Kayayyakin AI JAGORA

Hoton Rubutu-zuwa-Hoto

Hoton shine tsarin rubutu-zuwa hoto na Google wanda ke juya rubutattun bayanin zuwa hotuna na zahiri.

2 min karatuAn sabunta ta ƙarshe

Dubawa

Its headline finding was that a large frozen language model, not a bigger image network, was the biggest driver of quality.

Zurfafa nutsewa

An sanar da shi ta hanyar Google Bincike a cikin 2022, Imagen ya nuna cewa zurfin fahimtar hanzari yana da mahimmanci kamar zana shi da kyau. Madadin maƙallan rubutu na nau'in CLIP, Imagen yana amfani da babban riga-kafi na rubutu (T5-XXL) wanda aka ajiye a daskare, sannan yana ciyar da waɗancan abubuwan haɗar harshe masu wadata a cikin ƙirar yaduwa. Yana haifar da ƙaramin hoto mai girman 64x64 kuma yana amfani da matakai biyu masu girman ƙuduri don haɓakawa zuwa 1024x1024. Har ila yau, ƙungiyar ta gabatar da 'tsayi mai ƙarfi' don kiyaye launuka masu ƙarfi a babban jagora, kuma sun gina DrawBench, ma'auni na ƙididdige ƙididdigewa na gwaji, alaƙar sararin samaniya, da haɗin kai. Sigar baya, Imagen 2 da Imagen 3, dalla-dalla dalla-dalla, fassarar rubutu, da saurin aminci, kuma yanzu suna ba da ikon kayan aikin hoton Google.

Fahimtar Fasaha

Zaɓin fitaccen zaɓi na Imagen shine ƙaddamar da rikodin rubutu maimakon janareta na hoto. T5-XXL, wanda aka horar da shi kawai akan rubutu, yana samar da abubuwan haɗawa waɗanda ke ɗaukar harshe mara kyau, kuma masu binciken sun gano cewa faɗaɗa shi ya inganta daidaita rubutun-rubutu fiye da faɗaɗa samfurin watsawa. An ruɗe tsararru: ƙirar ƙirar tushe tana yin ƙaramin hoto mai ƙaranci, sannan samfuran watsa shirye-shiryen babban ƙuduri suna haɓaka shi a hankali, tare da matsananciyar matsar ƙimar pixel don guje wa sakamakon da aka wanke a ƙarƙashin jagora mai ƙarfi.

Dabarun Tasiri

Gudu da sikelin

Kayayyakin AI na iya sarrafa aiki da bincike, ganowa, da ayyuka masu alama a sikelin.

Gina zaɓuɓɓuka

Ƙungiyoyin ƙirƙira za su iya samar da ra'ayoyi cikin sauri tare da ƙarancin bita da hannu.

Ƙungiya da aikin aiki

Ayyuka na iya amfani da siginar hoto da bidiyo waɗanda a baya suke da wahalar aiwatarwa.

Makomar Hoton Rubutu-zuwa-Hoto

Zuriyar Imagen tana tafiya zuwa ga mafi kyawun ma'anar rubutu a cikin hotuna, saurin saurin biyo baya don fage mai rikitarwa, da yin samfur mai sauri. Yi tsammanin haɗuwa mai zurfi tare da ƙirar harshe don haka tsarin 'dalilan' game da buƙatun kafin zane, tare da alamar ruwa mai ƙarfi kamar SynthID don tabbatarwa. Yayin da yake haɗawa a cikin samfuran Google da Gemini, abin da aka fi mayar da hankali ya koma ga abin dogaro, aminci, tsarar da za a iya sarrafawa maimakon sabon sabon abu.

Aiwatar da Gaskiyar Duniya

Samar da abubuwan gani na tallace-tallace na hoto daga rubutaccen taƙaitaccen bayani ba tare da ɗaukar hoto ba

Ƙirƙirar misalan ra'ayi don ba da labari ko littattafan yara daga jimlolin siffantawa

Samar da izgili na samfur da bambancin yanayi don jerin kasuwancin e-commerce

Nuna tunanin kimiyya ko ilimi, kamar fassarar mai zane da aka bayyana a cikin yare bayyananne

Hatsari & Tsare-tsare

Haƙƙoƙin hoto da yarda na iya zama haxarin doka idan ba a fayyace ba.

Ayyukan samfuri na iya bambanta a ko'ina cikin haske, ƙididdiga, da mahalli.

Ƙarya tabbataccen ƙila ba za a iya lura da shi ba sai dai idan an kula da ƙofofin amincewa.

Taswirar Hanya

1

Ƙayyade ma'auni na karɓa don daidaito, tunowa, da farashi na kuskure.

2

Gwada tare da bayanan da suka dace da ainihin yanayin samarwa.

3

Ƙara bita na ɗan adam don ƙarancin amincewa ko tsinkaya mai tasiri.

4

Bi diddigin ƙirar ƙira kuma sake ingantawa bayan canje-canjen kamara ko saitin bayanai.

Ci gaba da Bincike

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 Imagen Text-to-Image quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

Fara tambayoyi

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

Jagora na gaba

Imagen 2 da Rarraba-Tuned Diffusion

Tambayoyin da ake yawan yi

What is Imagen Text-to-Image?

Hoton shine tsarin rubutu-zuwa hoto na Google wanda ke juya rubutattun bayanin zuwa hotuna na zahiri. Binciken kanun labaransa shine babban samfurin yare daskararre, ba babbar hanyar sadarwar hoto ba, shine babban direban inganci.

Menene binciken mafi tasiri na Imagen?

Imagen ya nuna cewa ƙaddamar da fahimtar harshe ta hanyar T5-XXL ya inganta sakamako fiye da ƙaddamar da samfurin watsawa.

Wanne mai rikodin rubutu Imagen ya dogara da shi?

Imagen yana amfani da babban, rubutun T5-XXL wanda aka riga aka horar da shi kawai, wanda aka adana shi a lokacin horo.

Ta yaya Imagen ya kai babban ƙuduri?

Yana haifar da hoto na 64x64 sannan ya haɓaka ta hanyar samfuran watsa shirye-shiryen babban ƙuduri zuwa 1024x1024.

Menene 'tsari mai ƙarfi' da ake amfani dashi a cikin Imagen?

Madaidaicin madaidaicin madaidaicin madaidaicin ƙimar pixel don haka babban jagora baya samar da hotunan da aka wanke.

Menene DrawBench?

DrawBench alama ce ta Google da aka gabatar tare da Imagen don gwada kirgawa, alaƙar sararin samaniya, da saƙon da ba kasafai ba.