Modèlu diffusion
Modèle diffusion yi dañuy defar ay nataal ci jàng ni ñuy soppi anam wu bari bruit, di soppi static yu bari yi ci nataal yu leer jéego ci jéego.
Résumé
They power today's leading text-to-image tools like Stable Diffusion, DALL-E, and Midjourney.
Plongeur bu xóot
Benn xeetu diffusion dañu koy tàggat ci ñaari yoon. Ci liggéey bi ñuy yóbbu kanam, dañuy yàq ndànk-ndànk nataal bu sell bi, ci yokk tuuti bruit bu bari ba keroog muy nekk static bu seer. Ginaaw loolu model bi jàng lu wuute: tàmbalee ci bruit, mu seetlu ba noppi dindi tuuti bruit ci jéego bu nekk, baamtu ko fukki-fukki yoon wala téemeeri yoon ba keroog muy génne nataal bu ñaw. Ngir mëna doxal lii, ab bind buy tegtal jéego bu nekk ci denoising, kon "astronaut bu war fas" dafay teg static bi ci nataal boobu. Sistem yu bees yu melni Stable Diffusion dañuy doxal liggéey bii ci barab bu nëbbu buñ komprime du ci pixel yu ñor, moo tax mu gëna gaaw. Buñu ko méngale ak GANs, xeetu diffusion yi ñoo gëna mëna tàggat seen bopp ba noppi di defar bariwaayu mbir yu bari, moo tax ñu romb GANs ci anam wu gëna am solo ci defar nataal yu baax ci 2022.
Gis-gis xarala
Li gëna am solo mooy reso bi du musa defar benn nataal ci benn yoon; Dafay jàng ludul xam bruit biy yokk ci benn jéego. Bu ñuy tàggat, dañuy yokk benn bruit buñ xam ci nataal bu dëggu bi, ba noppi ñu laaj model bi mu xayma bruit boobu; wuute gi mooy njuumti tàggat bi. Ci jamonoy defar, model bi dafay dindi bruit biñu ko séentu, di wane ndànk-ndànk structure bi. Kondisioneeru bind bi dañu koy dugal ci cross-attention, te tegtal bu amul classifier dafay yokk ni prompt bi di doxalee ci génn gi.
njeextalu pexe
dogal yu gëna leer
Daf lay jàppale nga tàqale kàddu yu leer ci wàllu xarala ak làkku fësal njaay.
Njëgg ak budget
Mën nga laaj laaj yu gëna baax ci samp gi balaa ngay dugal xaalis wala sa jotu liggéey.
Ekip ak def liggéey
Ekip yi bokk xam-xam ñoo gëna mëna jël yenn dogal ci wàllu produit, politik ak jàng.
Ëlëgu xeetu diffusion
Diffusion mooy anam wi art bi di doxee leegi ci wàllu nataal, di gëna bari wideo ak audio, ak jumtukaay yu melni Sora diko yokk ba ci yëngu. Push bu mag bi mooy gaawaay: pexe yu melni distillation ak model consistance yi dañu bëgga dagg téemeeri jéego denoising wàcci ba ci benn loxo wala sax benn, suko defee ñu mëna defar ci jamono dëgg. Xaarandil diffusion bi mu yaatu ba ci ay mbiri 3D, jëmmal gëstukat yu melni molecule ak protein, ak soppali buñu mëna saytu bu baax, boole ci nekk lu yomb lool ngir doxal ci telefon yi.
Doxal ci àdduna dëgg
Creating original artwork and images from text prompts in Stable Diffusion, DALL-E, and Midjourney
Inpainting ak outpainting, feesal wala yokk pàcci nataal ci anam wu jaar yoon
Generating video from text in tools like OpenAI's Sora
Nafar ay molecule ak ay protein yu bees ngir gëstu ci gis-gis drog
Risk yi ak balustrade yi
Ekip yu bari mën nañu jëfandikoo benn baat ci anam wu wuute, kon teela leeral yaatuwaayam.
Benchmark yi mën nañu nuru lu am doole waaye performance yi ci àdduna bi duñu tolloo.
Bëgg kalite done ak palaŋu jàngat dafay faral di jur njariñ yu yomba dagg.
Roadmap ngir samp gi
Tàmbaleel ci joxe leeral ci làkk wu leer ci njariñ li nga soxla.
Tannal benn metric bu baax ak benn anam bu baaxul balaa ngay saytu.
Doxal ab pilote bu ndaw ak ay done yu representatif, du ab demo bu leer.
Bindal fi Diffusion Models di jàppale ak fi pexe yu gëna yomba gëna baax.
Weyal di banneexu
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 Diffusion Models 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
Gis bi ci topp
Modèlu diffusion GLIDE
Laaj yi ñuy faral di laaj
What is Diffusion Models?
Modèle diffusion yi dañuy defar ay nataal ci jàng ni ñuy soppi anam wu bari bruit, di soppi static yu bari yi ci nataal yu leer jéego ci jéego. Dañuy dooleel jumtukaayi bind-ci-nataal yi gëna mag tay, lu ci melni Diffusion bu dëgër, DALL-E, ak Midjourney.
Lan mooy xalaat bu mag bi ci ginaaw ni modelu diffusion di defaree ab nataal?
Modèle diffusion dafay jàng ni ñuy delloosi ab bruit, tàmbalee ci bruit bu sell ak denoising jéego par jéego ba nataal bu leer feeñ.
Bu ñuy tàggat, lan la model bi di jàng xam luy waaja am ci jéego bu nekk?
Dañu jox model bi nataal bu bari bruit, mu jàng xayma bruit biñ yokk, suko defee mu mëna dindi bruit bi ci diiru defar bi.
Ak naka la ab mbind mëna amee ci nataal biñ defar?
Kondisioneeru mbind (jaaraleko ci cross-attention ak tegtal) nudges bépp denoising jéego suko defee nataal biy génn méngoo ak laaj bi.
Lan moo waral Diffusion Stable di doxal ci 'barab bu nëbbu'?
Liggéey ci barab bu nëbbu buñ komprime ci barabu pixel yu mat sëkk dafay wàññi bu baax xayma bi boole ci baña yàq kalite bi.
Lan mooy benn ci njariñu xeetu diffusion ci kaw GANs?
Modèle diffusion moytu jeu adversarial bu amul dal ak mode collapse bu GANs, di joxe tàggat bu gëna wóor ak bariwaayu génne.