Namuhla thina khuluma ngamaqembu adlule . Bazoba ababili babo kule habrapost. Esokuqala yi-Labra yokuqala, esakha isixazululo sokuqapha ukukhiqizwa kwabasebenzi. Okwesibili - ngesistimu yokubona ubuso yama-turnstiles.
Isithombe: /unsplash.com
I-Labra izokwandisa kanjani ukukhiqiza
Ukukhula komkhiqizo ezimakethe zaseNtshonalanga kwehlile. Ngu McKinsey, ekuqaleni kweminyaka yawo-2,4 lesi sibalo sasingu-2010%. Kodwa phakathi kuka-2014 no-0,5 lehle laya ku-2%. Abahlaziyi baphawula ukuthi isimo asishintshile kusukela lapho. Kodwa kunombono wokuthi izinhlelo ze-intelligence zokwenziwa zizosiza ukuxazulula inkinga. Ngosizo lwezinhlelo ze-AI, ukukhula komkhiqizo kulindeleke ukuthi kubuyele ku-XNUMX% phakathi neminyaka eyishumi. Ama-algorithms ahlakaniphile azosiza ukwenza imisebenzi evamile ibe ngokuzenzakalela futhi athuthukise izinqubo zomsebenzi.
Ucwaningo kulezi zindawo seluvele lwenziwa ngochwepheshe abavela , onjiniyela ngisho nabameleli . Umbono womshini uzodlala indima ebalulekile ekwandiseni ukukhula kokukhiqiza. Ubuchwepheshe busetshenziselwa ukuhlola ngokuzimela indawo yokusebenza nokusebenza kwabasebenzi. Izixazululo ezinjalo sezivele zenziwa yizinkampani zaseNtshonalanga - isibonelo, и .
Izinkampani zaseRussia nazo zakha izixazululo zokuhlola ukukhiqizwa kwabasebenzi. Isibonelo, i-Labra yokuqala, eyadlula yethu . Onjiniyela benza uhlelo lokugada amavidiyo ngenethiwekhi ye-neural ebona izenzo zabasebenzi bebhizinisi futhi ikwenza kucace kahle ukuthi basisebenzisa kanjani isikhathi sabo sokusebenza.
Indlela uhlelo olusebenza ngayo. I-Labra ingasebenza kunoma iyiphi inkampani esebenza ngomshini noma ngomshini abasebenzi bayo abangaphezu kwabayi-15. Ngosizo lwamakhamera, wenza lokho okubizwa ngokuthi - okungukuthi, ibhala konke okwenzeka ngesikhathi sokushintsha. Ngokuvamile, i-algorithm ibukeka kanje:
- Uhlelo luthwebula isithombe bese lumaka imisebenzi yomsebenzi;
- I-algorithm yokufunda yomshini ihlaziya ividiyo;
- I-algorithm ibe isikhiqiza isithombe sosuku lokusebenza;
- Okulandelayo, izibalo zibalwa ngokuzenzakalelayo;
- I-Labra ikhiqiza umbiko wokugcina onezincomo ezizokhuphula ukuphepha ebhizinisini futhi ithuthukise izinsiza zayo.
Ubani oseqenjini? Isiqalo sinezisebenzi zabantu abayisishiyagalombili: umphathi nomsunguli, onjiniyela ababili, ochwepheshe abathathu bezindinganiso zabasebenzi. Kukhona nomphathi wenkonzo yamakhasimende kanye ne-accountant. Abanye babo bahlanganisa umsebenzi wephrojekthi nezifundo zasenyuvesi. Ngakho-ke, wonke umuntu uqapha ukuqedwa kwemisebenzi neminqamulajuqu ngokuzimela. Kodwa-ke, iqembu libamba imihlangano kabili ngesonto ukuxoxa ngenqubekelaphambili nezinhlelo zentuthuko.
Amathemba. Ekuqaleni kukaSeptemba, ukuqaliswa kwethula iphrojekthi yayo . Lapho, onjiniyela babonisa amakhono omkhiqizo. I-Labra ihlela ukuqhubeka nokukhuthaza isisombululo futhi isebenza ngethemba lokubambisana namabhizinisi ezweni.
I-O.VISION izokusiza ukuthi ukhiphe okhiye namapasi
Ngo-2017, ukubuyekezwa kwe-MIT Technology ukuqashelwa kobuso kubuchwepheshe obuphambili obuyi-10 bempumelelo. Lesi sinqumo ngokwengxenye sasibangelwa ukusetshenziswa okubanzi kwezinhlelo ezinjalo. Ikakhulukazi, bangakwazi ukufaka esikhundleni sezihluthulelo ezijwayelekile futhi badlule lapho bengena esakhiweni - isibonelo, amabhange amaningi aseRussia asevele asebenzise intuthuko efanayo. Abadlali abasha nabo bayavela emakethe, isibonelo, isiqalisi sithuthukisa isisombululo esifanayo . Ithimba lenza isistimu yokufinyelela ngaphandle kokuthinta yama-turnstiles angafakwa emizuzwini engama-30.
Indlela uhlelo olusebenza ngayo. Ukuthuthukiswa kuyi-software ne-hardware complex efakwe endaweni yokuhlola. Isekelwe kumanethiwekhi amahlanu e-neural acubungula amafreyimu angawodwana kusukela kukhamera yesistimu ye-biometric. Ababhali bathi ukucubungula isithombe esisodwa kuthatha ngaphansi kwama-millisecond angu-200 (cishe amafreyimu amahlanu ngomzuzwana). Ithimba libhala wonke ama-algorithms wokuqaphela kanye nezixhumi ezibonakalayo ngokuzimela—abathuthukisi abazisebenzisi izixazululo zobunikazi. Qeqesha amanethiwekhi e-neural usebenzisa .
Ukucubungula idatha kwenzeka endaweni. Le ndlela inyusa ukuphepha kwedatha yomuntu siqu ye-biometric. I-Hardware ifaka ibhodi le-Jetson TX1 elivela ku-Nvidia, elakhelwe amadivayisi azimele. Uhlelo lwebhayomethrikhi futhi luqukethe isekethe edidiyelwe yomklamo wayo wokulawula ama-turnstiles nokuhlanganiswa nawo .
Isithombe: /unsplash.com
Abasebenzi bokuqala. Inhloko yenkampani ithi ukukhethwa kwenziwa ngokuvumelana nesimiso: ukhetho lwe-60 endaweni eyodwa. Le fomethi isivumele ukuthi siqashe abantu abanekhono kakhulu. Njengamanje, abahleli bohlelo abambalwa basebenza kuphrojekthi, abanesibopho sokufunda ngomshini ama-algorithms kanye nekhodi yamasistimu ashumekiwe. Kukhona futhi unjiniyela we-backend, uchwepheshe wezokuphepha kolwazi kanye nomklami. Abanye babasebenzi ngabafundi abahlanganisa umsebenzi neziqu ze-masters.
Amathemba. Izixazululo zanamuhla efakwe efekthri enkulu yekhofi eYurophu. Lo mkhiqizo futhi ulungiselelwa ukwethulwa kwesinye sezikhungo zokuzivocavoca zase-St. Petersburg kanye nase-Polytechnic University. Mhlawumbe esikhathini esizayo i-O.VISION izofakwa eNyuvesi yase-ITMO. Inhloko yenkampani ithi sebevele bexoxisana nezinkampani zaseRussia: Gazprom Neft, Beeline, Rostelecom kanye neRussia Railways. Ngokuzayo, sizongena ezimakethe zangaphandle.
Mayelana namanye amaphrojekthi wokusheshisa:
Izinto eziphathelene nomsebenzi we-ITMO University:
Source: www.habr.com