Umgomo weMicrosoft uwukunikeza wonke umuntu amandla nenhlangano emhlabeni ukuze kuzuzwe okwengeziwe. Imboni yezindaba iyisibonelo esihle sokwenza lo msebenzi ube ngokoqobo. Siphila enkathini lapho okuqukethwe okwengeziwe kwakhiwa futhi kudliwe, ngezindlela eziningi nangamadivayisi amaningi. Ku-IBC 2019, sabelane ngezinto ezintsha zakamuva esisebenza ngazo nokuthi zingasiza kanjani ekuguquleni umuzwa wakho wemidiya.
Imininingwane ngaphansi kokunquma!
Leli khasi livuliwe
I-Video Indexer manje isekela ukugqwayiza nokuqukethwe ngezilimi eziningi
Ngonyaka odlule kwa-IBC senze indondo yethu , futhi kulo nyaka kube ngcono nakakhulu. I-Video Indexer ikhipha ngokuzenzakalelayo ulwazi nemethadatha kumafayela emidiya, njengamagama akhulunywayo, ubuso, imizwa, izihloko, nemikhiqizo, futhi awudingi ukuba uchwepheshe wokufunda ngomshini ukuze ukusebenzise.
Iminikelo yethu yakamuva ihlanganisa ukubuka kuqala kwezici ezimbili ezifunwa kakhulu nezihlukanisiwe—ukubonwa kwezinhlamvu ezigqwayizayo nokulotshwa kwenkulumo ngezilimi eziningi—kanye nezengezo ezimbalwa kumamodeli akhona atholakala namuhla ku-Video Indexer.
Ukubonwa Kwezinhlamvu Ezinyakazayo
Okuqukethwe okugqwayizayo kungenye yezinhlobo ezidume kakhulu zokuqukethwe, kodwa amamodeli okubuka ajwayelekile ekhompyutha aklanyelwe ukubona ubuso bomuntu awasebenzi kahle kukho, ikakhulukazi uma okuqukethwe kuqukethe izinhlamvu ezingenazo izici zobuso bomuntu. Inguqulo entsha yokubuka kuqala ihlanganisa i-Video Indexer nesevisi ye-Azure Custom Vision ye-Microsoft, iletha isethi entsha yamamodeli azitholela ngokuzenzakalelayo futhi aqoqe izinhlamvu ezinyakazayo futhi azenze kube lula ukulebula nokuzibona kusetshenziswa amamodeli ombono angokwezifiso ahlanganisiwe.
Amamodeli ahlanganiswe abe yipayipi elilodwa, avumela noma ubani ukuthi asebenzise isevisi ngaphandle kolwazi lokufunda komshini. Imiphumela itholakala ngengosi ye-No-code Video Indexer noma nge-REST API ukuze ihlanganiswe ngokushesha ezinhlelweni zakho zokusebenza.
Sakhe lawa mamodeli ukuze asebenze nabalingiswa abagqwayizayo kanye nabathengi abathile abanikeze okuqukethwe kwangempela okupopayi ukuze kuqeqeshwe futhi kuhlolwe. Inani lomsebenzi omusha lifingqwe kahle ngu-Andy Gutteridge, umqondisi omkhulu wezobuchwepheshe besitudiyo kanye nokukhiqizwa kwangemuva kwe-Viacom International Media Networks, obengomunye wabahlinzeki bedatha: “Ukwengezwa kokutholakala kokuqukethwe okunamandla kwe-AI okunamandla kuzovumela. ukuze sithole ngokushesha nangempumelelo futhi senze ikhathalogi imethadatha yezinhlamvu kokuqukethwe komtapo wethu.
Okubaluleke kakhulu, kuzonikeza amaqembu ethu abadali ikhono lokuthola ngokushesha okuqukethwe akudingayo, kuncishiswe isikhathi esichithwa ekuphatheni imidiya nokubavumela ukuthi bagxile ekudaleni.”
Ungaqala ukujwayelana nokubonwa kwezinhlamvu ezigqwayizayo nge .
Ukuhlonza nokulotshwa kokuqukethwe ngezilimi eziningi
Ezinye izinsiza zemidiya, njengezindaba, umlando kanye nezingxoxo, ziqukethe okuqoshiwe kwabantu abakhuluma izilimi ezahlukene. Amakhono amaningi akhona enkulumo-kuya-umbhalo adinga ulimi oluqaphela umsindo ukuthi lucaciswe kusengaphambili, okwenza kube nzima ukuloba amavidiyo ezilimi eziningi.
Isici sethu esisha Sokuhlonza Ulimi Olukhulunywa Ngokuzenzakalelayo sezinhlobo ezihlukahlukene zokuqukethwe sisebenzisa ubuchwepheshe bokufunda komshini ukuhlonza izilimi ezitholakala kumafa wemidiya. Uma isitholiwe, ingxenye yolimi ngayinye idlula ngokuzenzakalela kunqubo yokubhala ngolimi olufanele, bese wonke amasegimenti ehlanganiswa abe ifayela elilodwa lokulotshwa ngezilimi eziningi.
Okulotshiweyo okuwumphumela kuyatholakala njengengxenye yokuphuma kwe-JSON Yenkomba Yevidiyo nanjengamafayela emibhalo engezansi. Okulotshiweyo okukhiphayo kuhlanganiswe noSesho lwe-Azure, okukuvumela ukuthi useshe ngokushesha izingxenye zolimi ezihlukene kumavidiyo akho. Ukwengeza, ukuloba ngezilimi eziningi kuyatholakala uma usebenza nengosi Yenkomba Yevidiyo, ukuze ukwazi ukubuka okulotshiweyo nolimi oluhlonziwe ngokuhamba kwesikhathi, noma ugxumele ezindaweni ezithile kuvidiyo yolimi ngalunye futhi ubone okulotshiweyo ngezilimi eziningi njengamagama-ncazo njengoba ividiyo idlala. Ungakwazi futhi ukuhumushela umbhalo owamukelwe kunoma yiziphi izilimi ezingu-54 ezitholakalayo ngokusebenzisa iphothali kanye ne-API.
Funda kabanzi mayelana nesici esisha sokubonwa kokuqukethwe ngezilimi eziningi nokuthi sisetshenziswa kanjani ku-Video Indexer .
Amamodeli engeziwe abuyekeziwe nathuthukisiwe
Futhi sengeza amamodeli amasha ku-Video Indexer futhi sithuthukisa akhona, kuhlanganise nalawo achazwe ngezansi.
Amabhizinisi akhiphayo ahlotshaniswa nabantu nezindawo
Sandise amandla ethu akhona okuthola uhlobo lomkhiqizo ukuze sifake amagama nezindawo ezaziwa kakhulu, njenge-Eiffel Tower eParis kanye neBig Ben eLondon. Uma zivela kumbhalo okhiqiziwe noma esikrinini kusetshenziswa i-optical character recognition (OCR), ulwazi olufanele luyengezwa. Ngalesi sici esisha, ungasesha bonke abantu, izindawo, nemikhiqizo evele kuvidiyo futhi ubuke imininingwane mayelana nabo, okuhlanganisa nezikhala zesikhathi, izincazelo, kanye nezixhumanisi zenjini yokusesha ye-Bing ukuze uthole ulwazi olwengeziwe.
Imodeli yokutholwa kozimele yomhleli
Lesi sici esisha sengeza isethi "yamathegi" kumethadatha enamathiselwe kuzimele ngamunye emininingwaneni ye-JSON ukumela uhlobo lwabo lokuhlela (isibonelo, ishothi enkulu, ishothi elimaphakathi, eliseduze, eliseduze kakhulu, amashothi amabili, abantu abaningi. , ngaphandle, endlini, njll.). Lezi zici zohlobo lwamashothi ziwusizo lapho uhlela ividiyo yeziqeshana nama-trailer, noma uma ufuna isitayela esithile sokushutha ngezinjongo zobuciko.
Ukutholwa kohlobo lozimele ku-Video Indexer.
Imbudumbudu yokumepha ye-IPTC ethuthukisiwe
Imodeli yethu yokutholwa kwesihloko inquma isihloko sevidiyo ngokususela ekulotshweni, ukuqashelwa kohlamvu lwe-optical (OCR), nosaziwayo abatholiwe, ngisho noma isihloko singacacisiwe ngokusobala. Senza imephu lezi zihloko ezitholiwe ezindaweni ezine zokuhlukaniswa: i-Wikipedia, i-Bing, i-IPTC, ne-IAB. Lokhu kuthuthukiswa kusivumela ukuthi sifake isigaba sesibili se-IPTC.
Ukusebenzisa lokhu kuthuthukiswa kulula njengokuphinda ukhombise ilabhulali yakho yamanje ye-Video Indexer.
Umsebenzi omusha wokusakaza bukhoma
Ekubukeni kuqala kwe-Azure Media Services, sinikeza nezici ezimbili ezintsha zokusakaza bukhoma.
Ukulotshwa kwesikhathi sangempela okunikwe amandla yi-AI kuthatha ukusakaza bukhoma kuye ezingeni elilandelayo
Usebenzisa i-Azure Media Services ukusakaza bukhoma, manje ungathola umfudlana ophumayo ohlanganisa ithrekhi yombhalo ekhiqizwe ngokuzenzakalelayo ngaphezu kokuqukethwe komsindo nevidiyo. Umbhalo udalwa kusetshenziswa ukulotshwa komsindo kwesikhathi sangempela okususelwe kubuhlakani bokwenziwa. Amasu angokwezifiso asetshenziswa ngaphambi nangemuva kokuguqulwa kwenkulumo-kuya-umbhalo ukuze kuthuthukiswe imiphumela. Ithrekhi yombhalo ihlanganiswe ku-IMSC1, TTML noma i-WebVTT, kuye ngokuthi ihlinzekwa nge-DASH, HLS CMAF noma i-HLS TS.
Umbhalo wekhodi wesikhathi sangempela wamashaneli angu-24/7 OTT
Usebenzisa ama-API ethu e-v3, ungakha, uphathe futhi usakaze iziteshi ze-OTT (phezulu), futhi usebenzise zonke ezinye izici ze-Azure Media Services ezifana nevidiyo ebukhoma lapho kudingeka khona (i-VOD, ividiyo efunwayo), ukupakishwa nokuphathwa kwamalungelo edijithali ( I-DRM, ukuphathwa kwamalungelo edijithali).
Ukuze ubone izinguqulo zokubuka kuqala zalezi zici, vakashela .
Amandla okukhiqiza amaphakheji amasha
Usekelo lwamathrekhi encazelo yomsindo
Okuqukethwe okusakazwa eziteshini zokusakaza kuvame ukuba nengoma yomsindo enezincazelo zomlomo zalokho okwenzekayo esikrinini ngaphezu kwesignali yomsindo evamile. Lokhu kwenza izinhlelo zifinyeleleke kalula ezibukelini ezingaboni kahle, ikakhulukazi uma okuqukethwe kubonwa ngokuyinhloko. Okusha ikuvumela ukuthi uchaze elinye lamathrekhi omsindo njengethrekhi echaza umsindo (AD, incazelo yomsindo), okuvumela abadlali ukuthi benze ithrekhi ye-AD itholakale kubabukeli.
Ifaka imethadatha ye-ID3
Ukubonisa ukufakwa kwezikhangiso noma imicimbi yemethadatha yangokwezifiso kusidlali seklayenti, abasakazi bavame ukusebenzisa imethadatha enesikhathi eshumekwe kuvidiyo. Ngokungeziwe kumamodi wokusayina we-SCTE-35, manje futhi siyasekela , echazwe umthuthukisi wohlelo lokusebenza ukuze isetshenziswe uhlelo lokusebenza lweklayenti.
Ozakwethu beMicrosoft Azure bakhombisa izixazululo zokuphela-to-ekupheleni
yethula i-Bitmovin Video Encoding kanye ne-Bitmovin Video Player ye-Microsoft Azure. Amakhasimende manje angakwazi ukusebenzisa lezi zixazululo zombhalo wekhodi nezokudlala ku-Azure futhi azuze ezicini ezithuthukile ezinjengombhalo wekhodi wezigaba ezintathu, ukusekelwa kwe-codec ye-AV1/VC, imibhalo engezansi yezilimi eziningi, nokuhlaziywa kwevidiyo okuhlanganiswe ngaphambili kwe-QoS, ukukhangisa, nokulandelela ividiyo.
ikhombisa I-User Lifecycle Management Platform ku-Azure. Njengomhlinzeki oholayo wemali engenayo kanye nezixazululo zokuphatha umjikelezo wokuphila kwamakhasimende, i-Evergent isebenzisa i-Azure AI ukusiza abahlinzeki bezokuzijabulisa be-premium bathuthukise ukutholwa nokugcinwa kwamakhasimende ngokwakha amaphakheji esevisi ahlosiwe kanye neminikelo ezindaweni ezibalulekile kumjikelezo wempilo wamakhasimende.
izokhombisa insizakalo yayo yemidiya esekwe emafini ehlakaniphile, i-SRT Hub, esiza amakhasimende ukuguqula ukugeleza komsebenzi kokuphela kuye ekupheleni kusetshenziswa. nokuguqula ukugeleza komsebenzi ngama-Hublets avela kwa-Avid, Telestream, Wowza, Cinegy kanye ne-Make.tv.
isungule uchungechunge lwezinsizakalo zemidiya ezisezingeni lokusakaza endaweni ye-Azure yesathelayithi kanye namakhasimende ayo ezinsiza zemidiya. I-SES izobonisa izixazululo zezinsizakalo zokudlala ezilawulwa ngokugcwele, okuhlanganisa ukudlala okuyinhloko, ukudlala kwendawo, ukutholwa kwesikhangiso nokushintshwa, kanye nombhalo wekhodi wekhwalithi ephezulu wesikhathi sangempela ongu-24x7 weziteshi eziningi ku-Azure.
yenza amathuluzi amafu alula kanye nobuchwepheshe bokuzenzakalela kwesiginesha butholakale ku-Azure. Lokhu kunikezwa kuzokwenza kube lula ezinhlanganweni zemidiya ukwengeza ngokuzenzakalelayo imibhalo engezansi, okuhlanganisa nemibhalo engezansi yolimi lwangaphandle, ekuhambeni kwazo okubukhoma nokungaxhunyiwe ku-inthanethi kwevidiyo ku-Azure.
inkampani yamazwe ngamazwe , inkampani yezinsizakalo zobuchwepheshe, ihlanganise inkundla yayo ye-OTT SaaS i-TEPlay ku-Azure Media Services ukuze ilethe okuqukethwe kwe-OTT okuvela emafini. I-Tata Elxsi futhi ilethe isisombululo sayo sokuqapha ikhwalithi ye-Falcon Eye (QoE) ku-Microsoft Azure, ihlinzeka ngezibalo namamethrikhi ukuze kuthathwe izinqumo.
yenza inkundla yayo yokusakaza itholakale ku-Azure njengokukhululwa kwe-beta. I-Verizon Media Platform isisombululo se-OTT esiphethwe ngokwebhizinisi esihlanganisa i-DRM, ukufakwa kwezikhangiso, izikhathi eziqondene nomuntu ngamunye, ukushintshwa kokuqukethwe okunamandla, nokulethwa kwevidiyo. Ukuhlanganiswa kwenza ukuhamba komsebenzi kube lula, ukusekelwa komhlaba wonke kanye nesikali, futhi kuvule amanye amakhono ayingqayizivele atholakala ku-Azure.
Source: www.habr.com
