Muva nje, ukuzivikela kobusika okulandelayo kwabathweswe iziqu ezintathu zamaphrojekthi wethu wezobuchwepheshe kwenzeka - i-Technopark (Bauman MSTU), i-Technosphere (i-Lomonosov Moscow State University) ne-Technotrek (MIPT). Amaqembu ethule kokubili ukuqaliswa kwemibono yawo kanye nezixazululo zezinkinga zangempela zebhizinisi eziphakanyiswe izigaba ezahlukene ze-Mai.ru Group.
Phakathi kwamaphrojekthi:
- Isevisi yokuthengisa izipho nge-augmented reality.
- Isevisi ehlanganisa amaphromoshini, izaphulelo nokunikezwa okuvela kuhlu lwamakheli.
- Ukusesha okubonakalayo kwezingubo.
- Isevisi yokuwela ibhuku nge-elekthronikhi ngenketho yokuqasha.
- Isithwebuli sokudla esihlakaniphile.
- Igayidi yomsindo yesimanje.
- Iphrojekthi "Imisebenzi ye-Mail.ru"
- Umabonakude weselula wesikhathi esizayo.
Sithanda ukukutshela kabanzi mayelana namaphrojekthi ayisithupha agqanyiswe ikakhulukazi amalungu nabeluleki bejaji.
Ukusesha okubonakalayo kwezingubo
Lo msebenzi wethulwe ithimba labathweswe iziqu eTechnosphere. Ngokusho kwabahlaziyi, imakethe yemfashini eRussia ngo-2018 yaba cishe ama-ruble angama-2,4 trillion. Abafana badale isevisi ebekwe njengomsizi ohlakaniphile wokuthenga ngezinhlobonhlobo ezinkulu zezimpahla. Lesi yisixazululo se-B2B esandisa ukusebenza kwezitolo eziku-inthanethi.
Ngesikhathi sokuhlolwa kwe-UX, abalobi bephrojekthi bathola ukuthi "ngokugqoka okufanayo" abantu baqonda ukufana hhayi ngombala noma iphethini, kodwa ngezimfanelo zezingubo. Ngakho-ke, abafana bahlakulela uhlelo olungaqhathanisi izithombe ezimbili kuphela, kodwa luqonda ukusondelana kwe-semantic. Ulayisha isithombe sempahla oyithandayo, futhi isevisi ikhetha imikhiqizo ehambisana nezici zayo.
Ngokobuchwepheshe uhlelo lusebenza kanje:
Inethiwekhi ye-Cascade Mask-RCNN ye-neural yaqeqeshelwa ukutholwa nokuhlukaniswa. Ukuze kunqunywe izibaluli nokufana kwezingubo, inethiwekhi ye-neural esekelwe ku-ResNext-50 enamakhanda ambalwa isetshenziselwa amaqembu ezibaluli, kanye nokulahlekelwa okuthathu kwezithombe zomkhiqizo owodwa. Yonke le phrojekthi yaqalwa ngokusekelwe ku-microservice architecture.
Ngokuzayo kuhlelwa:
- Yethula isevisi yazo zonke izigaba zezingubo.
- Yakha i-API yezitolo eziku-inthanethi.
- Thuthukisa ukukhohlisa kwesibaluli.
- Funda ukuqonda imibuzo ngolimi lwemvelo.
Ithimba lephrojekthi: Vladimir Belyaev, Petr Zaidel, Emil Bogomolov.
I-TV yeselula yesikhathi esizayo
Iphrojekthi yethimba le-Technopark. Abafundi badale uhlelo lokusebenza ngeshejuli ye-TV yeziteshi ezinkulu zokusakaza zedijithali zaseRussia, lapho kwenezelwa khona umsebenzi wokubuka iziteshi kusetshenziswa i-IPTV (iziteshi eziku-inthanethi) noma i-antenna.
Into enzima kakhulu kwakuwukunamathisela i-antenna kudivayisi ye-Android: ngenxa yalokhu basebenzisa i-tuner, lapho ababhali ngokwabo babhala khona umshayeli. Ngenxa yalokho, sithole ithuba lokubuka i-TV nokusebenzisa inkomba yohlelo lwe-TV ku-Android kuhlelo lokusebenza olulodwa.
Ithimba lephrojekthi: Konstantin Mitrakov, Sergey Lomachev.
Isevisi ehlanganisa amaphromoshini, izaphulelo nokunikezwa okuvela ohlwini lwama-imeyili
Lena iphrojekthi ezimpambanweni zokukhangisa nobuchwepheshe beposi. Amabhokisi ethu eposi agcwele ogaxekile nama-imeyili. Nsuku zonke sithola izincwadi ezinezaphulelo zomuntu siqu, kodwa sizivula kancane kancane, sizibona βnjengokukhangisa okungenamsebenzi.β Ngenxa yalokhu, abasebenzisi balahlekelwa izinzuzo futhi abakhangisi balahlekelwe. Ucwaningo olwenziwe yi-Mail.ru Mail lubonise ukuthi abasebenzisi bafuna ukubona isifinyezo sezaphulelo abanazo.
Le phrojekthi iqoqa ulwazi mayelana nezaphulelo namaphromoshini ephephabhukwini lakho lezindaba futhi ikubonise ngendlela yeribhoni yamakhadi ongaya kuyo uye kuwebhusayithi yokuphromotha noma i-imeyili. Uhlelo lungasebenza namabhokisi eposi ambalwa ngesikhathi esisodwa. Kukhona uhlu lwezitoko ezikhethiwe.
Le phrojekthi inokwakhiwa kwe-microservice futhi inezingxenye ezintathu eziyinhloko:
- Ukugunyazwa kwe-OAuth ukuze kuxhunywe kalula amabhokisi eposi.
- Ukuqoqwa nokuhlaziywa kwezinhlamvu ezinezinyuso.
- Ukugcina nokubonisa amakhadi esaphulelo.
Le phrojekthi isebenzisa ubuchwepheshe bokucubungula ulimi lwemvelo isebenzisa izinsiza ze-GPU: ama-accelerator ezithombe akwenze kwaba nokwenzeka ukwandisa isivinini sokucubungula izikhathi ezingu-50. I-algorithm isekelwe ohlelweni lwezimpendulo zemibuzo, olukuvumela ukuthi wengeze ngokushesha izigaba zesitoko ngokuhambisana nezidingo zebhizinisi elisha.
Leli qembu aligcinanga ngokuwina indawo emaqenjini aphezulu ngokusho kwejaji, kodwa futhi lanqoba umncintiswano we-"Digital Tops 2019". Lona umncintiswano wonjiniyela baseRussia abakha amathuluzi e-IT ukuthuthukisa ukusebenza kahle kwebhizinisi nama-ejensi kahulumeni, kanye nokukhulisa umkhiqizo womuntu siqu. Ithimba lethu liwine isigaba sabafundi.
Abafundi banezinhlelo ezinkulu zokuqhubeka nokuthuthuka kwephrojekthi, okulandelayo yilaba:
- Ukuhlanganiswa nezinsizakalo zemeyili.
- Ukuqaliswa kwesistimu yokuhlaziya izithombe.
- Kwethulwa iphrojekthi yezithameli eziningi.
Ithimba lephrojekthi: UMaxim Ermakov, uDenis Zinoviev, uNikita Rubinov.
Ngokwehlukana, sithanda ukukutshela mayelana namaqembu amathathu aqashelwe abeluleki be-Mail.ru Group abasebenze nabafundi kuyo yonke isemester. Kugxilwe kakhulu ekuxazululeni kwephrojekthi, ukuqaliswa nokusebenza ngokubambisana lapho kukhethwa amaphrojekthi.
Iphrojekthi "Imisebenzi ye-Mail.ru"
Le phrojekthi yaphawulwa yijaji kanye nabeluleki.
I-"Tasks Mail.ru" iyisevisi yokuqala ezimele yokugcina uhlu lwezinto okufanele zenziwe, ezakhiwe yinkampani. Ezinyangeni ezizayo, Imisebenzi izongena esikhundleni sohlu lwemisebenzi ku-Mail.ru Khalenda, futhi ngemva kokuba iphrojekthi isinikwe amandla kubo bonke abasebenzisi, izohlanganiswa ne-Mail.ru yeselula ne-Web Mail.
Iphrojekthi yaqalwa kusetshenziswa izindlela ezingaxhunyiwe ku-inthanethi zokuqala kanye nezamaselula-kuqala. Okusho ukuthi, ungasebenzisa uhlelo lwewebhu noma kunini, noma kuphi nanoma yini. Ukufinyelela i-inthanethi akusho lutho: idatha izogcinwa futhi ivumelaniswe. Ukuze kube lula kakhulu, ungakwazi "ukufaka" uhlelo lokusebenza kusuka kusiphequluli, futhi luzobukeka njengelomdabu.
Isithwebuli sokudla esihlakaniphile
Esitolo segilosa, asikwazi njalo ukunquma ngokushesha ukuthi umkhiqizo wokudla usifanele noma cha, ukuthi uphephe futhi unempilo kangakanani. Isimo siba nzima nakakhulu uma umuntu enemikhawulo yokudla, ukungezwani komzimba okuhlukahlukene, noma ekudleni. Uhlelo lokusebenza lwe-Foodwise Android likuvumela ukuthi uskene ibhakhodi yomkhiqizo futhi ubone kalula ukuthi iwufanele yini.
sebenzisa.
Uhlelo lokusebenza lunezigaba ezintathu eziyinhloko: "Iphrofayela", "Ikhamera" kanye "nomlando".
"Kuphrofayela" usetha okuncamelayo: esigabeni "Izithako" ungakhipha ekudleni kwakho noma yiziphi izithako ezingu-60 ezifakwe kusizindalwazi futhi ufunde ulwazi mayelana ne-E-supplements. "Amaqembu" akuvumela ukuthi ukhiphe ibhulokhi yezithako ngesikhathi esisodwa. Isibonelo, uma ucacisa "I-Vegetarianism," khona-ke yonke imikhiqizo equkethe inyama izogqanyiswa ngokubomvu.
Kukhona izindlela ezimbili esigabeni "Ikhamera": ukuskena amabhakhodi nokubona imifino nezithelo. Ngemva kokuskena ibhakhodi, uzothola lonke ulwazi mayelana nomkhiqizo. Izithako ozikhiphile zizogqanyiswa ngokubomvu.
Yonke imikhiqizo eskenwe ngaphambilini izolondolozwa Emlandweni. Lesi sigaba sihlonyiswe ngokusesha kombhalo nezwi.
Imodi yokuqaphela izithelo nemifino ikuvumela ukuthi uthole ulwazi mayelana nokudla kwakho kanye nenani lamandla. Isibonelo, i-apula elilodwa liqukethe cishe amagremu angu-25.
ama-carbohydrate, okungamukeleki kubantu abadla ukudla okune-carb ephansi.
Uhlelo lokusebenza lubhalwe nge-Kotlin; "Ikhamera" isebenzisa i-ML Kit ukuskena amabhakhodi futhi ihlonze izithelo nemifino. I-backend iqukethe izinsizakalo ezimbili: iseva ye-API ene-database,
egcina izithako ze-60 kanye nenhlanganisela yemikhiqizo ye-000, kanye nenethiwekhi ye-neural ebhalwe nge-Python ne-Tensorflow.
Ithimba lephrojekthi: Artyom Andryukhov, Ksenia Glazacheva, Dmitry Salman.
Isevisi yokuthengisa izipho nge-augmented reality
Wonke umuntu uthole izipho ezingokomfanekiso okungenani kanye empilweni yakhe. Ngokuvamile, kubantu, iqiniso lokunakwa libaluleke kakhulu kunesipho abasitholayo. Izipho ezinjalo azizuzisi, kodwa ukukhiqizwa nokulahlwa kwazo kunomthelela omubi esimweni seplanethi yethu. Lena yindlela ababhali bephrojekthi abaqhamuke ngayo nombono wokwenza isevisi yokuthengisa izipho ngeqiniso elingathandeki.
Ukuze sihlole ukuhlobana kombono, senze ucwaningo. U-82% wabaphenduli ubhekane nenkinga yokukhetha isipho. Ku-57% wabaphenduli, ubunzima obukhulu ekukhetheni kwakuwukwesaba ukuthi izipho zabo ngeke zisetshenziswe. U-78% wabantu ukulungele ukushintsha ukuxazulula izinkinga zemvelo.
Abaqambi babeke amaphuzu amathathu:
- Izipho zihlala emhlabeni obonakalayo.
- Abasithathi isikhala.
- Njalo eduzane.
Ukuze kusetshenziswe okungokoqobo okungathandwa kwabathelisi esikubona kuwebhu, ababhali bakhethe ilabhulali ye-AR.js, equkethe izingxenye ezimbili eziyinhloko:
- Owokuqala unesibopho sokudweba imidwebo phezu kokusakaza kwekhamera usebenzisa i-A-Frame noma i-Three.js.
- Ingxenye yesibili i-ARToolKit, enesibopho sokubona umaka (uhlamvu olukhethekile olungaphrintwa noma luboniswe esikrinini senye idivayisi) ekusakazweni kokuphuma kwekhamera. Umaka usetshenziselwa ukubeka imidwebo. Ukuba khona kwe-ARToolKit akukuvumeli ukuthi udale okungokoqobo okungathandwa kwabathelisi esikubona usebenzisa i-AR.js.
I-AR.js ifihla izingibe eziningi. Isibonelo, ukusetshenziswa kwayo kanye ne-A-Frame "kunganqamula" izitayela kusayithi lonke. Ngakho-ke, ababhali basebenzise "inqwaba" ye-AR.js + Three.js, eyasiza ukuxazulula ezinye zezinkinga. Futhi ukuze sishumeke i-AR.js esekelwe ku-Three.js ku-React, lapho kubhalwe khona iwebhusayithi yephrojekthi, bekufanele sakhe indawo yokugcina ye-AR-Test-2 (), esebenzisa ingxenye ehlukile ye-React yokusebenzisa i-AR.js esekelwe ku-Three.js. Ukubuka imodeli kokungokoqobo okungathandwa kwabathelisi esikubona kanye ne-3D (yamadivayisi angenayo ikhamera) kwasetshenziswa.
Kodwa-ke, kamuva kwavela ukuthi abasebenzisi abaqondi ukuthi yini umaka nokuthi usetshenziswa kanjani. Ngakho-ke, ababhali bashintshele kwezobuchwepheshe , okwamanje ethuthukiswa ngokusebenzayo yi-Google. Isebenzisa i-ARKit (iOS) noma i-ARCore (Android) ukuze inikeze amamodeli nge-AR ngaphandle komaka. Ubuchwepheshe busekelwe ku-Three.js futhi buhlanganisa isibukeli esiyimodeli ye-3D. Ukusebenziseka kohlelo sekuthuthuke kakhulu, nokho, ukuze ubuke okwangempela okungathandwa kwabathelisi esikubona, udinga idivayisi ene-iOS 12 noma eyakamuva.
Iphrojekthi isiyatholakala ku- (), lapho ungathola khona isipho sakho sokuqala.
Ithimba lephrojekthi: uDenis Stasyev, u-Anton Chadov.
Ungafunda okwengeziwe ngamaphrojekthi ethu ezemfundo ku . Futhi vakashela isiteshi kaningi , amavidiyo amasha okufundisa mayelana nezinhlelo, ukuthuthukiswa kanye neminye imikhakha avela lapho njalo.
Source: www.habr.com