VQGAN iyo Buug-gacmeedka Sawirka
VQGAN waxay ku cadaadisaa sawirada isku xidhka calaamado kala duwan oo laga soo qaatay buug kood la bartay, taas oo u ogolaanaysa transformer-ku inuu soo saaro sawiro si la mid ah qaababka luqaddu u soo saaraan qoraalka.
quusid qoto dheer
VQGAN, oo lagu soo bandhigay xaashida 2021 'Taming Transformers for High-Resolution Image Synthesis', waxay isku daraysaa qalab-ku-tiri-ku-tiriseedka autoencoder (VQVAE) oo leh tababbar iska soo horjeeda iyo garashada. Encoder-ku wuxuu sawiraa sawir sixir yar oo muuqaal ah; vector kastaa waxa la jeexjeexay meesha ugu dhow ee buug-codee ee la bartay, dheh, 1024 codes, oo sawirka u beddelaya taxane calaamado ah. Qalab-sameeyaha ayaa dib-u-dhis ku sameeya sawirka calaamadahaas, oo lagu tababaray takoorka GAN iyo luminta garashada si dib-u-dhisku u ekaado mid fiiqan halkii uu ka dhalan lahaa. Sababtoo ah sawiradu hadda waa taxane calaamado kala duwan, transformer autoregressive wuxuu u qaabayn karaa sida luqadda, isaga oo saadaaliya calaamado mid mid ah. VQGAN waxay si caan ah u xoojisay qoraal-u-sawir hore qalab farshaxan markii lagu lammaaniyay hagida CLIP.
Aragtida Farsamada
Hawlgalka xudunta u ah waa qiyaasida vector: soo saarida codeeyaha joogtada ah waxaa lagu badalaa vectors-kooda buuga koodka ee kuugu dhow, oo leh 'toos-dhex-dhexaad ah' qiyaase si uu codeeyaha wali wax u baran karo inkasta oo aan la kala duwanayn. Ku darista midabtakoorka GAN-ku-salaysan ee dusha sare ee autoencoder-ka ayaa ah waxa u ogolaanaya VQGAN inay adeegsato shabaqooda calaamad aad uga yar (tusaale 16x16) marka loo eego VQVAE iyadoo la ilaalinayo textures, taasoo ka dhigaysa qaab beddelka qaabaynta.
Saamaynta Istiraatijiyadeed
Xawaaraha iyo miisaanka
Visual AI wuxuu si otomaatig ah u samayn karaa baadhista, ogaanshaha, iyo sumadaynta hawlaha miisaanka.
Xulashada dhismayaasha
Kooxaha hal-abuurka leh waxay hindise karaan fikradaha si dhakhso leh iyagoo leh dib-u-eegis buugeed yar.
Kooxda iyo socodka shaqada
Hawlgalladu waxay isticmaali karaan calaamadaha muuqaalka iyo muuqaalka kuwaas oo markii hore adkeyd in la farsameeyo.
Mustaqbalka VQGAN iyo Buug-gacmeedka Sawirka Sawirka
VQGAN's calaamo-calaamad gooni ah waxay noqotay aasaaska sawirka calaamooyinka ku salaysan iyo moodooyinka fiidyaha, laga bilaabo MaskGIT ilaa nidaamyada hab-dhaqameedka badan ee isku dara sawirka iyo calaamadaha qoraalka hal beddelka. Cilmi baaristu hadda waxay u riixdaa buug- codeedyo waaweyn, heer-kooban ama-fiirin-la'aan ah oo ka fogaanaya burburka buug-gacmeedka iyo dhinaca moodooyinka midaysan halkaas oo erayo isku mid ah ay ka kooban yihiin sawirro, maqal, iyo luqad, taasoo awood u siinaysa jiil kasta.
Dhaqangelinta Adduunka-dhabta ah
Ku dhejinta sawirka 16x16 grid ee calaamadaha buuga koodka si uu transformerku u qaabeeyo oo dib u soo kiciyo
Isku-dubbaridka VQGAN iyo hagida CLIP si loo abuuro farshaxanimada 'VQGAN+CLIP' AI ee dhabta ah ee fayraska galay 2021
Ku cadaadinta sawirada koodh kooban oo is haysta oo kaydinta hufan ama tababbarka wax-soo-saarka hoose
U adeega sida calaamadeeyaha sawirka gudaha soosaarayaasha ku saleysan calaamada sida MaskGIT iyo transformers multimodal
Khatarta & Dariiqyada Ilaalada
Xuquuqda sawirka iyo ogolaanshaha waxay noqon kartaa khataro sharci ah haddii caddayntu aanay caddayn.
Waxqabadka moodeelku wuu ku kala duwanaan karaa iftiinka, tirakoobka, iyo deegaanka.
Wanaagga beenta ah waxa laga yaabaa inaan la dareemin ilaa xadka kalsoonida aan la kormeerin.
Qorshe Hawleedka Dhaqangelinta
Qeex shuruudaha aqbalida ee saxnaanta, dib u celinta, iyo kharashyada khaladka.
Ku tijaabi xogta ku habboon xaaladaha wax soo saarka dhabta ah.
Ku dar dib u eegis bini'aadamka si aad u hesho kalsoonida hoose ama saameeynta sare.
Lasoco moodeel dhaqaaqa oo dib u cusboonaysii kamarada ama xogta kaydinta ka dib.
Sii wad Sahaminta
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 VQGAN and Codebook Image Synthesis 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
Hagaha xiga
SPADE Semantic Image Synthesis
Su'aalaha soo noqnoqda
What is VQGAN and Codebook Image Synthesis?
VQGAN waxay ku cadaadisaa sawirada isku xidhka calaamado kala duwan oo laga soo qaatay buug kood la bartay, taas oo u ogolaanaysa transformer-ku inuu soo saaro sawiro si la mid ah qaababka luqaddu u soo saaraan qoraalka.
Maxay VQGAN u isticmaashaa inay u matasho sawirka calaamado gooni ah?
VQGAN waxay qiyaasaysaa astaamaha codeeyaha meelaha ugu dhow ee ku jira buug-koodeedka la bartay, isaga oo sawirka u rogaya taxane tusaha koodka gaarka ah.
Muxuu VQGAN ugu darayaa takoore GAN dusheeda VQVAE?
Khasaaraha iska soo horjeeda iyo garashada ayaa u ogolaanaya VQGAN inay dib u dhisto sawiro qafiif ah oo ka soo jeeda calaamado is haysta, kaas oo VQVAE kaligiis u janjeero inuu blursado.
Mar haddii sawirku yahay taxane calaamado ah, noocee ah ayaa soo saara sawirro cusub?
Sababtoo ah sawiradu waxay noqdaan taxane calaamado kala duwan, transformer-ku wuxuu u qaabayn karaa si toos ah, sida soo saarista qoraalka.
Farsamo noocee ah ayaa u oggolaanaysa gradients inay dhex maraan tillaabada xisaabinta ee aan kala duwanayn?
Tiro-ururinta Vector waa mid aan la kala sooci karin, sidaa darteed VQGAN waxay isticmaashaa qiyaasaha gradient-ka tooska ah si ay u tababarto cod-bixiyaha.
Waa maxay isbeddelka fanka AI ee caanka ah ee VQGAN ka caawiyay abuurista 2021?
Isku-xidhka VQGAN iyo hagida CLIP waxa ay soo saartay farshaxanka 'VQGAN+CLIP' ee si weyn loo wadaago kaas oo caan baxay jiilka sawir-qaadista qoraalka ku salaysan.