Posachedwapa, chitetezo chotsatira chachisanu cha omaliza maphunziro atatu a teknoloji yathu chinachitika - Technopark (Bauman MSTU), Technosphere (Lomonosov Moscow State University) ndi Technotrek (MIPT). Maguluwa adapereka kukhazikitsidwa kwa malingaliro awo ndi mayankho kumavuto enieni abizinesi omwe aperekedwa ndi magawo osiyanasiyana a Mai.ru Group.
Zina mwa ntchito:
- Service kugulitsa mphatso ndi augmented zenizeni.
- Ntchito yomwe imaphatikiza zokwezedwa, kuchotsera ndi zotsatsa kuchokera pamndandanda wamakalata.
- Kusaka zovala zowoneka bwino.
- Ntchito yowoloka mabuku pakompyuta ndi njira yobwereka.
- Smart food scanner.
- Kalozera wamakono wamawu.
- Ntchito "Mail.ru Tasks"
- Televizioni yam'manja yam'tsogolo.
Tikufuna kukuuzani mwatsatanetsatane za mapulojekiti asanu ndi limodzi omwe adawonetsedwa makamaka ndi mamembala a jury ndi alangizi.
Kusaka zovala zowoneka bwino
Ntchitoyi idaperekedwa ndi gulu la omaliza maphunziro a Technosphere. Malinga ndi akatswiri, msika wamafashoni ku Russia mu 2018 unali pafupifupi ma ruble 2,4 thililiyoni. Anyamatawo adapanga ntchito yomwe imayikidwa ngati wothandizira wanzeru pogula zinthu zosiyanasiyana. Ili ndi yankho la B2B lomwe limakulitsa magwiridwe antchito amasitolo apaintaneti.
Pakuyesa kwa UX, olemba pulojekitiyi adapeza kuti ndi "chovala chofanana" anthu amamvetsetsa kufanana osati mtundu kapena chitsanzo, koma mu zikhumbo za zovala. Choncho, anyamatawo adapanga dongosolo lomwe silimangofanizira zithunzi ziwiri, koma limamvetsetsa kuyandikana kwa semantic. Mumayika chithunzi cha zovala zomwe mukufuna, ndipo ntchitoyo imasankha zinthu zomwe zimagwirizana ndi zomwe zimapangidwira.
Mwaukadaulo, dongosololi limagwira ntchito motere:
Cascade Mask-RCNN neural network idaphunzitsidwa kuti izindikire ndikuyika magulu. Kuti mudziwe mawonekedwe ndi kufanana kwa zovala, neural network yozikidwa pa ResNext-50 yokhala ndi mitu ingapo imagwiritsidwa ntchito pamagulu azinthu, ndi kutayika kwa Triplet pazithunzi za chinthu chimodzi. Ntchito yonseyi idakhazikitsidwa motengera kamangidwe ka microservice.
M'tsogolomu zakonzedwa:
- Yambitsani ntchito yamagulu onse azovala.
- Pangani API yogulitsira pa intaneti.
- Limbikitsani kusintha kwa mawonekedwe.
- Phunzirani kumvetsetsa mafunso muchilankhulo chachilengedwe.
Gulu la polojekiti: Vladimir Belyaev, Petr Zaidel, Emil Bogomolov.
TV yam'manja yam'tsogolo
Project ya timu ya Technopark. Ophunzira adapanga pulogalamu yokhala ndi pulogalamu yapa TV yamakanema akulu aku Russia owulutsa digito, omwe adawonjezedwa ntchito yowonera makanema pogwiritsa ntchito IPTV (njira zapaintaneti) kapena mlongoti.
Chovuta kwambiri chinali kulumikiza mlongoti ku chipangizo cha Android: chifukwa cha ichi adagwiritsa ntchito tuner, yomwe olemba okha adalemba dalaivala. Zotsatira zake, tinali ndi mwayi wowonera TV ndikugwiritsa ntchito kalozera wa pulogalamu ya TV pa Android mu pulogalamu imodzi.
Gulu la polojekiti: Konstantin Mitrakov, Sergey Lomachev.
Ntchito yomwe imaphatikiza kukwezedwa, kuchotsera ndi zotsatsa kuchokera pamndandanda wamakalata
Iyi ndi projekiti yomwe ili pamphambano zamatekinoloje otsatsa ndi ma positi. Mabokosi athu amadzaza ndi sipamu ndi maimelo. Tsiku lililonse timalandira makalata okhala ndi kuchotsera kwaumwini, koma timatsegula pang'onopang'ono, tikumawona ngati "kutsatsa kopanda ntchito." Chifukwa cha izi, ogwiritsa ntchito amataya phindu ndipo otsatsa amawonongeka. Kafukufuku wa Mail.ru Mail adawonetsa kuti ogwiritsa ntchito akufuna kuwona chidule cha kuchotsera komwe ali nako.
Ntchitoyi amasonkhanitsa zambiri za kuchotsera ndi kukwezedwa kuchokera m'makalata anu ndikuziwonetsa ngati riboni yamakhadi omwe mutha kupitako patsamba lotsatsa kapena imelo. Pulogalamuyi imatha kugwira ntchito ndi ma mailbox angapo nthawi imodzi. Pali mndandanda wazinthu zosankhidwa.
Ntchitoyi ili ndi kamangidwe ka microservice ndipo ili ndi magawo atatu:
- Chilolezo cha OAuth kuti mulumikizidwe mosavuta pamabokosi a makalata.
- Kusonkhanitsa ndi kusanthula makalata okhala ndi zotsatsa.
- Kusunga ndi kuwonetsa makadi ochotsera.
Pulojekitiyi imagwiritsa ntchito ukadaulo wokonza zilankhulo zachilengedwe pogwiritsa ntchito zida za GPU: ma accelerator azithunzi adapangitsa kuti ziwonjezeke liwiro la kukonza ndi nthawi 50. The aligorivimu imachokera pamayankho a mafunso, omwe amakulolani kuti muwonjezere mwachangu magulu amasheya malinga ndi zofunikira zabizinesi.
Gululi silinangopeza malo m'magulu apamwamba malinga ndi oweruza, komanso linapambana mpikisano wa "Digital Tops 2019". Uwu ndi mpikisano wa otukula aku Russia omwe amapanga zida za IT kuti apititse patsogolo magwiridwe antchito abizinesi ndi mabungwe aboma, komanso kukulitsa zokolola zamunthu. Gulu lathu lapambana gulu la ophunzira.
Ophunzirawo ali ndi mapulani akuluakulu opititsa patsogolo ntchitoyo, otsatirawa ndi awa:
- Kuphatikiza ndi ntchito zamakalata.
- Kukhazikitsa dongosolo losanthula zithunzi.
- Kukhazikitsa pulojekiti kwa anthu ambiri.
Gulu la polojekiti: Maxim Ermakov, Denis Zinoviev, Nikita Rubinov.
Payokha, tikufuna kukuuzani zamagulu atatu omwe adazindikiridwa ndi alangizi a Mail.ru Group omwe adagwira ntchito ndi ophunzira mu semesita yonse. Chisamaliro chinaperekedwa ku zovuta za polojekiti, kukhazikitsa ndi kugwira ntchito pamodzi posankha ntchito.
Ntchito "Mail.ru Tasks"
Ntchitoyi idadziwika ndi oweruza komanso alangizi.
"Tasks Mail.ru" ndi ntchito yoyamba yodziyimira payokha yosunga mndandanda wazomwe mungachite, wopangidwa ndi kampaniyo. M'miyezi ikubwerayi, Ntchito zidzalowa m'malo mwa mndandanda wa ntchito mu Kalendala ya Mail.ru, ndipo pulojekitiyo itatsegulidwa kwa ogwiritsa ntchito onse, idzaphatikizidwa mu Mail.ru mafoni ndi intaneti Mail.
Pulojekitiyi idakhazikitsidwa pogwiritsa ntchito njira za Offline-zoyamba komanso za Mobile-zoyamba. Ndiye kuti, mutha kugwiritsa ntchito intaneti nthawi iliyonse, kulikonse komanso pachilichonse. Kufikira pa intaneti zilibe kanthu: deta idzasungidwa ndikulumikizidwa. Kuti mumve zambiri, mutha "kukhazikitsa" pulogalamuyi kuchokera pasakatuli, ndipo imawoneka ngati yamba.
Smart food scanner
Mu golosale, sitingathe kudziwa mwachangu ngati chakudya chili choyenera kwa ife kapena ayi, kuti ndi chotetezeka komanso chathanzi bwanji. Zinthu zimakhala zovuta kwambiri ngati munthu ali ndi zoletsa pazakudya, ziwengo zosiyanasiyana, kapena ali pazakudya. Pulogalamu ya Foodwise Android imakupatsani mwayi wosanthula barcode yazinthu ndikuwona ngati kuli koyenera.
gwiritsani ntchito.
Pulogalamuyi ili ndi magawo atatu: "Profile", "Camera" ndi "History".
Mu "Profile" mumayika zomwe mumakonda: mu gawo la "Ingredients" mutha kuchotsera pazakudya zanu zilizonse za 60 zophatikizika m'dawunilodi ndikuwerenga zambiri za E-supplements. "Magulu" amakupatsani mwayi wopatula zosakaniza zonse nthawi imodzi. Mwachitsanzo, ngati mungatchule "Vegetarianism," ndiye kuti zonse zomwe zili ndi nyama zidzawonetsedwa zofiira.
Pali mitundu iwiri mugawo la "Kamera": kusanthula ma barcode ndikuzindikira masamba ndi zipatso. Mukayang'ana barcode, mupeza zonse zokhudzana ndi malonda. Zosakaniza zomwe mudazipatula zidzawonetsedwa zofiira.
Zogulitsa zonse zomwe zidasinthidwa kale zidzasungidwa mu Mbiri. Gawoli lili ndi mawu komanso mawu.
Kuzindikirika kwa zipatso ndi ndiwo zamasamba kumakupatsani mwayi wodziwa zambiri zazakudya komanso mphamvu zake. Mwachitsanzo, apulo imodzi ili ndi pafupifupi magalamu 25.
Zakudya zopatsa mphamvu, zomwe sizovomerezeka kwa anthu omwe amadya zakudya zochepa zama carb.
Ntchitoyi idalembedwa ku Kotlin; "Kamera" imagwiritsa ntchito ML Kit kusanthula ma barcode ndikuzindikira zipatso ndi ndiwo zamasamba. Kumbuyo kumakhala ndi ntchito ziwiri: seva ya API yokhala ndi database,
yomwe imasunga zosakaniza za 60 ndi zolemba za 000, komanso neural network yolembedwa mu Python ndi Tensorflow.
Gulu la polojekiti: Artyom Andryukhov, Ksenia Glazacheva, Dmitry Salman.
Service kugulitsa mphatso ndi augmented zenizeni
Munthu aliyense walandira mphatso zophiphiritsa kamodzi pa moyo wake. Nthawi zambiri, kwa anthu, chidwi chimakhala chofunikira kwambiri kuposa mphatso yomwe amalandira. Mphatso zoterezi sizopindulitsa, koma kupanga ndi kutaya kwawo kumakhudza kwambiri chilengedwe cha dziko lathu lapansi. Umu ndi momwe olemba polojekitiyi adafikira ndi lingaliro lopanga ntchito yogulitsa mphatso ndi zenizeni zenizeni.
Kuti tione kufunika kwa lingalirolo, tinachita phunziro. 82% ya omwe adafunsidwa adakumana ndi vuto losankha mphatso. Kwa 57% ya omwe adafunsidwa, vuto lalikulu pakusankha linali kuopa kuti mphatso zawo sizidzagwiritsidwa ntchito. 78% ya anthu ali okonzeka kusintha kuti athetse mavuto a chilengedwe.
Olembawo apereka mfundo zitatu:
- Mphatso zimakhala m'dziko lenileni.
- Satenga malo.
- Nthawi zonse pafupi.
Kuti akwaniritse zowona zenizeni pa intaneti, olemba adasankha laibulale ya AR.js, yomwe ili ndi magawo awiri akulu:
- Woyamba ali ndi udindo wojambula zithunzi pamwamba pa mtsinje wa kamera pogwiritsa ntchito A-Frame kapena Three.js.
- Gawo lachiwiri ndi ARToolKit, yomwe ili ndi udindo wozindikira cholembera (chinthu chapadera chomwe chitha kusindikizidwa kapena kuwonetsedwa pazenera la chipangizo china) mumayendedwe a kamera. Cholembera chimagwiritsidwa ntchito kuyika zojambulazo. Kukhalapo kwa ARToolKit sikukulolani kuti mupange chowonadi chosazindikirika pogwiritsa ntchito AR.js.
AR.js imabisa misampha yambiri. Mwachitsanzo, kugwiritsidwa ntchito kwake limodzi ndi A-Frame kumatha "kuswa" masitaelo patsamba lonse. Choncho, olembawo adagwiritsa ntchito "mtolo" wa AR.js + Three.js, zomwe zinathandiza kuthetsa mavuto ena. Ndipo kuti tiyike AR.js kutengera Three.js mu React, momwe tsamba la projekitiyo linalembedwa, tidayenera kupanga nkhokwe ya AR-Test-2 (), yomwe imagwiritsa ntchito gawo lina la React pakugwiritsa ntchito AR.js kutengera Three.js. Kuyang'ana kwachitsanzo mu zenizeni zenizeni ndi 3D (zazida zopanda kamera) zidakhazikitsidwa.
Komabe, pambuyo pake zidapezeka kuti ogwiritsa ntchito samamvetsetsa kuti cholembera ndi chiyani komanso momwe angachigwiritsire ntchito. Chifukwa chake, olembawo adasinthira kuukadaulo , yomwe ikupangidwa mwachangu ndi Google. Imagwiritsa ntchito ARKit (iOS) kapena ARCore (Android) kuti ipereke zitsanzo mu AR popanda chikhomo. Tekinolojeyi idakhazikitsidwa pa Three.js ndipo imaphatikizapo chowonera cha 3D. Kugwiritsa ntchito pulogalamuyi kwasintha kwambiri, komabe, kuti muwone zenizeni zenizeni, mufunika chida chokhala ndi iOS 12 kapena mtsogolo.
Ntchitoyi tsopano ikupezeka ku (), komwe mungalandire mphatso yanu yoyamba.
Gulu la polojekiti: Denis Stasyev, Anton Chadov.
Mutha kuwerenga zambiri zamapulojekiti athu a maphunziro pa . Ndipo yenderani tchanelo pafupipafupi , mavidiyo atsopano a maphunziro okhudza mapulogalamu, chitukuko ndi machitidwe ena amawonekera kumeneko nthawi zonse.
Source: www.habr.com