Kediska Gelida Qoraalka Nul-Text

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

1. Muxuu rogaal-celisku dhab ahaantii hagaajiyaa?
   • Ku-xidhka qoraalka madhan (null) ee loo adeegsaday hagitaan-free-classifier
   • Miisaanka U-Net faafinta oo dhan
   • pixels ee sawirka gelin si toos ah
   • Codeeyaha sawirka CLIP
2. Waa maxay sababta roganta cad ee DDIM ay u noqoto mid aan la isku halayn karin marka loo isticmaalo tafatirka?
   • Waxa kaliya oo ay ka shaqeysaa sawirada synthetic
   • Tilmaanta xorta ah ee kala-saaruhu waxay kordhisaa dib-u-dhiskeeda dib-u-dhiskeeda
   • Waxay u baahan tahay in dib loo tababaro qaabka
   • Kuma socon karo meel qarsoon
3. Waa maxay hadafka ugu weyn ee rogida sawirka dhabta ah ka hor intaanan tafatirin?
   • Si loo cadaadiyo sawirka kaydinta
   • Si meesha looga saaro xogta badan ee sawirka
   • Si aad u hesho barta bilawga sawaxan/dhammaanad daahsoon oo soo cusboonaysiisa sawirka saxda ah
   • Si loogu beddelo miisaan cawl
4. Ku dhawaad sidee buu u rogitaan qoraal-la'aan ah u maareeyaa wanaajinta guud ahaan hannaanka fidinta?
   • Waxa ay wanaajisaa hal-xidhka la wadaago ee sawirka oo dhan
   • Waxay gabi ahaanba ka boodaa tayaynta
   • Waxa ay wanaajisaa kaliya wakhtiga ugu dambeeya
   • Waxay wanaajisaa isku-xidhka null gaar ah ee marxalad kasta oo beeninaysa
5. Dhanka hoose ee ugu muhiimsan rogaal celinta qoraalka asalka ah waa in ay tahay:
   • Hagaajinta sawir kasta, taasoo ka dhigaysa mid gaabis ah
   • Aan awoodin in la isticmaalo tilmaamaha qoraalka
   • Kaliya waxaa lagu isticmaali karaa 8x8 thumbnails
   • Aan ku habboonayn Faafinta Deggan
6. What does Prompt-to-Prompt use to control edits while retaining parts of an image layout?
   • A trainable 3D Gaussian representation
   • Cross-attention maps associated with prompt tokens
   • A classifier that labels every pixel as a new object
   • A new text encoder trained for each edited image
7. What role does pivotal inversion play in null-text inversion?
   • It replaces the diffusion model with a newly trained classifier
   • It deletes the input image before reconstruction
   • It provides a per-timestep noise trajectory that anchors later optimization
   • It selects a new text prompt without using the image
8. During null-text optimization, which components does the method keep fixed?
   • Only the input image dimensions
   • The model weights and every unconditional embedding at initialization
   • The image caption and all diffusion timesteps
   • The pretrained model weights and the conditional text embedding