Muse Masked Generative Imaging
Muse bụ ihe nlere ederede gaa na onyonyo sitere na Google nke na-ewepụta foto site na imeju ihe onyonyo nkpuchi n'otu oge, na-eme ka ọ dị ngwa ngwa karịa mgbasa nke nzọụkwụ site na nzọụkwụ.
Nchịkọta
It matters because it showed you can get high-quality, well-aligned images without the slow iterative denoising that most generators rely on.
Ime miri emi
Muse na-arụ ọrụ na oghere akara ngosi nke onyonyo. VQGAN a zụrụ azụ na-atụgharị foto ka ọ bụrụ grid nke akara akara integer, dị ka okwu okwu nke ngọngọ ụlọ a na-ahụ anya. N'oge ọzụzụ, a na-ekpuchi akụkụ buru ibu nke akara ngosi ndị a, onye ntụgharị na-amụta ịkọ ha azụ, dabere na ntinye ederede site na ụdị asụsụ oyi kpọnwụrụ (T5-XXL). N'oge ọgbọ, Muse na-amalite site na grid niile nwere mkpuchi ma depụta koodu n'otu oge, na-ebu amụma ọtụtụ akara n'otu nzọụkwụ ma na-emegharị ndị na-enweghị obi ike. Nhazi nke ọkwa abụọ na-ebu ụzọ wepụta grid token dị ala, mgbe ahụ, ụdị mkpebi siri ike na-ejupụta grid mkpebi dị elu. N'ihi na ọtụtụ tokens na-edozi n'otu oge, ụdị paramita 900M na 3B na-emepụta ihe oyiyi 256 ma ọ bụ 512 pikselụ n'ime naanị njuaka na-aga n'ihu.
Nghọta nka nka
Isi aghụghọ bụ ngbanwe ngbanwe na ngbanwe dabere na ntụkwasị obi, nke a na-akpọkarị ụdị ụdị MaskGIT. Kama ịkọ otu token n'otu oge (autoregressive) ma ọ bụ na-agbagha ọtụtụ narị ugboro (mgbasa), Muse na-ebu amụma ihe mkpuchi niile, na-edobe ndị kasị nwee obi ike, ma megharịa ihe ndị ọzọ maka agba na-esote. Iji ihe ngbanwe ederede T5-XXL oyi kpọnwụrụ na-enye nghọta asụsụ siri ike n'efu, yana ịrụ ọrụ na akara ngosi pụrụ iche na-eme ka ihe atụ ahụ tụgharịa uche gbasara onyonyo ka okwu.
Mmetụta atụmatụ
Ọsọ na ọnụ ọgụgụ
Visual AI nwere ike megharịa nyocha, nchọpụta na mkpado ọrụ n'ọtụtụ.
Mee nhọrọ
Otu ndị na-emepụta ihe nwere ike imepụta echiche ngwa ngwa site na ngbanwe akwụkwọ ntuziaka ole na ole.
Team na usoro ọrụ
Ọrụ nwere ike iji onyonyo na akara vidiyo siri ike ịhazi.
Ọdịnihu nke Muse Masked Generative Imaging
Ndozi ndakọrịta nke masked na-atụ aka n'ebe ndị na-emepụta ihe dị oke mma yana ngwa ngwa, nke dị mkpa maka ndezi mmekọrịta yana iji ngwaọrụ. Na-atụ anya ka echiche amụma token ga-ejikọta ya na mgbasa ozi na ụzọ vidiyo na-akpaghasị anya, na iji mee ka eserese ozugbo, eserese na ndezi n'enweghị ihe mkpuchi. Ka tokenizers pụrụ iche na-akawanye mma, onyonyo nwere mkpuchi nwere ike ịgbatị n'ime vidiyo na 3D, ebe ngbanwe ngbanwe nwere ike belata ọnụ ahịa imepụta ọtụtụ okpokolo agba ma ọ bụ nlele.
Mmejuputa n'ezie n'ụwa
Nkà echiche ngwa ngwa na bọọdụ ọnọdụ ebe onye na-ese ihe chọrọ ọtụtụ ọdịiche onyonyo n'ime sekọnd karịa nkeji.
Ihe osise efu, dị ka iwepụ ihe na ime ka ihe nlereanya jupụta mpaghara mkpuchi na gburugburu ya.
Ihe osise na-esetịpụ foto gafere ókèala mbụ ya maka ọkọlọtọ ma ọ bụ akụkụ dị iche iche.
Ndozi na-enweghị ihe mkpuchi, dị ka ịgbanwe agba nkịta ma ọ bụ mbara igwe gaa na ọdịda anyanwụ site na-edezi ngwa ngwa ederede yana degharịa akara ngosi emetụtara.
Ihe ize ndụ & okporo ụzọ nche
Ikike onyonyo na nkwenye nwere ike bụrụ ihe egwu dị n'iwu ma ọ bụrụ na edoghị anya.
Ọrụ nlereanya nwere ike ịdịgasị iche n'ofe ọkụ, igwe mmadụ, na gburugburu.
Enwere ike ghara ịhụ ihe dị mma ma ọ bụrụ na enyochaghị oke ntụkwasị obi.
Map mmejuputa
Kọwaa ụkpụrụ nnabata maka nkenke, icheta, na ụgwọ njehie.
Nwalee na data dabara na ọnọdụ mmepụta n'ezie.
Tinye nyocha mmadụ maka obere obi ike ma ọ bụ amụma mmetụta dị elu.
Sochie ihe nlere anya wee megharịa ka emechara mgbanwe igwefoto ma ọ bụ dataset.
Nọgide na-eme nchọpụta
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Muse Masked Generative Imaging quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Ntuziaka na-esote
Ihe mkpuchi Autoencoders
Ajụjụ a na-ajụkarị
What is Muse Masked Generative Imaging?
Muse bụ ihe nlere ederede gaa na onyonyo sitere na Google nke na-ewepụta foto site na imeju ihe onyonyo nkpuchi n'otu oge, na-eme ka ọ dị ngwa ngwa karịa mgbasa nke nzọụkwụ site na nzọụkwụ. Ọ dị mkpa n'ihi na ọ gosipụtara na ị nwere ike nweta onyonyo dị elu, dabara nke ọma na-enweghị ngwa ngwa ikwupụta nke ọtụtụ ndị na-emepụta ọkụ na-adabere na ya.
Kedu ihe Muse na-ebu amụma n'oge ọgbọ kama ịjụ pikselụ?
Muse na-arụ ọrụ na oghere akara ngosi pụrụ iche, na-ebu amụma ihe onyonyo kpuchie nke ihe tokenizer VQGAN mepụtara kama ịrụ ọrụ na pikselụ ozugbo.
Kedu ihe kpatara Muse ji dị ngwa karịa ụdị mgbasa ozi ọkọlọtọ?
Muse na-ejupụta n'ọtụtụ akara mkpuchi mkpuchi n'otu oge yana ọ chọrọ naanị obere mkpirisi nhazi nhazi, n'adịghị ka usoro mgbasa ozi n'usoro n'usoro.
Ebee ka Muse nwetara nghọta ya maka ngwa ngwa ederede?
Ọgbọ ọnọdụ Muse na ntinye ederede sitere na ụdị asụsụ T5-XXL a zụrụ azụ, na-enye ya nghọta asụsụ siri ike.
Gịnị bụ ọrụ nke remasking dabeere obi ike na Muse?
Mgbe amụma ọ bụla yiri ya gachara, Muse na-ejigide akara kacha nwee ntụkwasị obi ma mekwaa ndị nwere obi ike ka ha megharịa n'ọkwa na-esote.
Kedu ka Muse si ejikwa imepụta onyonyo dị elu?
Muse buru ụzọ wepụta grid token dị ala, mgbe ahụ ụdị mkpebi nke abụọ na-ewepụta grid token nwere mkpebi dị elu.