Abanditsi:  (1) Hoon Kim, Beeble AI, kandi yatanze umusanzu ungana muriyi mirimo;  (2) Minje Jang, Beeble AI, kandi yatanze umusanzu ungana muriyi mirimo;  (3) Wonjun Yoon, Beeble AI, kandi yagize uruhare runini muri iki gikorwa;  (4) Jisoo Lee, Beeble AI, kandi yatanze umusanzu ungana muriyi mirimo;  (5) Donghyun Na, Beeble AI, kandi yatanze umusanzu ungana muriyi mirimo;  (6) Sanghyun Woo, Kaminuza ya New York, kandi yagize uruhare runini muri iki gikorwa.   Icyitonderwa cya Muhinduzi: Iki nigice cya 1 kuri 14 cyubushakashatsi butangiza uburyo bwo kunoza uburyo urumuri nigicucu bishobora gukoreshwa kumashusho yabantu mumashusho ya digitale. Soma ibisigaye hepfo.  Imbonerahamwe   Ibisobanuro na 1. Intangiriro   2. Imirimo ifitanye isano   3. Guhindura urumuri na 3.1. Amajonjora   3.2. Gutegura Ikibazo   3.3. Ubwubatsi   3.4. Intego   4. Multi-Masked Autoencoder Mbere yo guhugura   5. Amakuru   6. Ubushakashatsi   7. Umwanzuro  Umugereka   A. Ibisobanuro birambuye   B. Imigaragarire Yabakoresha   C. Kwerekana Video     D. Ibisubizo Byiyongereye Byibisubizo & Reba  Ibisobanuro   Turashiraho uburyo bwateguwe bwo gushushanya amashusho yumuntu ahuza imyubakire iyobowe na fiziki hamwe nuburyo bwo gutangira amahugurwa. Dushushanyije kuri moderi yerekana Cook-Torrance, twashizeho ubwitonzi igishushanyo mbonera kugirango twigane neza imikoranire-yumucyo. Byongeye kandi, kugira ngo tuneshe imbogamizi zidafite ubuziranenge bwo mu rwego rwo hejuru, twashyizeho ingamba zo kwihugura mbere yo guhugura. Iki gitabo gishya cyerekana uburyo bwiza bwo kwerekana imiterere no kwagura imyitozo dataset ishyiraho igipimo gishya mugusobanura realism.  1. Intangiriro  Kuruhuka birenze igikoresho cyiza; ifungura uburyo butagira ingano bwo kuvuga kandi igafasha guhuza amasomo ahantu hatandukanye (reba Ishusho 1). Iri terambere ryumvikanisha icyifuzo cyacu kavukire cyo kurenga imbogamizi zifatika zumwanya nigihe, mugihe tunatanga ibisubizo bifatika kubibazo bifatika mugukora ibintu. Ihindura cyane cyane mubikorwa (VR) hamwe nukuri kwagutse (AR), aho kumurika byorohereza igihe nyacyo cyo guhuza urumuri, kwemeza ko abakoresha nibintu bya digitale bibana mubisanzwe mubidukikije, bitanga urwego rukurikira rwa telepresence.  Muriyi mirimo, turibanda kumashusho yumuntu. Mugihe umurimo wibanze usaba muburyo budasobanutse gusobanukirwa geometrike, imiterere yibintu, no kumurika, ikibazo kirushijeho kwiyongera mugihe ukemura ibibazo byabantu, bitewe nibiranga umwihariko wuruhu rwuruhu kimwe nuburyo butandukanye hamwe nibintu byerekana ibintu byinshi. imyambaro, imisatsi, nibikoresho. Ibi bintu bikora muburyo bugoye, bikenera algorithms zateye imbere zishobora kwigana imikoranire yoroheje yumucyo hamwe nubuso butandukanye.  Kugeza ubu, uburyo butanga ikizere burimo gukoresha imiyoboro yimbitse yamahugurwa yatojwe kubiri kumashusho meza yo mu rwego rwo hejuru yerekana amashusho hamwe nibiranga imiterere yabyo, biva mubyiciro byoroheje [10]. Imbaraga zambere zegereye inzira yo kumurika nk 'agasanduku kirabura' [45, 48], utiriwe winjira muburyo bwibanze. Nyuma yaho iterambere ryashushanyije igishushanyo mbonera cya fiziki cyerekanwe na fiziki, gikubiyemo kwerekana neza imiterere yimiterere yimiterere na fiziki yo gushushanya [32]. Pandey n'abandi. [34] yatanze igitekerezo cyubwubatsi bwuzuye (TR), nacyo kiyobowe na fiziki, cyangiza ishusho yinjiza mubisanzwe bisanzwe hamwe namakarita ya albedo, kandi ikora ibyerekanwa bishingiye kumiterere ya Phong idasanzwe. Ubwubatsi bwa TR bwabaye icyitegererezo cyo kwerekana amashusho, hamwe nubwubatsi bwa vuba kandi bugezweho bwubaka ku ihame ryayo [23, 31, 52].  Dukurikije uburyo buyobowe na fiziki, uruhare rwacu ruri mu gishushanyo mbonera cyubwubatsi hamwe na gahunda yo kwiyobora ubwayo. Ubwa mbere, imyubakire yacu igenda ihinduka yerekeza kumiterere nyayo igaragara muguhuza uburyo bwa Cook-Torrance bwerekana uburyo bwo kwerekana ibintu [8], byerekana iterambere ryibonekeje ryikigereranyo cya Phong cyihariye [37] gikoreshwa mubwubatsi bwuzuye. Moderi ya Cook-Torrance igereranya ubuhanga imikoranire yumucyo na microfacets yo hejuru, ikabarirwa muburyo butandukanye kandi bugaragara. Icya kabiri, uburyo bwo kwitegura buringaniza inzira yo kwiga birenze mubisanzwe bigoye kubona amakuru yoroheje. Mugusubiramo ibice byashizweho na autoencoder (MAE) [19], turabihuza numurimo wo gutanga ibitekerezo. Izi mpinduka zakozwe kugirango zikemure ibibazo byihariye biterwa niki gikorwa, bituma icyitegererezo cyacu twigira kumibare itemewe kandi inonosora ubushobozi bwayo bwo gukora amashusho yukuri mugihe cyo gutunganya neza. Dukurikije ubumenyi dufite, ni ubwambere dushyira mu bikorwa imyitozo ubwayo yiyobora mbere yimyitozo yihariye.  Mu ncamake, umusanzu wacu ni ebyiri. Ubwa mbere, mugutezimbere uburyo bwo kwerekana ibintu, twashyizeho urwego rushya rwa realism mubisohoka. Icya kabiri, mugukoresha imyigire yigenga, twaguye igipimo cyamakuru yamahugurwa kandi tunoza imvugo yumucyo mubintu bitandukanye byukuri. Hamwe na hamwe, iri terambere ryayoboye uburyo bwa SwitchLight kugirango tugere ku buryo bushya bugezweho mu kwerekana amashusho y’abantu.  Uru rupapuro   munsi ya CC BY-NC-SA 4.0. ruraboneka kuri arxiv

Part of HackerNoon's growing list of open-source research papers, promoting free access to academic material.

AutoEncoder.tech

Research & publications on Auto Encoders, revolutionizing data compression and feature learning techniques.

Auto Encoder's blog

Aya majwi yakozwe mururimi rwumwimerere rwinkuru!

Birebire cyane; Gusoma

Make resilience your competitive advantage

Ibikorwa bishya byakozwe nabashakashatsi ba Beeble basezeranya kuzana umucyo nyawo ku mashusho ya Digital ukoresheje AI

About Author

IBITEKEREZO

HANG TAGS

IYI ngingo YATANZWE MU

Related Stories

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps