatira kwathu kungathandizenso kwambiri kuwongolera njira zolowera, mwachitsanzo, kumalo osungira katundu kapena malo ena ofunikira aofesi. Ndipo zowonadi, kutsatira anthu ndi nkhope kumathandiza kuthetsa mavuto achitetezo. Kodi mwapezapo munthu akuba m'sitolo yanu? Onjezani PersonID yake, yomwe idabwezedwa ndi Vision, pamndandanda wakuda wa pulogalamu yanu yowunikira makanema, ndipo nthawi ina dongosololi lidzachenjeza nthawi yomweyo chitetezo ngati mtundu uwu uwonekeranso.

2.4. Mu malonda

Mabizinesi ogulitsa ndi ntchito zosiyanasiyana ali ndi chidwi ndi kuzindikira pamzere. Mothandizidwa ndi Masomphenya, mutha kuzindikira kuti ili si gulu la anthu mwachisawawa, koma mzere, ndikuzindikira kutalika kwake. Kenako dongosololi limadziwitsa omwe akuyang'anira pamzere kuti athe kudziwa momwe zinthu zilili: mwina pamakhala kuchuluka kwa alendo komanso antchito owonjezera omwe akufunika kuyitanidwa, kapena wina akuchepetsa ntchito yawo.

Ntchito ina yosangalatsa ndikulekanitsa antchito akampani mu holo ndi alendo. Nthawi zambiri, dongosololi limaphunzitsidwa kulekanitsa zinthu zovala zovala zina (zovala) kapena zokhala ndi mawonekedwe apadera (scarf yodziwika bwino, beji pachifuwa, ndi zina zotero). Izi zimathandiza kuwunika molondola kuchuluka kwa opezekapo (kuti ogwira ntchito "asafufuze" ziwerengero za anthu muholoyo mwa kupezeka kwawo).

Pogwiritsa ntchito kuzindikira kumaso, mutha kuwunikanso omvera anu: kukhulupirika kwa alendo ndi chiyani, ndiko kuti, ndi anthu angati omwe amabwerera ku malo anu komanso pafupipafupi. Werengerani kuchuluka kwa alendo apadera omwe amabwera kwa inu pamwezi. Kuti muwongolere mtengo wokopa komanso kusunga, mutha kudziwanso kusintha kwa magalimoto kutengera tsiku la sabata komanso nthawi ya tsiku.

Ma franchisor ndi makampani amaketani amatha kuyitanitsa kuwunika kwa chithunzi cha mtundu wa malonda ogulitsa osiyanasiyana: kukhalapo kwa logo, zizindikilo, zikwangwani, zikwangwani, ndi zina zotero.

2.5. Ndi transport

Chitsanzo china chowonetsetsa chitetezo pogwiritsa ntchito kusanthula kwamavidiyo ndikuzindikira zinthu zomwe zasiyidwa m'mabwalo a ndege kapena masitima apamtunda. Masomphenya amatha kuphunzitsidwa kuzindikira zinthu zamagulu mazanamazana: mipando, zikwama, masutukesi, maambulera, mitundu yosiyanasiyana ya zovala, mabotolo, ndi zina zotero. Ngati makina anu owerengera makanema azindikira chinthu chopanda umwini ndikuchizindikira pogwiritsa ntchito Vision, amatumiza chizindikiro kuchitetezo. Ntchito yofananayi imagwirizanitsidwa ndi kuzindikira kwachilendo kwa zochitika zachilendo m'malo opezeka anthu ambiri: wina akumva kudwala, kapena wina amasuta pamalo olakwika, kapena munthu amagwera panjanji, ndi zina zotero - machitidwe onsewa amatha kudziwika ndi machitidwe owonetsera mavidiyo. kudzera mu Vision API.

2.6. Document flow

Ntchito ina yosangalatsa yamtsogolo ya Vision yomwe tikupanga pano ndikuzindikira zolemba komanso kusanja kwawo kukhala nkhokwe. M'malo molowetsa pamanja (kapena choyipirapo, kulowa) mndandanda wopanda malire, manambala, masiku otulutsidwa, manambala a akaunti, zambiri zakubanki, masiku ndi malo obadwira ndi zina zambiri zokhazikika, mutha kusanthula zikalata ndikuzitumiza zokha panjira yotetezeka kudzera pa API ku mtambo, komwe dongosolo lidzazindikira zolemba izi pa ntchentche, kuzigawanitsa ndi kubwezera yankho ndi deta mumtundu wofunikira kuti mulowemo mu database. Masiku ano Masomphenya akudziwa kale kugawa zikalata (kuphatikiza PDF) - amasiyanitsa mapasipoti, SNILS, TIN, zikalata zobadwa, zikalata zaukwati ndi zina.

Zachidziwikire, neural network siyimatha kuthana ndi zovuta zonsezi. Pazochitika zonse, chitsanzo chatsopano chimapangidwira makasitomala enieni, zinthu zambiri, ma nuances ndi zofunikira zimaganiziridwa, ma data amasankhidwa, ndipo kubwereza kwa maphunziro, kuyesa, ndi kasinthidwe kumachitika.

3. Ndondomeko ya ntchito ya API

"Chipata cholowera" cha Vision kwa ogwiritsa ntchito ndi REST API. Itha kulandira zithunzi, mafayilo amakanema ndi kuwulutsa kuchokera ku makamera a netiweki (mitsinje ya RTSP) ngati yolowera.

Kuti mugwiritse ntchito Vision, muyenera lowani mu ntchito ya Mail.ru Cloud Solutions ndikulandila ma tokeni (client_id + client_secret). Kutsimikizira kwa ogwiritsa ntchito kumachitika pogwiritsa ntchito protocol ya OAuth. Zomwe zili m'matupi a zopempha za POST zimatumizidwa ku API. Ndipo poyankha, kasitomala amalandira kuchokera ku API zotsatira zodziwika mu mtundu wa JSON, ndipo yankho limakonzedwa: liri ndi chidziwitso chokhudza zinthu zomwe zapezeka ndi makonzedwe awo.

Yankho lachitsanzo

{
   "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
}

Yankho lili ndi chidwi chochititsa chidwi - ichi ndi "kuzizira" kwa nkhope mu chithunzi, ndi chithandizo chake timasankha kuwombera bwino kwa nkhope kuchokera pamndandanda. Tidaphunzitsa neural network kuti ilosere mwayi woti chithunzi chizikondedwa pamasamba ochezera. Ubwino wa chithunzicho komanso nkhope ikumwetulira, zimakulitsa kudabwitsa.

API Vision imagwiritsa ntchito lingaliro lotchedwa space. Ichi ndi chida chopangira mitundu yosiyanasiyana ya nkhope. Zitsanzo za malo ndi mndandanda wakuda ndi woyera, mndandanda wa alendo, antchito, makasitomala, ndi zina zotero. Pa chizindikiro chilichonse mu Vision, mukhoza kupanga malo okwana 10, malo aliwonse akhoza kukhala ndi anthu 50 zikwi za PersonID, ndiye kuti, mpaka 500 zikwi. pa chizindikiro. Komanso, chiwerengero cha zizindikiro pa akaunti si malire.

Masiku ano API imathandizira njira zotsatirazi zozindikirira ndi kuzindikira:

• Zindikirani/Ikani - kuzindikira ndi kuzindikira nkhope. Imagawira munthu ID kwa munthu aliyense payekhapayekha, kubweza PersonID ndikugwirizanitsa anthu omwe apezeka.
• Chotsani - kuchotsa PersonID inayake pankhokwe ya munthu.
• Truncate - imachotsa malo onse kuchokera ku PersonID, yothandiza ngati idagwiritsidwa ntchito ngati malo oyesera ndipo muyenera kukonzanso nkhokwe kuti mupange.
• Zindikirani - kuzindikira kwa zinthu, zochitika, ma laisensi, malo, mizere, ndi zina zotero. Kubwezeretsanso gulu la zinthu zomwe zapezedwa ndi zolumikizira
• Kuzindikira zikalata - amazindikira mitundu yeniyeni ya zikalata Russian Federation (amasiyanitsa pasipoti, SNILS, nambala chizindikiritso msonkho, etc.).

Posachedwapa tikhala tikumaliza ntchito pa njira za OCR, kudziwa jenda, zaka ndi malingaliro, komanso kuthetsa mavuto ogulitsa, ndiko kuti, kuwongolera zokha zowonetsera katundu m'masitolo. Mutha kupeza zolemba zonse za API apa: https://mcs.mail.ru/help/vision-api

4. Kutsiliza

Tsopano, kudzera pagulu la API, mutha kuzindikira mawonekedwe a nkhope muzithunzi ndi makanema; chizindikiritso cha zinthu zosiyanasiyana, ma laisensi, zizindikiro, zikalata ndi zochitika zonse zimathandizidwa. Ntchito zochitika - nyanja. Bwerani, yesani ntchito yathu, ikhazikitseni ntchito zovuta kwambiri. Zoyamba za 5000 ndi zaulere. Mwina idzakhala "chosowa" pamapulojekiti anu.

Mutha kupeza API nthawi yomweyo mukalembetsa ndikulumikizana. Vision. Ogwiritsa ntchito onse a Habra amalandila nambala yotsatsira pazowonjezera zina. Chonde ndilembereni imelo adilesi yomwe mudagwiritsa ntchito polembetsa akaunti yanu!

Source: www.habr.com