HejuruUbuyobozi bukurikira
Nigute ushobora guhindura Video uhindura inyandiko hamwe na AI
AI igaragara
Amabwiriza ya AI
Umwandiko-ujya kuri AI utanga animasiyo ya 3D uhereye kumyandikire yanditse nka "umuntu asimbukira hejuru yicyuzi," agatanga urukurikirane rwimyanya ya skeletale aho kuba pigiseli ya videwo.
Umva iki gitaboAI-byavuzwe · ~3 min
Ibikoresho bifitanye isano bivana gufata amashusho muri videwo isanzwe ya terefone. Hamwe na hamwe baha abategura indie, animateurs hamwe nabarezi byihuse umushinga wambere wimiterere yimiterere idafite ikositimu ya mocap cyangwa ingengabihe y'intoki.
Sisitemu-kuri-sisitemu sisitemu itanga umubiri, ntabwo ari pigiseli. Buri kintu gisobanura skeleton: imizi yumwanya hamwe nicyerekezo wongeyeho kuzunguruka kuri buri rugingo, akenshi bigaragarira kumurongo wimiterere nka SMPL. Clip rero ni urusobekerane rwamakadiri kubiranga, niyo mpamvu tekinoroji yo kubyara yubatswe kumashusho nururimi bishobora guhuzwa no kugenda. Uburyo buzwi cyane ni Motion Diffusion Model (MDM), yasohowe nabashakashatsi bo muri kaminuza ya Tel Aviv mu 2022. Itangirira ku rusaku rudasanzwe rumeze nka clip yerekana kandi ikoresha transformateur kugirango ikureho urusaku intambwe ku yindi, iyobowe ninyandiko yashyizwe muri CLIP. Moderi nkiyi yigira kuri datasets ihuza gufata hamwe nibisobanuro byanditse, cyane cyane HumanML3D, amashusho yayo yakuwe ahanini mubikusanyamakuru bya AMASS. Umuryango wa kabiri ufata icyerekezo nkururimi: T2M-GPT na MoMask compress yimuka mubimenyetso byihariye hamwe na VQ-VAE yuburyo bwo kubara, hanyuma bikabyara ibimenyetso nkuko urugero rwururimi rutanga amagambo. Ishami ryihariye risiba inyandiko rwose. Ibikoresho byo gufata AI nka Move.ai, DeepMotion, Rokoko Vision na Wonder Studio (ubu Autodesk Flow Studio) igereranya amashusho ya 3D kuva kuri videwo isanzwe. Batahura 2D umubiri wingenzi muri buri kadamu, ukazamura muri 3D, kandi ugahindura ibisubizo mugihe. Kubikorwa byinshi ibi bisimbuza imyenda yo gufata hamwe na sitidiyo ya marikeri. Muri ibyo bihe byombi, ibisohoka bigomba gusubirwamo: gushushanywa ku nyuguti yerekana amazina yamagufwa, ibipimo hamwe nuburuhukiro butandukanye ninkomoko ya skeleton. Igitekerezo gikocamye ni uko ibyo bikoresho bitanga imikorere yarangiye. Inenge zisanzwe zirimo kunyerera ibirenge, aho ibirenge bishushanya hasi, ibirenge bikarohama hasi, jitter, n'amaboko akomeye, kubera ko imibare myinshi yandika intoki nabi. Amashusho yakozwe mubusanzwe amasegonda make gusa, kandi ibisobanuro birimo ibikorwa birebire cyangwa ibintu bikomeza kuba ingorabahizi. Sitidiyo nyinshi zifata ibisohoka nkinzira yambere animateur atunganya mubikoresho nka Maya, Blender cyangwa MotionBuilder.
Umushinga wumukino wa indie yerekana icyerekezo cyo gukwirakwiza "umusirikare unaniwe acumbagira imbere akicara ku gisanduku," yohereza clip nka FBX, hanyuma ayisubiza mubumwe bwubumwe nkumwanya wa mbere mbere ya animasiyo yanyuma.
Sitidiyo nto ya animasiyo yerekana umukinnyi mucyumba cyo kuraramo hamwe na terefone imwe, ikoresha amashusho binyuze muri serivisi yo gufata amashusho ashingiye kuri videwo, kandi igahanagura skeleton yavuye muri Blender aho gukodesha icyiciro cyo gufata.
Umuhanzi previsualisation ahagarika imirwano mugukora amashusho icumi ngufi-y-amashusho, akayadoda hagati, kandi akerekana umuyobozi igihe nigihe cyo gutegura mbere yuko urufunguzo rwose rushushanywa.
Umushakashatsi wubuvuzi bwumubiri agereranya uko umurwayi agenda, yafashwe kuri videwo y’amavuriro hamwe n’ikigereranyo cyerekana ko atigeze agaragara, bitandukanye n’urugendo rwo kugenda kugira ngo akurikirane impinduka mu mpande zombi mu byumweru.
Ubushakashatsi bugenda bugana kumashusho maremare, inyuguti nyinshi zikorana, icyerekezo cyubaha imbogamizi zumubiri, ninyuguti zikorana nibintu nibintu aho kwimukira mumwanya muto. Imashini yimikino hamwe na animasiyo yamashanyarazi itangiye guhuza ibisekuruza hamwe no gufata amashusho nkibintu byubatswe, bishobora gutuma animasiyo ikabije igera kubarema kure cyane. Imipaka yingenzi isigaye: amakuru yimyitozo ni mato ugereranije namakuru yamakuru, yerekana ubwoko bwinshi bwumubiri, imico nuburyo bwo kugenda, hamwe nubwiza bwamaboko no mumaso bikiri inyuma yumubiri wuzuye. Tegereza animateur kugirango igumane uruhare rwibanze mu kuyobora, gukosora no kongerera imikorere ibikorwa byakozwe.
Ibaruwa y'amakuru y'ubuntu
Inkuru eshatu zizewe za AI buri gitondo cy'iminsi y'akazi, zanditse mu rurimi rworoshye. Ubuntu iteka ryose, nta kwamamaza.
Imeri imwe buri munsi y'akazi. Kwiyimura mu buryo bumwe. Ntitwigera tugurisha cyangwa ngo dusangire aderesi yawe.
Isuzume
Ibisubizo ako kanya kuri buri kibazo, n'icyemezo gishobora gusangirwa gifite ID yagenzurwa umaze gutsinda isomo.
Fasha uburezi bwa AI bw'ubuntu. AI Understanding ni umuryango utegamiye kuri leta wa 501(c)(3) — nta kwamamaza, nta kwishyura, iteka ryose. Tanga impano
Umwandiko-ujya kuri AI utanga animasiyo ya 3D uhereye kumyandikire yanditse nka "umuntu asimbukira hejuru yicyuzi," agatanga urukurikirane rwimyanya ya skeletale aho kuba pigiseli ya videwo. Ibikoresho bifitanye isano bivana gufata amashusho muri videwo isanzwe ya terefone. Hamwe na hamwe baha abategura indie, animateurs hamwe nabarezi byihuse umushinga wambere wimiterere yimiterere idafite ikositimu ya mocap cyangwa ingengabihe y'intoki.
Inyandiko-Kuri-Icyitegererezo cyerekana umubiri usohora ikadiri kumurongo, mubisanzwe umuzi umwanya hamwe nicyerekezo wongeyeho kuzunguruka. Gutanga no gushushanya imiterere bibaho ahandi mumiyoboro.
MDM isobanura transformateur denoiser kumyandiko yashyizwe muri CLIP, ihuza ururimi nubusobanuro bwize kubishusho hamwe ninyandiko zombi.
HumanML3D ihuza ibisobanuro byanditse hamwe na clips yerekana yakuwe ahanini muri AMASS, icyegeranyo kinini cyamakuru asanzwe afata amakuru.
Kuberako umuyoboro usohora imyifatire ifatika, imyitozo irashobora guhana byimazeyo ibisubizo bidashoboka kumubiri, nkibirenge bigenda mugihe bigomba guterwa.
Izi moderi ziga amagambo yikimenyetso hanyuma ikabyara ibimenyetso bikurikirana, nkururimi rwururimi rutanga inyandiko.
Komeza wige
Abandi bayobozi batoranijwe kuriyi ngingo
HejuruUbuyobozi bukurikira
Nigute ushobora guhindura Video uhindura inyandiko hamwe na AI
AI igaragara